Author: Richard B. Langley

Innovation: Reducing the Jitters
Chip-scale atomic clock.

How a Chip-Scale Atomic Clock Can Help Mitigate Broadband Interference

Small low-power atomic clocks can enhance the performance of GPS receivers in a number of ways, including enhanced code-acquisition capability that precise long-term timing allows. And, it turns out, such clocks can effectively mitigate wideband radio frequency interference coming from GPS jammers. We learn how in this month’s column.

By Fang-Cheng Chan, Mathieu Joerger, Samer Khanafseh, Boris Pervan, and Ondrej Jakubov

INNOVATION INSIGHTS by Richard Langley

THE GLOBAL POSITIONING SYSTEM is a marvel of science and engineering. It has become so ubiquitous that we are starting to take it for granted. Receivers are everywhere. In our vehicle satnav units, in our smart phones, even in some of our cameras. They are used to monitor the movement of the Earth’s crust, to measure water vapor in the troposphere, and to study the effects of space weather. They allow surveyors to work more efficiently and prevent us from getting lost in the woods. They navigate aircraft and ships, and they help synchronize mobile phone and electricity networks, and precisely time financial transactions.

GPS can do all of this, in large part, because the signals emitted by each satellite are derived from an onboard atomic clock (or, more technically correct, an atomic frequency standard). The signals from all of the satellites in the GPS constellation need to be synchronized to within a certain tolerance so that accurate (conservatively stated as better than 9 meters horizontally and 15 meters vertically, 95% of the time), real-time positioning can be achieved by a receiver using only a crystal oscillator. This requires satellite clocks with excellent long-term stability so that their offsets from the GPS system timescale can be predicted to better than about 24 nanoseconds, 95% of the time. Such a performance level can only be matched by atomic clocks.

The very first atomic clock was built in 1949. It was based on an energy transition of the ammonia molecule. However, it wasn’t very accurate. So scientists turned to a particular energy transition of the cesium atom and by the mid-1950s had built the first cesium clocks. Subsequently, clocks based on energy transitions of the rubidium and hydrogen atoms were also developed. These initial efforts were rather bulky affairs but in the 1960s, commercial rack-mountable cesium and rubidium devices became available. Further development led to both cesium and rubidium clocks being compact and rugged enough that they could be considered for use in GPS satellites. Following successful tests in the precursor Navigation Technology Satellites, the prototype or Block I GPS satellites were launched with two cesium and two rubidium clocks each. Subsequent versions of the GPS satellites have continued to feature a combination of the two kinds of clocks or just rubidium clocks in the case of the Block IIR satellites.

While it is not necessary to use an atomic clock with a GPS receiver for standard positioning and navigation applications, some demanding tasks such as geodetic reference frame monitoring use atomic frequency standards to control the operation of the receivers. These standards are external devices, often rack mounted, connected to the receiver by a coaxial cable—too large to be embedded inside receivers.

But in 2004, scientists demonstrated a chip-scale atomic clock, and by 2011, they had become commercially available. Such small low-power atomic clocks can enhance the performance of GPS receivers in a number of ways, including enhanced code-acquisition capability that precise long-term timing allows. And, it turns out, such clocks can effectively mitigate wideband radio frequency interference coming from GPS jammers. We learn how in this month’s column.

“Innovation” is a regular feature that discusses advances in GPS technology and its applications as well as the fundamentals of GPS positioning. The column is coordinated by Richard Langley of the Department of Geodesy and Geomatics Engineering, University of New Brunswick. He welcomes comments and topic ideas. Write to him at lang @ unb.ca.

Currently installed Local Area Augmentation System (LAAS) ground receivers have experienced a number of disruptions in GPS signal tracking due to radio frequency interference (RFI). The main sources of RFI were coming from the illegal use of jammers (also known as personal privacy devices [PPD]) inside vehicles driving by the ground installations. Recently, a number of researchers have studied typical properties of popular PPDs found in the market and have concluded that the effect of PPD interference on the GPS signal is nearly equivalent to that of a wideband signal jammer, to which the current GPS signal is most vulnerable. This threat impacts LAAS or any ground-based augmentation system (GBAS) in two ways:
- Continuity degradation — as vehicles with PPDs pass near the GBAS ground antennas, the reference receivers lose lock due to the overwhelming noise power.
- Integrity degradation — the code tracking error will increase when the noise level in the tracking loop increases.
Numerous interference mitigation techniques have been studied for broadband interference. The interference mitigation methods can be separated according to the two fundamental stages of GPS signal tracking: the front-end stage, in which automatic gain control and antenna nulling/beam forming techniques are relevant, and the baseband stage, where code and carrier-tracking loop algorithms and aiding methods are applicable.

In our current work, the baseband strategy and resources that are practically implementable at GBAS ground stations are considered. Among those resources, we focus on using atomic clocks to mitigate broadband GNSS signal interference. For GPS receivers in general, wide tracking loop bandwidths are used to accommodate the change in signal frequencies and phases caused by user dynamics. Unfortunately, wide bandwidths also allow more noise to enter into the tracking loop, which will be problematic when wideband inference exists. The general approach to mitigate wideband interference is to reduce the tracking loop bandwidth. However, a reference receiver employing a temperature-compensated crystal oscillator (TCXO) needs to maintain a minimum loop bandwidth to track the dynamics of the clock itself, even when all other Doppler effects are removed. The poor stability of TCXOs fundamentally limits the potential to reduce the tracking loop bandwidth. This limitation becomes much less constraining when using an atomic clock at the receiver, especially in the static, vibration-free environment of a GBAS ground station.

Integrating atomic clocks with GPS/GNSS receivers is not a new idea. Nevertheless, the practical feasibility of such integration remained difficult until recent advancements in atomic clock technology, such as commercially available compact-size rubidium frequency standards or, more recently, chip-scale atomic clocks (CSACs). Most of the research using atomic clock integrated GPS receivers aims to improve positioning and timing accuracy, enhance navigation system integrity, or coast through short periods of satellite outages. In these applications, the main function of the atomic clock is to improve the degraded system performance caused by bad satellite geometries. As for using narrower tracking loop bandwidths to obtain better noise/jamming-resistant performance, the majority of work in this area has focused on high-dynamic user environments with extra sensor aiding, such as inertial navigation systems, pseudolites, or other external frequency-stable radio signals. These aids alone do not permit reaching the limitation of tracking loop bandwidth reduction since the remaining Doppler shift from user dynamics still needs to be tracked by the tracking loop itself. Our research intends to explore the lower end of the minimum tracking loop bandwidth for static GPS/GNSS receivers using atomic clocks.

High-frequency-stability atomic clocks naturally reduce the minimum required bandwidth for tracking clock errors (since clock phase random variations are much smaller). We have conducted analyses to obtain the theoretical minimum tracking loop bandwidths using clocks of varying quality. Carrier-phase tracking loop performance under deteriorated C/N₀ conditions (that is, during interference) was investigated because it is the most vulnerable to wideband RFI. The limitations on the quality of atomic clocks and on the receiver tracking algorithms (second- or third-order tracking loop bandwidths) to achieve varying degrees of interference suppression at the GBAS reference receivers are explored. The tracking loop bandwidth reductions and interference attenuations that are achievable using different qualities of atomic clocks, including CSACs and commercially available rubidium receiver clocks, are also discussed in this article.

In addition to the theoretical analyses, actual GPS intermediate frequency (IF) signals have been sampled using a GPS radio frequency (RF) frond-end kit, which is capable of utilizing external clock inputs, connected to a commercially available atomic clock. The sampled IF data are fed into a software receiver together with and without simulated wideband interference to evaluate the performance of interference mitigation using atomic clocks. The wideband interference is numerically simulated based on deteriorated C/N₀. The actual tracking errors generated from real IF data are used to validate the system performance predicted by the preceding broadband interference mitigation analyses.

Signal Tracking Loop and Tracking Error

The carrier-phase tracking phase lock loop (PLL) is introduced first to understand the theoretical connection between the carrier-phase tracking errors and the signal noise plus receiver clock phase errors. A simplified PLL is shown in FIGURE 1 with incoming signals set to zero. In the figure, n(s), c(s), and δθ(s) are receiver white noise, clock phase error or clock disturbance, and tracking loop phase error respectively, with s being the Laplace transform parameter. G(s) is the product of the loop filter F(s) and the receiver clock model 1/s.

FIGURE 1. Simplified tracking loop diagram.

From Figure 1, the transfer functions relating the white noise and clock disturbance to the output can be derived as:
(1)

The frequency response of H(s) is complementary to 1-H(s). Therefore, the PLL tracking performance is a trade-off between the noise rejection performance and the clock disturbance tracking performance.

Total PLL errors resulting from different error sources are presented as phase jitter, which is the root-mean-square (RMS) of resulting phase errors. Equation (2) shows the definition of the standard deviation of phase jitter resulting from the error sources considered in this work:
(2)

where , and are standard deviations of receiver white noise, receiver clock errors, and satellite clock error, respectively, for static receivers.

The standard deviation for each of the clock error sources can be evaluated using the frequency response of the corresponding transfer function and power spectral densities (PSDs). The equations to evaluate the phase error from each error source are:
(3)

where S_rx and S_sv are one-sided PSDs for receiver clock and satellite clock, respectively. B_w is the bandwidth of the tracking loop and T_c is the coherent integration time.

Receiver and Satellite Clock Models

In general, the receiver noise can be reasonably assumed to be white noise with constant PSD with magnitude (noise density) of N₀. However, it is not the case for clock errors. The clock frequency error PSD is usually formulated in the form of a power-law equation and has been used to describe the time and frequency behaviors of the random clock errors in a free running clock:

(4)

where s_y(f) represents the PSD of clock frequency errors and is a function of frequency powers.

The clock phase error PSD can be analytically derived from the frequency PSD equation because the phase error is the time integral of the frequency error:
(5)

where f₀ is the nominal clock frequency. The h coefficients of the clock phase error PSD are the product of the h coefficients from the clock frequency error PSD and the nominal frequency.

We have adopted the PSD clock error models in our work to perform tracking loop performance analysis. The PSD of the CSAC is derived from an Allan deviation figure published by the manufacturer and is shown in FIGURE 2. We took three piecewise Allan deviation straight lines, which are slightly conservative, and converted them to a PSD.

FIGURE 2. Allan deviations for chip-scale atomic clock.

Three PSDs of clock error models are listed in TABLE 1, which represent spectrums of the well known TCXO, the CSAC, and a rubidium standard. Phase noise related h₀ and h₁ coefficients in the CSAC model are assumed to be the same as the TCXO because they can’t be obtained from the Allan deviation figure. The rubidium clock phase noises resulting from h₀ and h₁ coefficients are assumed to be two times smaller than those of the TCXO, and the same model is also used as the satellite clock error model in our tracking loop analysis.

TABLE 1. Coefficients of power-law model.

Theoretical Carrier Tracking Loop Performance

Second- and third-order PLLs are used to study the tracking loop performance. The loop filters for each PLL are given by:
(6)

where F₂(s) and F₃(s) are second- and third-order loop filters respectively. Typical coefficients for the second- and third-order loop filters are a₂ = 1.414; w_o,2 = 4×B_w,2× a₂/[(a₂)²+1]; a₃ = 1.1; b₃ = 2.4; w_o,3 = B_w,3/0.7845. B_w,2 and B_w,3 are the second- and third-order tracking loop bandwidths accordingly.

As stated earlier, three error sources are considered for static receivers. Using the clock error models described earlier, the contribution of different error sources to phase jitter is a function of PLL tracking bandwidth. The resulting phase tracking errors from different error sources are evaluated based on Equation (3) and shown in FIGURE 3.

FIGURE 3. Phase error contribution from different error sources.

The third-order PLL performance using 2-, 1-, 0.5- and 0.1-Hz tracking loop bandwidths were analyzed as a function of C/N₀ and are shown in FIGURES 4 and 5. For each selected bandwidth, three different qualities of receiver clocks were analyzed, and a conventional 15-degree performance threshold was adopted. The second-order PLL performs similarly to the third-order PLL. However, the phase jitter tends to be more biased when the tracking loop bandwidth becomes smaller. This phenomenon will be observed later on using signal data for performance validation. Therefore, only the third-order loop performance analysis is shown in Figures 4 and 5. It is obvious from these two figures that the minimum tracking loop bandwidth for a TCXO receiver PLL is about 2 Hz, and the PLL can work properly only while C/N₀ is above 24 dB-Hz.

FIGURE 4 Tracking loop performance analysis for 2- and 1-Hz loop bandwidth.

FIGURE 5. Tracking loop performance analysis for 0.5- and 0.1-Hz loop bandwidth.

As for the receiver using atomic clocks, CSAC and a rubidium frequency standard in our analysis, the PLL bandwidth can be reduced down to at least 0.1 Hz while C/N₀ is above 15 dB-Hz.

Experimental Tracking Loop Performance

Experimental data were collected at Nottingham Scientiﬁc Limited. The experiment was conducted using a GPS/GNSS RF front end with a built-in TCXO clock. The RF front end also has the capability of accepting atomic clock signals through an external clock input connector to which the CSAC (see Photo) was connected during data collection. All data (using the built-in TCXO clock or the CSAC) were sampled at a 26-MHz sampling rate and at a 6.5-MHz IF with 2-MHz front-end bandwidth and four quantization levels.

A MatLab-coded software defined receiver (SDR) was used to process collected IF samples for tracking loop performance validation. TCXO phase jitters resulting from different tracking loop bandwidths are shown in FIGURE 6 for a typical second-order PLL under a nominal C/N₀, which is about 45 dB-Hz. A 45-degree loss-of-lock threshold was adopted (three times larger than the standard deviation threshold used in an earlier performance analysis). In our work, all code tracking delay lock loops (DLLs) are implemented using a second-order loop filter with 20-millisecond coherent integration time and 0.5-Hz loop bandwidth without any aiding. The resulting phase jitters in the figure become biased when the tracking loop bandwidth is reduced. This observed phenomenon implies that a second-order PLL time response cannot track the clock dynamics when the loop bandwidth approaches the minimum loop bandwidth (where loss of lock occurs).

FIGURE 6. Second-order PLL phase jitter using TCXO.

The same IF data was re-processed by the SDR using the third-order PLL with the same range of tracking loop bandwidths. The resulting phase jitters are shown in FIGURES 7 and 8. There is no observable phase jitter bias before the PLLs lose lock in the figures. These results demonstrate that a third-order PLL performs better in terms of capturing the clock dynamics when the tracking loop bandwidth is reduced close to the limitation. Therefore, only the third-order PLL will be considered further.

FIGURE 7. Third-order PLL phase jitter using TCXO.

FIGURE 8. Third-order PLL phase jitter using CSAC.

The performance of the TCXO PLL can be evaluated from the results in Figure 7. It demonstrates that the minimum loop bandwidth is 2 Hz, which is consistent with the previous analysis shown in figure 4. However, the minimum bandwidth using the CSAC is shown to be 0.5 Hz in Figure 8. This result does not meet the performance predicted by the analysis, which shows that the working bandwidth can be reduced to 0.1 Hz.

Analysis and Tracking Performance under PPD Interference

The motivation of our work, as described earlier, is to improve the receiver signal tracking performance under PPD interference, or equivalently, wideband interference. We carried out a simple analysis first to understand how much signal deterioration a GBAS ground receiver could expect. A 13-dBm/MHz PPD currently available on the market was used to analyze the signal deterioration based on the distance between the PPD and the GBAS ground receiver. A simple analysis using a direct-path model shows that noise power roughly 30 dB higher than the nominal noise level (about -202 dBW/Hz) could be experienced by the GBAS ground receiver if the nearest distance is assumed to be 0.5 kilometers. In this case, any wideband interference mitigation method to address PPD interference has to handle C/N₀ as low as 10 to 15 dB-Hz.

Gaussian distributed white noises were simulated and added on top of the original IF samples, then re-quantized to the original four quantization levels to mimic the PPD interference signal condition. A 20-dB higher noise level was simulated to demonstrate the effectiveness of this signal deterioration technique.

The tracking loop performance using the third-order PLL under low C/N₀ conditions was evaluated using the IF sampling and PPD interference simulation technique just described. The evaluation results show that the minimum PLL bandwidth using the TCXO is still 2 Hz. This result is roughly consistent with a previous analysis showing a 24-dB-Hz C/N₀ limitation using 2-Hz tracking bandwidth. The PLL using the CSAC performs better than that using the TCXO, which is expected.

After raising the noise level 5 dB higher to achieve an average of C/N₀ of 18 dB-Hz, phase jitters using the TCXO exceed the threshold at all bandwidths as shown in FIGURE 9. The same magnitude of noise was also added to the CSAC IF samples. The resulting phase jitters are shown in FIGURE 10, which demonstrates that the minimum bandwidth is 1 Hz for this deteriorated signal condition. Any further increase in noise level will result in loss of lock for PLLs using a CSAC at all tracking bandwidths.

FIGURE 9. Phase jitter using TCXO under 18 dB-Hz C/N₀.

FIGURE 10. Phase jitter using CSAC under 18 dB-Hz C/N₀.

Summary and Future Work

We explored a baseband approach for an effective wideband interference mitigation method in this article. We have presented the theoretical analysis and actual data validation to study the possible improvement of the PLL tracking performance under PPD interference, which has been experienced by LAAS ground receivers.

The limitations of reducing PLL tracking loop bandwidths using different qualities of receiver clocks have been analyzed and compared with the experimental results generated by processing IF samples using an SDR. We conclude that the PLL tracking performance using a TCXO is consistent between theoretical prediction and data validation under both nominal and low C/N₀ conditions. However, the PLL tracking performance using the CSAC was not as good as the analysis prediction under both conditions.

In our future work, to understand the reason for the tracking performance inconsistency using the CSAC, we will carefully examine and evaluate the hardware components in line between the external clock input and the IF sampling chip. In this way, we will exclude the clock performance degradation due to any hardware incompatibility.

Other types of high quality clocks, such as extra-low-phase-noise oven-controlled crystal oscillators and low-phase-noise rubidium oscillators, will also be tested to explore the limitation of PLL tracking bandwidth reduction. If the results using other clocks exhibit good consistency between performance analysis and data validation, it is highly possible that the CSAC clock error model mis-represents the available commercial products.

In our future work, we will also consider simulating PPD interference more closely to the real scenario, by adding analog interference signals on top of GPS/GNSS analog signals before taking digital IF samples.

Acknowledgments

The authors would like to thank the Federal Aviation Administration for supporting the work described in this article. Also, the authors would like to extend their thanks to all members of the Illinois Institute of Technology NavLab and to the collaborators from Nottingham Scientific Limited for their insightful advice. This article is based on the paper “Using a Chip-scale Atomic Clock-Aided GPS Receiver for Broadband Interference Mitigation” presented at ION GNSS+ 2013, the 26th International Technical Meeting of the Satellite Division of The Institute of Navigation held in Nashville, Tennessee, September 16–20, 2013.

Manufacturers

The CSAC used in our tests is a Symmetricom Inc., now part of Microsemi Corp. (www.microsemi.com), model SA.45s. We used a Nottingham Scientific Ltd. (www.nsl.eu.com) Stereo GPS/GNSS RF front end with the MatLab-based SoftGNSS 3.0 software from the Danish GPS Center at Aalborg University (gps.aau.dk).

FANG-CHENG CHAN is a senior research associate in the Navigation Laboratory of the Department of Mechanical and Aerospace Engineering at the Illinois Institute of Technology (IIT) in Chicago. He received his Ph.D in mechanical and aerospace engineering from IIT in 2008. He is currently working on GPS receiver integrity for Local Area Augmentation System (LAAS) ground receivers, researching GPS receiver interference detection and mitigation to prevent unintentional jamming using both baseband and antenna array techniques, and developing navigation and fault detection algorithms with a focus on receiver autonomous integrity monitoring or RAIM.

MATHIEU JOERGER obtained a master’s in mechatronics from the National Institute of Applied Sciences in Strasbourg, France, in 2002, and M.S. and Ph.D. degrees in mechanical and aerospace engineering from IIT in 2002 and 2009 respectively. He is the 2009 recipient of the Institute of Navigation Bradford Parkinson award, which honors outstanding graduate students in the field of GNSS. He is a research assistant professor at IIT, working on multi-sensor integration, on sequential fault-detection for multi-constellation navigation systems, and on relative and differential RAIM for shipboard landing of military aircraft.

SAMER KHANAFSEH is a research assistant professor at IIT. He received his M.S. and Ph.D. degrees in aerospace engineering at IIT in 2003 and 2008, respectively. He has been involved in several aviation applications such as autonomous airborne refueling of unmanned air vehicles, autonomous shipboard landing, and ground-based augmentation systems. He was the recipient of the 2011 Institute of Navigation Early Achievement Award for his contributions to the integrity of carrier-phase navigation systems.

BORIS PERVAN is a professor of mechanical and aerospace engineering at IIT, where he conducts research focused on high-integrity satellite navigation systems. Prof. Pervan received his B.S. from the University of Notre Dame, M.S. from the California Institute of Technology, and Ph.D. from Stanford University.

ONDREJ JAKUBOV received his M.Sc. in electrical engineering from the Czech Technical University (CTU) in Prague in 2010. He is a postgraduate student in the CTU Department of Radio Engineering and he also works as a navigation engineer for Nottingham Scientiﬁc Limited in Nottingham, U.K. His research interests include GNSS signal processing algorithms and receiver architectures.

FURTHER READING

• Authors’ Conference Paper

“Performance Analysis and Experimental Validation of Broadband Interference Mitigation Using an Atomic Clock-Aided GPS Receiver” by F.-C. Chan, S. Khanafseh, M. Joerger, B. Pervan and O. Jakubov in the Proceedings of ION GNSS+ 2013, the 26th International Technical Meeting of the Satellite Division of The Institute of Navigation, Nashville, Tennessee, September 16–20, 2013, pp. 1371–1379.

• Chip-Scale Atomic Clocks

“The SA.45s Chip-Scale Atomic Clock–Early Production Statistics” by R. Lutwak in the Proceedings of the 43rd Annual Precise Time and Time Interval (PTTI) Systems and Applications Meeting, Long Beach, California, November 14–17, 2011, pp. 207–219.

“Time for a Better Receiver: Chip-Scale Atomic Frequency References” by J. Kitching in GPS World, Vol. 18, No. 11, November 2007, pp. 52–57.

“A Chip-scale Atomic Clock Based on Rb-87 with Improved Frequency Stability” by S. Knappe, P.D.D. Schwindt, V. Shah, L. Hollberg, J. Kitching, L. Liew, and J. Moreland in Optics Express, Vol. 13, No. 4, 2005, pp. 1249–1253, doi: 10.1364/OPEX.13.001249.

• Atomic Clocks and GNSS Receivers

“Three Satellite Navigation in an Urban Canyon Using a Chip-scale Atomic Clock” by R. Ramlall, J. Streter, and J.F. Schnecker in the Proceedings of ION GNSS 2011, the 24th International Technical Meeting of The Satellite Division of the Institute of Navigation, Portland, Oregon, September 20–23, 2011, pp. 2937–2945.

“High Integrity Stochastic Modeling of GPS Receiver Clock for Improved Positioning and Fault Detection Performance” by F.-C. Chan, M. Joerger, and B. Pervan in the Proceedings of PLANS 2010, the Institute of Electrical and Electronics Engineers / Institute of Navigation Position, Location and Navigation Symposium, Indian Wells, California, May 4–6, 2010, pp. 1245–1257, doi: 10.1109/PLANS.2010.5507340.

“Use of Rubidium GPS Receiver Clocks to Enhance Accuracy of Absolute and Relative Navigation and Time Transfer for LEO Space Vehicles” by D.B. Cox in the Proceedings of ION GNSS 2007, the 20th International Technical Meeting of the Satellite Division of The Institute of Navigation, Fort Worth, Texas, September 25–28, 2007, pp. 2442–2447.

• Clock Stability

“Signal Tracking,” Chapter 12 in Global Positioning System: Signals, Measurements, and Performance, Revised Second Edition by P. Misra and P. Enge. Published by Ganga-Jamuna Press, Lincoln, Massachusetts, 2011.

“Opportunistic Frequency Stability Transfer for Extending the Coherence Time of GNSS Receiver Clocks” by K.D Wesson, K.M. Pesyna, Jr., J.A. Bhatti, and T.E. Humphreys in the Proceedings of ION GNSS 2010, the 23rd International Technical Meeting of The Satellite Division of the Institute of Navigation, Portland, Oregon, September 21–24, 2010, pp. 2937–2945.

“Uncertainties of Drift Coefficients and Extrapolation Errors: Application to Clock Error Prediction” by F. Vernotte, J. Delporte, M. Brunet, and T. Tournier in Metrologia, Vol. 38, No. 4, 2001, pp. 325–342, doi: 10.1088/0026-1394/38/4/6.

• Tracking Loop Filters and Inertial Navigation System Integration

“Kalman Filter Design Strategies for Code Tracking Loop in Ultra-Tight GPS/INS/PL Integration” by D. Li and J. Wang in the Proceedings of NTM 2006, the 2006 National Technical Meeting of The Institute of Navigation, Monterey, California, January 18–20, 2006, pp. 984–992.

“Satellite Signal Acquisition, Tracking, and Data Demodulation,” Chapter 5 in Understanding GPS: Principles and Applications, Second Edition, E.D. Kaplan and C.J. Hegarty, Editors. Published by Artech House, Norwood, Massachusetts, 2006.

“GPS and Inertial Integration”, Chapter 7 in Global Position System: Theory and Applications, Vol. 2, by R.L. Greenspan. Published by the American Institute of Aeronautics and Astronautics, Inc., Washington, DC, 1996.

• GNSS Jamming

“Know Your Enemy: Signal Characteristics of Civil GPS Jammers” by R.H. Mitch, R.C. Dougherty, M.L. Psiaki, S.P. Powell, B.W. O’Hanlon, J.A. Bhatti, and T.E. Humphreys in GPS World, Vol. 23, No. 1, January 2012, pp. 64–72.

“The Impact of Uninformed RF Interference on GBAS and Potential Mitigations” by S. Pullen, G. Gao, C. Tedeschi, and J. Warburton in the Proceedings of ION GNSS 2012, the 25th International Technical Meeting of the Satellite Division of The Institute of Navigation, Nashville, Tennessee, September 17–21, 2012, pp. 780–789.

“Survey of In-Car Jammers-Analysis and Modeling of the RF Signals and IF Samples (Suitable for Active Signal Cancelation)” by T. Kraus, R. Bauernfeind, and B. Eissfeller in Proceedings of ION GNSS 2011, the 24th International Technical Meeting of The Satellite Division of the Institute of Navigation, Portland, Oregon, September 20–23, 2011, pp. 430–435.
May 1, 2014
Delay in Setting Recently Launched Block IIF Satellite Healthy

A United Launch Alliance Delta IV lifts off from Space Launch Complex-37 with the Air Force’s Global Positioning System (GPS) IIF-5 satellite. This launch marked the 25th Delta IV flight since the first flight in 2002.

The latest GPS Block IIF satellite, IIF-5 or SVN64 (operating as PRN30), was launched on February 21, 2014. Typically, GPS satellites are checked out and made operational within about a month following launch. SVN64 has yet to be set healthy.

The delay is due to an extended navigation test being performed by the GPS master control station. A navigation upload for SVN64 was performed in March with ephemeris and clock data as usual streching weeks in advance. However, unlike with operational satellites, no further updated uploads have been performed. The aging ephermis and clock data gradually becomes less and less accurate as time goes by but should degrade gracefully.

Inquisitive observers will have noticed that the received navigation data from SNV64 changes infrequently. Currently, the navigation data changes once per day with an epoch of 13:00 GPS Time unlike every two hours with operational satellites. And the data fit interval is 26 hours, compared to four hours.

The test is scheduled to run until mid-May.

April 17, 2014
GLONASS Loses Control Again

The GLONASS constellation has suffered a major problem for the second time this month.

On Monday, April 14, eight GLONASS satellites were simultaneously set unhealthy for about half an hour, meaning that most GLONASS or multi-constellation receivers would have ignored those satellites in positioning computations. In addition, one other satellite in the fleet was out of commission undergoing maintenance. This might have left too few healthy satellites to compute GLONASS-only receiver positions in some locations.

The unhealthy status of the satellites was noted in the monitoring information provided by the Roscosmos GLONASS Information-Analytical Centre website. The problem was also reported by The Moscow Times, an English-language daily published in Russia.

April 16, 2014
Innovation: Ground-Based Augmentation
Combining Galileo with GPS and GLONASS

By Mirko Stanisak, Mark Bitter, and Thomas Feuerle

INNOVATION INSIGHTS by Richard Langley

GPS = SAFER FLIGHT. While reviewing material for an article celebrating the 25th anniversary of the launch in February 1989 of the first Block II or operational GPS satellite, I was yet again annoyed by many articles on the Web stating that GPS only became available for civil use after the launch of this satellite. Some sources get closer to the truth when they say that GPS was opened for civil use in 1983, following the shoot-down of the Korean Airlines Flight 007. In fact, GPS was designed to serve the needs of both the military and civil communities from the outset. A government memo from April 1973 clearly states: “Civil user needs should be considered in the design of the spaceborne equipment.”

One of the first demonstrations of the use of GPS for aircraft navigation occurred in July 1983, when a Sabreliner business jet was flown in stages from Cedar Rapids, Iowa, to the Paris Air Show, flying only when a sufficient number of the experimental or Block I satellites were in view. The first standalone GPS receivers certified for aviation use (with Receiver Autonomous Integrity Monitoring or RAIM) became available by the mid-1990s. But already the Federal Aviation Administration had been looking into the development of a system to provide higher accuracies and better integrity than that afforded by standalone receivers. In 1994, the FAA announced the development of the Wide Area Augmentation System, its brand of a system generically known as satellite-based augmentation. Geostationary satellites transmit corrections and integrity information to GPS receivers, permitting GPS use for en route navigation all the way down to traditional Category I approach and landing. CAT I approaches can be flown down to a decision height of 61 meters (200 feet). WAAS was declared operational on July 10, 2003, but enhancements to the system continue. Japan, Europe, and India also have operational SBAS based on GPS.

Ground-based GPS augmentation was first developed for maritime applications with the U.S. Coast Guard’s low-frequency system coming on line in the mid-1990s. Also in the mid-1990s, the FAA began the development of the Local Area Augmentation System, generically known as a ground-based augmentation system (GBAS), to provide aircraft with approach and landing capabilities from CAT I down through CAT II (30-meter or 100-foot decision height) and CAT III (no decision height but certain visual range minima) using a VHF datalink. Initial CAT I systems are being operated at Bremen, Germany, and at Newark Liberty International Airport and Houston George Bush Intercontinental Airport.

While a GPS-based GBAS will definitely offer improved navigation services for aircraft, might these services be even better if the systems were to use satellites from other constellations besides GPS? In this month’s column, we look at a straw-man concept for modifying the GBAS protocols to accommodate multiple constellations and the results of preliminary tests using GPS, GLONASS, and Galileo simultaneously.

“Innovation” is a regular feature that discusses advances in GPS technology and its applications as well as the fundamentals of GPS positioning. The column is coordinated by Richard Langley of the Department of Geodesy and Geomatics Engineering, University of New Brunswick. He welcomes comments and topic ideas. Write to him at lang @ unb.ca.

Ever since the declaration of Full Operational Capability (FOC) of the U.S. Global Positioning System in April 1995, GPS has dominated satellite navigation, especially in aviation applications. By contrast, the Russian GLONASS system cannot be used in western aviation because no approval guidelines exist for GLONASS equipment. Thus GPS has been the de-facto standard in aviation for years.

However, within the last few years, major changes have evolved in the field of GNSS, providing a wide variety of useable satellite navigation systems. The European Union launched its Galileo project, which will provide global multi-frequency services in the near future. China is upgrading its BeiDou system (formerly called Compass) to provide global coverage with more medium-Earth-orbit (MEO) satellites. The operators of GPS and GLONASS have started modernization programs that will enable multi-frequency operations in the future, too. Therefore, a large number of usable satellites and signals from multiple systems will soon be available.

In aviation, almost all phases of flight can be assisted by satellite navigation systems nowadays. The most challenging phase of flight with respect to accuracy, continuity, availability, and integrity is the approach and landing phase. The Ground Based Augmentation System (see FIGURE 1; courtesy of the European Organization for Civil Aviation Equipment) allows precision approaches to be performed using satellite navigation. It uses a VHF data link to broadcast differential GNSS corrections, integrity information, and approach definitions to approaching aircraft. These aircraft combine the differential corrections with their own GNSS measurements, calculate a GBAS-corrected position solution, and determine path deviations based on the selected approach.

FIGURE 1. GBAS principle. (Source: EUROCAE WG 28, ED-114)

From a technical perspective, GBAS can use either GPS or GLONASS for differential corrections. For this, the International Civil Aviation Organization (ICAO) Standards and Recommended Practices (SARPs) include GPS and GLONASS side by side. On the other hand, some standardization documents (for example, those from RTCA) are limited to GPS only, effectively excluding GLONASS from being used in the western world. Nevertheless, Russian GBAS systems provide differential corrections for GPS and GLONASS, and are expected to be certified in Russia in the near future. Additional GNSS such as Galileo or BeiDou are not yet included within these documents, as these systems are not approved for aviation use themselves. This article will focus on how a multi-constellation GBAS with GPS, GLONASS, and Galileo could work.

GBAS installations can provide multiple services for different kinds of operation, based on GNSS L1 corrections only. On the one hand, the differentially corrected positioning service (DCPS) is intended to be a generic service for high accuracy positioning. On the other hand, two different GBAS approach services have been defined. GBAS Approach Service Type C (GAST-C) allows Category I (CAT I) procedures and is already in operation. GAST-D is still under development and will enable precision approaches and landings down to CAT II/III minima once certified. To mitigate all possible hazards, GAST-D will require some additional broadcast messages.

VHF Data Broadcast

The VHF Data Broadcast (VDB) is used to communicate binary GBAS messages to approaching aircraft. It operates in the VHF band (108.025 – 117.975 MHz) and uses time-division multiple access (TDMA) to allow the operation of multiple GBAS ground stations on a single frequency. As shown in FIGURE 2, VDB uses UTC time to have a common time frame. Two frames are transmitted each second, lasting 0.5 seconds each. Within each frame, eight slots with durations of 62.5 milliseconds can be used for transmission. Binary application data is encoded using a differentially encoded eight-phase-shift-keying modulation (D8PSK) and a symbol rate of 10,500 symbols per second. With three bits transmitted per symbol, up to 31,500 bits per second can be transmitted. Each slot can contain up to 222 bytes of binary application data. Usually, only a subset of slots is allocated to a particular ground facility. This way, multiple GBAS ground facilities can share a common VDB frequency.

FIGURE 2. VDB timing structure. (Source: RTCA SC-159, DO-246D)

Within each slot, multiple VDB messages can be transmitted as application data. The coding of information in VDB messages is defined in the RTCA’s GNSS-Based Precision Approach Local Area Augmentation System (LAAS) Signal-in-Space Interface Control Document (ICD) and depends on the VDB message type. (LAAS is the U.S. GBAS.) Currently, message types (MT) 1, 2, 3, 4 and 11 are defined. Figure 2 is derived from this document.

Message Type 1 – MT1. Within VDB Message Type 1, differential corrections based on 100-second smoothing are transmitted. These corrections are required by all GBAS approach services (GAST-C and GAST-D). Aside from the differential corrections, additional information for the first broadcast satellite is transmitted. This includes an ephemeris cyclic redundancy check (CRC), mitigating the effects of wrongly received GNSS navigation data, and the Issue of Data (IOD) flag, indicating the time of applicability for the ephemeris data to be used. To transmit this information for all satellites, the satellite for which differential corrections are transmitted first has to be alternated continuously.

Each MT1 message can contain up to 18 pseudorange- and range-rate corrections for individual satellites. Nevertheless, it is possible to link two consecutive MT1 messages using the Additional Message Flag (AMF). The value of this parameter indicates whether this is a single message (0), or the first (1) or second (3) part of a linked MT1 message. Up to 36 differential corrections can be transmitted using two consecutive VDB time slots with 18 corrections each.

All MT1 measurement blocks must be transmitted at least once per frame. The maximum transmission rate is once per slot for all measurement blocks.

Message Type 2 – MT2. VDB Message Type 2 contains station and integrity parameters such as the coordinates of the reference point to which all differential corrections refer. MT2 messages can include (next to a “core” MT2 message) multiple Additional Data Blocks (ADBs) to transmit information required for different GBAS services. At the moment, the Additional Data Blocks 1, 3, and 4 are defined.

ADB1 contains the maximum distance to the reference point at which the corrections may be used (Dmax) as well as parameters to calculate the remaining risk of incorrect GNSS ephemeris data (Kmd,e). Within ADB3, additional information required for GAST-D is transmitted. ADB4 implements the VDB authentication feature. If this ADB is broadcast by a ground facility, MT2 messages must be transmitted first and contain additional indications about which VDB slots are allocated to the ground facility.

MT2 messages must be transmitted at least each 20th frame, but may be repeated up to once per frame.

Message Type 3 – MT3. The VDB Message Type 3 is a fill message, which is only used in conjunction with the GBAS authentication feature (MT2, ADB4). Among other things, this feature requires a minimum slot occupancy of at least 95 percent. Thus, MT3 messages are broadcast only by ground facilities that support the authentication feature and are completely ignored by airborne GBAS receivers.

Message Type 4 – MT4. With VDB Message Type 4, approach information can be broadcast to approaching aircraft. A pilot can select a specific approach by simply tuning to a given channel number.

Currently, GBAS only uses Instrument Landing System look-alike straight-in approaches called Final Approach Segments (FAS). Each FAS represents one approach. This way, a single GBAS ground facility can provide multiple approaches for all runways of an airport. All approaches must be broadcast at least once per 20 consecutive frames.

Message Type 11 – MT11. The VDB Message Type 11 provides differential corrections in a way very similar to MT1 messages. The main difference is that MT11 corrections are based on 30-second smoothing, which is required for GAST-D service. As for MT1, all MT11 measurement blocks must be transmitted at least once per frame.

Enhancements for GBAS with Galileo

At the moment, the GBAS standardization documents include information on GPS, GLONASS, and SBAS ranging sources. No information on Galileo or other constellations has been added yet. Thus, to include Galileo for GBAS, some Galileo-specific experimental additions to the standards are necessary. These proposed modifications have been made in such a way as to keep as close to the other system standards as possible to preserve consistency. This way, hardly any new functionality is added, but additional satellites can be used. The additional Galileo signals (E5a, E5b, E6) are not used at the moment; however, they might be highly beneficial for multi-frequency applications in the future.

All modifications presented here are purely experimental and will most probably not be exactly the same as those in future standards documents. Nevertheless, they provide a way to test Galileo together with GPS and GLONASS for GBAS on an experimental basis.

Ranging Source ID. The Ranging Source ID uniquely addresses a single satellite. It is used in MT1 and MT11 to transmit the differential corrections and other information for each ranging source. In ICAO Annex 10, Standards and Recommended Practices, the Ranging Source ID is defined for GPS, GLONASS, and SBAS only. To provide Galileo corrections as well, an experimental mapping for Galileo satellites was added; see TABLE 1.

TABLE 1. GBAS Ranging Source IDs.

In this way, up to 36 Galileo satellites can be addressed.

Navigation Data. Galileo provides two different sets of navigation data. The I/NAV data corresponds to the Safety-of-Life (SoL) service and is broadcast on E1 and E5b. The F/NAV data corresponds to the Open Service (OS) and is broadcast on E5a. In order to remain as close as possible to the legacy navigation systems, we selected the I/NAV navigation data for use, as it is broadcast on the E1 frequency and can thus be received with an L1-only GNSS receiver.

The navigation data is primarily used in VDB MT1. For the first transmitted correction in this message, the ephemeris set that shall be used in the aircraft is identified via the Issue of Data (IOD) field. To be consistent with the GPS ephemeris, we used Galileo’s IODnav parameter.

Together with the identification of the navigation data, a CRC parameter is transmitted in MT1 for the first satellite within the differential corrections. This parameter ensures that the receiver as well as the ground facility use identical navigation data for all calculations. The CRC algorithm uses the raw navigation data to generate a distinct CRC value.

For GPS and GLONASS, two ephemeris masks are defined. These masks ensure that only information relevant for GBAS processing are covered by the CRC. For Galileo, a similar mask had to be designed.

Additional Data Blocks in MT2. Within VDB MT2, station parameters and integrity information are transmitted. Some parameters for the over-bounding of possible ephemeris errors are specific to each satellite navigation system.

To extend MT2 to Galileo, parameters for the DCPS, GAST-C, and GAST-D must be added for Galileo. For downward compatibility, these parameters cannot be included in the existing Additional Data Blocks beside the existing parameters. Thus, a new Additional Data Block (ADB5) was defined on an experimental basis. This Additional Data Block is dedicated to Galileo and is structured as shown in TABLE 2. The coding of all values corresponds to the coding of the parameters for the existing systems.

TABLE 2. Additional Data Block 5 in Message Type 2 for Galileo parameters.

Optimized VDB Transmission Scheme

Having available a large number of ranging sources for differential corrections, the VHF VDB is a bottleneck for the transmission of this data. To demonstrate this, we first consider the number of visible satellites that there will be in the future. This leads to construction rules for an optimal VDB transmission scheme, which allows transmitting the maximum number of differential corrections.

Number of Satellites Available. To demonstrate the number of differential corrections enabled by the different systems in the future, we computed the number of visible satellites over a day for a stationary GNSS receiver in Braunschweig, Germany. Even though only four Galileo satellites were in orbit at that time, up to 26 different satellites (GPS, GLONASS, and Galileo) were in view simultaneously. Keeping in mind the preliminary Galileo constellation, it is obvious that more than 30 satellites will be available simultaneously in the future — considering only GPS, GLONASS, and Galileo. Adding BeiDou satellites for GBAS would further boost these numbers.

The broadcast of such a large number of differential corrections is limited by the capacity of the VDB and thus by the number of slots assigned to a GBAS ground facility. The number of assigned slots for a facility should be limited as far as possible to be able to use the same frequency for other GBAS ground facilities. Thus, the available capacity must be used as effectively as possible.

Number of Bytes Required. Each VDB message is framed by a message block header (6 bytes) and the message block CRC (4 bytes).

The length of each message depends on the message type and the amount of information to be transmitted. The resulting length for a message of each type is given in TABLE 3.

TABLE 3. Size of different VDB message types (including message block header and CRC). Variable length message types are dependent on the number of corrections, N.

VDB Constraints. A GBAS ground facility must transmit the VDB data following some constraints. These are:
- MT2 messages (including all Additional Data Blocks required) must be transmitted at least each 20th frame (that is, every 10 seconds).
- If authentication is required, each MT2 message must be transmitted in the first slot assigned to the GBAS ground facility.
- All differential corrections (both MT1 and MT11) must be transmitted at least once in each frame. However, it is possible to split the differential corrections into two adjacent slots using the Additional Message Flags in MT1 and MT11 messages.
- Within each MT1 message, the ephemeris decorrelation parameter (P_eph), the Issue of Data (IOD), and the ephemeris CRC is transmitted for the first satellite in the message. Thus, the first satellite must be alternated in order to broadcast the ephemeris information for all satellites.
- Approach definitions are transmitted in MT4 messages. All MT4 messages must be transmitted within at least each 20th slot.
Based on these constraints, a VDB encoding scheme has been developed, which allows us to fulfill all the requirements listed above while optimizing the number of differential corrections that can be transmitted. Even though it is optimized for GAST-D-like services (including authentication parameters, MT11 messages, and experimental Galileo extensions), it can be used for legacy GAST-C systems, too.

Rules for Optimal VDB Transmission. To fulfill the requirement for the MT2 message to be transmitted first, a complete MT2 message must be transmitted each 20th frame at the beginning of the first slot assigned. If no MT2 message has to be transmitted, an MT4 message is transmitted instead. Thus, all messages are arranged in proper order by three simple rules:
1. MT2 (each 20th frame) or MT4 (otherwise)
2. MT11 (all corrections; can be split into two messages)
3. MT1 (all corrections; can be split into two messages).
Additionally, two more rules must be fulfilled. On the one hand, if supporting the authentication feature, each slot in which the ground facility may transmit VDB data must be filled to at least 95 percent. For this, MT3 null messages may be used to ensure that each slot is filled sufficiently. On the other hand, an additional rule for MT1 messages is necessary if more than three slots are assigned to the GBAS ground facility. In this case, to maximize the number of differential corrections the MT1 messages may be transmitted in the last two assigned slots only. This rule is necessary because the Additional Message Flag is limited to two slots for differential corrections.

Using this transmission scheme, the number of differential corrections is maximized while fulfilling the minimum requirements on the VDB data. Even in case of the maximum number of differential corrections, MT4 approach definitions can still be broadcast. However, in this case, the number of transmittable FAS segments is limited to 19. If more approaches (or different approach types such as Terminal Area Paths (TAPs)) have to be transmitted, the VDB generation scheme must be adapted.

Number of Transmittable Corrections. Using the optimized transmission scheme explained earlier, the number of transmittable corrections can be calculated easily for different numbers of assigned slots for GAST-C as well as for GAST-D services (see TABLE 4).

TABLE 4. Number of differential corrections that can be broadcast.

The exact distribution of VDB messages for the maximum number of differential corrections (18) is shown in FIGURE 3 for an MT1/MT11 configuration and two assigned slots.

FIGURE 3. VDB messages for two slots and 18 satellites (MT1 and MT11).

Experimental Realization of Multi-Constellation GBAS

The experimental GBAS multi-constellation extensions described earlier have been implemented in software for further testing. As these enhancements are purely experimental and might change in the future, we have ensured that these definitions can be changed easily.

Navigation Software. The Institute of Flight Guidance at Technische Universität Braunschweig has been developing an experimental navigation framework for many years. This software, called TriPos, can handle and combine different navigation technologies. TriPos can be used for simulations, post-processing of recorded data, and even for live (online) processing. It is written in C++ and supports various platforms.

The navigation framework can be extended easily. Originally, only GPS was supported within the software, but support for GLONASS and Galileo as well as augmentation systems like SBAS and GBAS were added over the past few years. Additionally, the software handles GNSS data of multiple frequencies internally and can thus be used for multi-constellation and multi-frequency applications. TriPos includes decoders for the binary protocols of most GNSS receivers currently available.

For GBAS research, two components can be simulated using the software. On the one hand, the Ground Facility simulation calculates the differential corrections and provides simulated VDB data. On the other hand, the GBAS receiver simulation emulates the behavior of an airborne GBAS receiver and uses VDB data and GNSS measurements to calculate a GBAS solution. Both simulations can use either recorded data in post-processing or live data for online-processing. This allows complete simulation of GBAS.

Multi-Constellation GBAS Ground Facility Simulation. The GBAS ground facility simulation uses raw binary data from multiple stationary GNSS receivers to calculate binary VDB data. The simulation can be freely configured to process either live or pre-recorded GNSS data. Even though it features all algorithms required by the standards, it does not contain additional monitor algorithms at the moment.

Nevertheless, it can provide a valid VDB signal-in-space (SIS), which can be used by GBAS receivers and simulation tools (such as Eurocontrol’s PEGASUS tool). The ground facility simulation supports legacy GBAS CAT-I (GAST-C) as well as GAST-D (including all additional VDB information required) using GPS and GLONASS. Support for Galileo has been added according to the experimental definitions described earlier. In addition to FAS data blocks, the ground facility simulation is also capable of providing curved approaches using TAP data blocks.

Multi-Constellation Airborne GBAS Receiver Simulation. The GBAS receiver simulation has been used for various GBAS-related projects. It supports GAST-C as well as GAST-D and can be configured flexibly to use GPS, GLONASS, and/or Galileo (using the experimental enhancements as described earlier). For GAST-D, all airborne monitoring algorithms required are present. Thus, the aircraft-specific parameters (for example for the airborne geometry screening) can be configured together with the other parameters.

Flight Trials

The practicability of the multi-constellation GBAS approach has been tested in flight trials. To ensure that all four Galileo satellites were in view and capable of providing valid data during our trials, an orbit prediction tool and the Notice Advisory to Galileo Users (NAGU) service of the European GNSS Service Center (GSC) were used prior to the flight.

The data processing configuration is shown in FIGURE 4 and includes the GBAS simulation components explained earlier. All processing is done in real time while recording all data for later post processing.

FIGURE 4. Schematic data processing for the flight experiments (ground components in orange, airborne components in blue).

Ground Processing. On the ground, two Septentrio AsteRx3 GNSS receivers connected to two roof-top antennas were used. The GNSS receivers were connected to the GBAS ground facility simulation via a network and provided binary GPS, GLONASS, and Galileo raw measurements with an update rate of 2 Hz as well as navigation data. Using this data, the ground facility simulation generated binary VDB data. The GBAS ground facility simulation was configured to generate multi-constellation GAST-D VDB data for a three-slot configuration. All required messages (MT1, MT2 including all required ADBs, MT3, MT4 and MT11) were generated and sent to the telemetry facility via the network.

Telemetry. Official VHF data broadcasts operate in a frequency band between 108 and 118 MHz, which is reserved for authorized aviation applications. However, for our experimental system, an alternative data link was used. The Institute of Flight Guidance operates a full-duplex telemetry system to share data between ground and aircraft. Even though the operating frequencies are different, the telemetry system allows the generated binary VDB data to be transmitted to research aircraft. The airborne telemetry receiver outputs data as if it were a VDB receiver to allow us to switch between a real VDB receiver and the telemetry receiver easily.

Research Aircraft. The Institute of Flight Guidance operates the research aircraft of the Technische Universität Braunschweig. The Dornier Do 128-6 with the call sign D-IBUF (see FIGURE 5) is a twin-engine turboprop aircraft without a pressurized cabin and has been used multiple times for GBAS-related research over the years.

FIGURE 5. Research aircraft D-IBUF (Dornier Do 128-6).

The research aircraft allows us to flexibly integrate experimental equipment for specific flight trials. For the multi-constellation GBAS flights, a JAVAD Delta GNSS receiver (capable of multiple constellations and frequencies), a telemetry receiver, and an experimental cockpit display were installed temporarily.

Airborne Processing. The online GBAS receiver simulator uses GNSS data from the JAVAD Delta GNSS receiver together with the VDB data received via telemetry. The receiver was configured to output raw GPS, GLONASS, and Galileo measurements with an update rate of 10 Hz. The simulator was configured to use this data to calculate a multi-constellation GAST-D solution. Based on the selected approach definition, the resulting information (deviations, distance to threshold, and so on) was displayed in the cockpit using an experimental cockpit display.

Results. The flight test was conducted in the evening of November 6, 2013 (16:52 – 17:58 UTC), at Research Airport Braunschweig (EDVE). We performed five approaches with a 10 nautical mile final segment. The flight path as calculated by the GBAS receiver subsystem is shown in FIGURE 6.

FIGURE 6. Flight trial trajectory. (Map data © OpenStreetMap contributors)

FIGURE 7 shows the number of satellites used for the GBAS receiver simulation, and distinguishes between the different satellite navigation systems used. Up to 22 satellites have been used simultaneously for GBAS processing, including up to 10 GPS satellites, eight GLONASS satellites, and four Galileo satellites.

FIGURE 7. Number of satellites used by the multi-constellation GBAS receiver simulation.

Even though no certified GBAS equipment was used for the flight trials, FIGURE 8 shows the resulting vertical and lateral protection levels (VPL and LPL) of the online multi-constellation GBAS receiver simulation. Both values fluctuate due to the differences between 100- and 30-second smoothing position solutions, which have to be added to the protection levels for GAST-D. Nevertheless, both sets of values remain clearly below the corresponding Alert Limits (FAS Lateral Alarm Limit (FASLAL): 40 meters, FAS Vertical Alarm Limit (FASVAL): 10 meters). A valid GAST-D service was achieved continuously.

FIGURE 8. Vertical and lateral protection levels (VPL and LPL).

FIGURE 9 shows a vertical integrity diagram, commonly known as a Stanford plot, for the integrity of the multi-constellation GBAS simulation. This plot shows the Vertical Protection Level (VPL) as determined by the GBAS receiver simulation against the actual Vertical Position Error (VPE). The Vertical Position Error is a direct measure for the Vertical Navigation System Error (V-NSE). This has been determined using a precise point positioning reference trajectory. Both values are normalized by the current VAL as these values change during the approaches. During the flight, the GBAS online processing ran at a rate of 10 Hz, resulting in 43,670 GAST-D epochs and an availability of 100 percent.

FIGURE 9. Normalized vertical Stanford plot of flight trials (GAST-D using GPS, GLONASS, and Galileo). Color scale indicates number of occurrences.

Of course, these results must not be misinterpreted as a multi-constellation GBAS performance assessment. The ground facility simulation was highly experimental and lacked any kind of long-term analysis. Even the GNSS antennas used do not meet formal requirements. However, aside from a quantitative judgment, these results show the practicability of this multi-constellation GBAS approach on an experimental basis.

Conclusion and Outlook

In this article, experimental extensions to GBAS have been developed to support GPS, GLONASS, and Galileo simultaneously. Based on these extensions, an optimized VDB transmission scheme has been created. In this way, the number of transmittable differential corrections could be maximized. Using flight trials, the multi-constellation GBAS concept has successfully been verified. The experimental airborne GBAS subsystem was able to calculate a valid GBAS solution including GPS, GLONASS, and Galileo satellites continuously.

It has been shown that multi-constellation GBAS is possible from a purely technical perspective. On the other hand, neither operational nor approval aspects for satellite navigation systems other than GPS have been addressed yet. Additionally, further testing would be necessary to ensure the compatibility with legacy GPS-only GBAS equipment. However, in theory, all modifications for Galileo are backward compatible. Nevertheless, it has to be assured that certified GBAS multi-mode receivers only use the GPS part of the VDB data and are not disturbed by additional VDB messages or additional ranging sources, for example. The required tests are planned for the future.

The operational benefit of multi-constellation GBAS systems cannot be foreseen yet. A certification for this will take several years and could only be addressed by the GBAS community after the completion of the GAST-D certification. Most probably, the use of GNSS signals on multiple frequencies could provide a highly improved GBAS service and will allow much more operational benefit. Many of the satellite navigation systems have already introduced additional frequencies, including signals in the protected L5 aviation band. The use of multiple frequencies for satellite navigation in aviation can remove most ionospheric errors effectively and mitigate a major source of uncertainty. Thus, multi-constellation GBAS can just be seen as a preliminary step on the way towards multi-frequency GBAS. The concepts and infrastructure described in this article will serve as a basis for more research in this area.

Acknowledgments

Most of our work on multi-constellation GBAS was done within the research project “Bürgernahes Flugzeug,” which was established in 2009 and is partly funded by the German federal state of Lower Saxony. This is gratefully acknowledged by the authors. Additionally, the authors would like to thank all colleagues involved for constructive discussions and their support. This article is based on the paper “Mulitple Satellite Navigation for the Ground Based Augmentation System” presented at ITM 2014, The Institute of Navigation 2014 International Technical Meeting, held in San Diego, California, January 27-29, 2014.

MIRKO STANISAK is a research assistant at the Institute of Flight Guidance (IFF) at the Technische Universität (TU) Braunschweig in Germany. He received his diploma in mechanical engineering (Dipl.-Ing.) in 2009 from TU Braunschweig.

MARK BITTER holds a Dipl.-Ing. in mechanical engineering from TU Braunschweig and has been employed as a research engineer at TU Braunschweig IFF since 2003.

THOMAS FEUERLE received his Dipl.-Ing. in mechanical engineering in 1997 from TU Braunschweig. He joined the TU Braunschweig IFF in May 1997. Since 2005, he has been the leader of the Air Traffic Management Team at the IFF. In April 2010, he completed his Ph.D. dissertation at TU Braunschweig.

FURTHER READING

• Authors’ Conference Paper

“Multiple Satellite Navigation Systems for the Ground Based Augmentation System,” by M. Stanisak, M. Bitter, and T. Feuerle in Proceedings of ITM 2014, the 2014 International Technical Meeting of The Institute of Navigation, San Diego, California, January 27–29, 2014, pp. 254–264.

• Standards Documents

Aeronautical Communications, Vol. 1, Radio Navigation Aids, Annex 10 to the Convention on International Civil Aviation, International Standards and Recommended Practices, International Civil Aviation Organization, Montreal, Draft Version, May 2010.

GNSS-Based Precision Approach Local Area Augmentation System (LAAS) Signal-In Space Interface Control Document (ICD), DO-246D, RTCA Special Committee 159, Global Positioning Systems, RTCA Inc. Washington, D.C., December 2008.

Minimum Operational Performance Standards for GPS Local Area Augmentation System Airborne Equipment, DO-253C, RTCA Special Committee 159, Global Positioning Systems, RTCA Inc. Washington, D.C., December 2008.

Minimum Operational Performance Specification for Global Navigation Satellite Ground Based Augmentation System Ground Equipment to Support Category I Operations, ED-114, EUROCAE Working Group 28 on Global Navigation Satellite System, European Organisation for Civil Aviation Equipment, Malakoff, France, September 2003.

• GBAS Research and Development

“Conception, Implementation and Validation of a GAST-D Capable Airborne Receiver Simulation” by M. Stanisak, R. Schork, M. Kujawska, T. Feuerle, and P. Hecker in Proceedings of ION GNSS 2012, the 25th International Technical Meeting of the Satellite Division of The Institute of Navigation, Nashville, Tennessee, September 17–21, 2012, pp. 250–257.

“Making the Case for GBAS: Experimental Aircraft Approaches in Germany,” by U. Bestmann, P.M. Schachtebeck, T. Feuerle, and P. Hecker in Inside GNSS, Vol. 1, No. 7, October 2006, pp. 42–45.

“Initial GBAS Experiences in Europe” by A. Lipp, A. Quiles, M. Reche, W. Dunkel, and S. Grand-Perret in Proceedings of ION GNSS 2005, the 18th International Technical Meeting of the Satellite Division of The Institute of Navigation, Long Beach, California, September 13–16, 2005, pp. 2911–2922.

• GPS Use in Aviation

“Aircraft Landings: The GPS Approach,” by G. Dewar in GPS World, Vol. 10, No. 6, June 1999, pp. 68–74.

“GPS in Civil Aviation” by K.D. McDonald in GPS World, Vol. 2, No. 8, September 1991, pp. 52–59.
April 2, 2014
Innovation: A PET Project from Finland
Automating GNSS Receiver Testing

By Sarang Thombre, Jussi Raasakka, Tommi Paakki, Francescantonio Della Rosa, Mikko Valkama, Laura Ruotsalainen, Heidi Kuusniemi, and Jari Nurmi

INNOVATION INSIGHTS by Richard Langley

WE HAVE A CAT. My wife and I do, that is. One with a voracious appetite. She likes to be fed on demand, even at the most inopportune times. Like three o’clock in the morning. No, it doesn’t help to close the bedroom door. Her squeaking (yes, some cats squeak) still wakes us up. I was designated as the one to get up in the night to feed her. Sometimes twice. Each night, every night. That got tiresome (literally) very quickly. Automation came to the rescue. We now have a microprocessor-controlled cat feeder, which rotates food compartments into the feeding position at pre-programmed times. Just fill up one or two of the compartments with “crunchies” before retiring, set the activation time to 3:00 a.m., say, and no more middle-of-the-night squeaking interrupting blissful sleep.

This is just one example of how automation — machines replacing (or supplementing) human activity to perform repetitive, difficult, undesirable, or even humanly-impossible tasks — can affect (and benefit) our everyday lives.

As noted on Wikipedia, two common types of automation are ones that involve feedback control, which is usually continuous and involves making measurements using one or more sensors and computing adjustments to keep the measured variables within a set range, and those that involve sequence control, in which a programmed sequence of discrete operations is performed, often based on system logic. An aircraft autopilot is an example of the former while our cat-feeding machine is an example of the latter. Some systems, such as Earth-orbiting satellites, can involve both types.

Automation applications range from the (now) mundane (such as point-and-shoot cameras, smart phones, home control, and factory assembly lines) to the (now) exotic (such as robots to assist the elderly and the infirm and robots to explore space). Laboratories have also benefited from increasing automation, making rapid clinical and analytical testing, for example, possible.

The testing of GNSS receivers can also benefit from automation. This work typically requires the active participation of humans to initiate, control, monitor, and terminate test cases. These manual operations are often inefficient and inaccurate, rendering the test results unreliable. Furthermore, accessing the internal signals of a receiver at different stages of processing is necessary to pinpoint the exact location of any anomalies. Using traditional black-box testing techniques, it is only possible to test the final outputs of a receiver. In this month’s column, we take a look at an automated test bench for analyzing the overall performance of multi-frequency, multi-constellation GNSS receivers. The system includes a data-capture tool to extract internal process information and controlling software, called the Automated Performance Evaluation Tool or AutoPET, which is able to communicate between all modules of the system for hands-free, one-button-click testing of GNSS receivers. Would my cat appreciate the benefit? Likely not, but GNSS engineers and scientists certainly will.

“Innovation” is a regular feature that discusses advances in GPS technology and its applications as well as the fundamentals of GPS positioning. The column is coordinated by Richard Langley of the Department of Geodesy and Geomatics Engineering, University of New Brunswick. He welcomes comments and topic ideas.

The prototype GNSS receiver developed at the Department of Electronics and Communications Engineering of Tampere University of Technology (TUT), called TUTGNSS, is now in the performance-testing phase. TUTGNSS is a GPS L1/L5 + Galileo E1/E5a dual-frequency dual-constellation receiver jointly developed by TUT and its international partners under two European Union Framework Programme research grants.

During the manual testing of the receiver, it was noticed that the results were often contaminated with errors due to imprecise time-keeping and inconsistent test environments.

It was also strenuous and time consuming to perform repetitive tests over multiple iterations, with decreasing personnel efficiency as the number of iterations increased. The aforementioned problems led to the results being deemed unreliable and unrepeatable. There was thus a need to innovate and automate the testing process and environment. In addition, there was also the need to study the signals as they flowed through the internal signal processing chain, so that the exact location of anomalies could be detected.

Currently, few solutions are available in the commercial and academic domains, which can perform end-to-end fully automated, yet customizable testing of GNSS receivers. A couple of commercial testing tools were recently unveiled, which claim to perform similar automated testing of GNSS receivers. However, these are not fully customizable by the end-user, having the limitation that they can be used only with their parent company’s proprietary signal simulators. Other commercial automated testing tools are available nowadays. However, they are targeted towards electronic systems other than GNSS receivers. It was due to these reasons that we decided to implement an in-house solution. Consequently, we devised the Automated Performance Evaluation Tool (AutoPET), along with a data capture tool.

AutoPET is implemented completely in software (Qt, with C++) and communicates with the receiver under test (RUT) via RS-232 and a National Marine Electronics Association (NMEA) protocol and with a commercial GNSS signal simulator via an RS-232 link. It handles the GNSS test cases with user-defined iterations and other system settings. AutoPET has already been used for making test runs on the TUTGNSS receiver with positive results. It is possible to initiate the overall testing of the receiver with a single button-click and the results are stored in the computer without any human intervention. Test scenarios currently included in the tool’s library are: time-to-first-fix (TTFF), position accuracy, acquisition sensitivity, tracking sensitivity, and reacquisition time. By changing the scenarios in this library, the tool can be used with different simulator models. Another innovative aspect of AutoPET is that it uses multi-threading to perform the receiver testing. Multiple software processing threads are necessary to keep track of the receiver operations and simulator feeds simultaneously, so that an appropriate interrupt can be generated when the receiver has performed the desired operation. This feature is explained in further detail later on.

Data Capture Tool (dCAP) is a hybrid (software-controlled hardware) entity capable of extracting the user-defined internal process data from the different modules (acquisition, tracking, bit decoding, and so on) of the GNSS RUT and stores it in a computer via a 100-Mbps Ethernet link. The dCAP hardware is independent of the receiver module (although implemented on the same softcore) and operates through minimal interference with the receiver operation. This data can then be post-processed to monitor and record the behavior of the receiver and to investigate any anomalies in its intermediate stages. An experimental version of dCAP has already been used to monitor the carrier-to-noise-density ratio (C/N₀), carrier Doppler, and code delay from the internal tracking channels, and the raw GNSS signals in I/Q format entering the baseband processing unit (BPU) of the TUTGNSS receiver from its radio front end.

The benefits of AutoPET over state-of-art approaches are that it is portable (software platform independent), easy to use, suitable for testing most receivers using a variety of simulators (provided each of them can communicate with the outside world using some form of communication protocol), and its operational parameters are easy to modify through an external configuration file. dCAP is designed specifically for the TUTGNSS receiver; however, it can be easily replicated for most experimental embedded system receivers. Once implemented, dCAP offers a clear view of the internal operation of the receiver by accessing intermediate signals between the input and output terminals. The speed and size of data capture are limited only by the type of Ethernet connection and the size of the internal and external memories. Additional details of AutoPET and dCAP are provided in the next two sections of this article, while the third section describes the application of these tools in testing the GPS L1 operation of the TUTGNSS receiver.

Automated Performance Evaluation Tool

AutoPET is a software program developed using the Qt platform and the C++ language, which communicates between the GNSS receiver, signal simulator, and its associated computer through a remote PC that houses AutoPET. The set-up is shown in FIGURE 1. This figure also denotes the different communication protocols used between the different modules.

FIGURE 1. Block schematic of the AutoPET assembly.

At the center is the GNSS receiver, which accepts RF signals from the GNSS signal simulator. These signals represent signals from the sky in accordance with the scenario loaded in the simulator, and therefore represent unidirectional communication. On the other hand, the receiver communicates with the remote PC housing AutoPET using the NMEA-0183 protocol. This is bidirectional communication, as the receiver continuously updates its status via NMEA messages to AutoPET and, in turn, AutoPET sends a response/control command to the receiver. The receiver sends the $GPGGA NMEA message every second, and through reading this message, AutoPET can determine the current status (acquisition, tracking, position fix, and so on) of the receiver.

The TUTGNSS receiver has the capability to perform a cold start to initiate the next test iteration when commanded by AutoPET. For this purpose, we have designed a simple custom message string, which can be identified by the TUTGNSS receiver as a cold-start command. In response, the receiver sends a custom NMEA message, $GPTXT, which identifies that it has successfully performed a cold start. Performing a cold start involves erasing all a priori navigation-related information from the receiver memory. This includes erasing the ephemeris, almanac, and timing information, and ensuring that all satellite tracking is lost.

AutoPET communicates with the GNSS signal simulator through its controlling computer, called the Sim-PC (which runs the control software for the simulator). This communication is bidirectional using a 100-Mbps Ethernet link. The AutoPET library holds the scenario files, through which it remotely controls the simulator. In turn, the Sim-PC returns responses or error messages in the form of Extensible Markup Language (XML) strings to the AutoPET. The communication between the Sim-PC and the simulator is through its proprietary protocols.

AutoPET makes extensive use of multi-threading. The receiver, AutoPET, and the simulator function autonomously of each other and hence are independently controlled using their own processing threads running in parallel. Examples of some processing threads are:
- Thread 1 – monitors the receiver operation through the received NMEA messages. This thread is responsible for identifying, for example, if the receiver achieves a position fix or if it performs a successful cold start.
- Thread 2 – monitors the simulator through the received XML error messages and response messages from the Sim-PC. It is responsible for identifying, for example, if the simulator scenario is successfully set up or if the satellite signals are turned on and off when demanded by the test case.
- Thread 3 – monitors the internal operation of AutoPET itself to check, for example, if a timer has expired or if the user performs any operation on the GUI during the progress of a test.
Each thread generates an internal software interrupt within AutoPET based on which future course of action has to be dynamically determined.

Later in the article, the application of AutoPET for single-frequency, single-constellation operation and testing of the TUTGNSS receiver is described. However, it can just as easily be applied for more complex, multi-frequency, multi-constellation testing. The scenarios are stored in the library of AutoPET, and they can be easily updated without requiring any changes in the tool itself. On the other hand, the receiver operation needs to be updated to perform position fixes with multiple signals and constellations. If the receiver allows updating of its operation mode using software commands, as is the case in TUTGNSS, these commands can also be included within AutoPET.

In the case of TUTGNSS, two configuration settings control the mode of operation and the manner in which it has to be turned on (cold, warm, or hot start) via a 32-bit control word. Table 1 describes the various options and the digital control word bits corresponding to each option. There are eight possible modes of operation that would require three bits to be uniquely represented. However, we have assigned five bits, to accommodate any planned future increase in operating modes. Similarly, there are three ways to turn on the TUTGNSS receiver, and they can be uniquely represented by two bits. Therefore, out of the 32 available bits, only seven bits are currently utilized. The rest of the bits are in reserve for future use. The mode selection bits are in least significant bit positions of the control word. For example, if the receiver should perform a position fix after a warm start using GPS L1 and Galileo E1 signals, the 32 bit control word would be 00000000_0000000_00000000_00100010. Using this control word at the beginning of every test, AutoPET can be used for a simple single constellation or more advanced multi-constellation testing of the receiver.

TABLE 1. Control words for multi-frequency, multi-constellation testing of TUTGNSS.

Data Capture Tool

The overall set up of dCAP is shown in FIGURE 2. The TUTGNSS receiver consists of the radio front end and the BPU implemented on an Altera Stratix-II development board. This board consists of the NIOS-II softcore embedded processor controlled by the MicroC operating system within a field-programmable gate array (FPGA) board. The hardware is programmed using VHSIC Hardware Description Language (VHDL) and consists of the system entity and a few peripheral entities, such as a phase-locked loop (PLL), which are not shown in the figure for sake of simplicity. The system entity consists of (among others) two software-controlled hardware entities, one for the TUTGNSS receiver BPU and the other for the dCAP server, called CPU-0 and CPU-1 respectively. The Control-PC is responsible for the overall programming of the FPGA board through a USB link. It also holds a Qt-based user interface acting as the dCAP client implementation.

FIGURE 2. Overall block schematic of the dCAP assembly.

The dCAP client (in the Control-PC) establishes an Ethernet connection with the dCAP server (on the FPGA) and requests a user-specified internal data sample. As an example, let us assume the user requests raw I/Q samples input to the TUTGNSS BPU from the radio front end. The dCAP server software communicates with the TUTGNSS software, which in turn allows the dCAP server hardware access to the requested data from the appropriate region of the TUTGNSS hardware, similar to how a signal across a resistor on a dense printed circuit board is viewed by placing oscilloscope probes across it. The only limitation with dCAP is that the user has to predict, in advance, which internal data parameters are of interest and create access points within the correct hardware entities. The dCAP server hardware will connect to the respective access point when demanded by the client.

This data snapshot is first buffered in the local shared memory entity on the FPGA board due to the requirements of speed, size, and time synchronization. The dCAP server software is responsible for transferring this data from the internal memory to the Control-PC through the Ethernet link. The data is stored on the Control-PC hard drive in the form of a *.bin file. Therefore, the size of each data-packet that can be accessed at a time is limited by the size of the FPGA memory entity, while the total data size is limited only by the size of the hard drive of the Control-PC. The speed of data capture is restricted by the maximum speed of the Ethernet link between the dCAP client and server.

In FIGURE 3, the internal operation of the dCAP server is illustrated, assuming that we would like to access the raw samples from the radio front end. The first block that the samples enter inside the TUTGNSS BPU is the baseband converter unit (BCU). This is where the dCAP hardware probes listen in on the signal samples. Through these probes, the signals are diverted to the first-in-first-out (FIFO) data collector on the dCAP server (CPU-1) in addition to their usual route through the further baseband processing blocks of the receiver. After the FIFO collector, the data undergoes clock arbitration, time synchronization, and master-slave synchronization, before being buffered into the on-chip Synchronous Dynamic Random Access Memory (SDRAM), where it waits until the dCAP server transfers it through the Ethernet-based local network to the requesting dCAP client within the Control-PC. In the case where different internal data has to be monitored, the probes simply reorient to the correct access point within the correct hardware entity (for example, to monitor the signal C/N₀, the probes access the tracking loops).

FIGURE 3. Block schematic of an example of the dCAP internal operation.

TUTGNSS Receiver Performance Testing

During the GPS L1 performance testing of the TUTGNSS receiver, the reference receiver position in the simulator was set randomly. Ionosphere and troposphere errors were turned off in the simulator. On average, 100 iterations were performed for each test, and the total duration to complete all tests was two weeks. dCAP was used in monitoring the tracking channels and extracting information such as the C/N₀, carrier Doppler, and code-delay estimates for the satellites being tracked. Access to these parameters enabled testing the acquisition and tracking sensitivity of the TUTGNSS receiver, thus confirming the results of the tests performed using AutoPET.

Acquisition Sensitivity. Acquisition sensitivity for the TUTGNSS receiver was measured to be -141.5 dBm via AutoPET and -141 dBm via dCAP. Each coherent integration interval was 4 milliseconds, and 256 such intervals were integrated non-coherently. Using AutoPET, 100 acquisition iterations were performed at every power level, and the average number of satellites acquired was recorded. It was observed that no satellites were acquired at -142 dBm. The acquisition sensitivity test using dCAP involved extracting the carrier Doppler and code-delay estimates. A successful acquisition was assumed only if the code-delay estimate error was less than ±1 chip (300 meters) and the carrier Doppler estimate error was less than ±150 Hz. Based on these criteria, 96.72% of acquisitions were found to be successful when the satellite power was maintained at -141 dBm in the simulator as shown in the histograms in FIGURES 4 and 5.

FIGURE 4. Code-delay estimate within ±1 chip (300 meters).

FIGURE 5. Carrier Doppler estimate within ±150 Hz.

Tracking Sensitivity. Tracking sensitivity for the TUTGNSS receiver was measured to be -151 dBm via both tools, assuming a coherent integration interval of 20 milliseconds. Using AutoPET, 100 tracking iterations were performed at every power level and the average number of satellites tracked was recorded. Using dCAP, this test was performed by selecting one satellite and observing how the receiver C/N₀ tracked this satellite during high and low signal power conditions. Twenty tracking iterations of 90 seconds each were performed for a particular satellite. In each iteration, the satellite power in the simulator was maintained at the nominal condition of -130 dBm (equivalent to 38 dB C/N₀ in the receiver) for the first 30 seconds. Subsequently, the power of the satellite was dropped to -151 dBm (equivalent to 17 dB C/N₀ in the receiver).

As visible in Figure 6, the receiver was able to continue tracking the satellite at -51 dBm in 19 out of the 20 iterations. In the case where tracking was lost, the C/N₀ can be seen to diverge rapidly to 0. To make sure that in the rest of the 19 cases the receiver was really tracking the satellite at low power, the power of the satellite was increased again after an additional 30 seconds. In each of the 19 cases, the receiver successfully continued to track the satellite.

FIGURE 6. Tracking C/N₀ in one tracking channel using dCAP.

3D Position Accuracy and TTFF. Computation of the position fix was performed using a least-squares algorithm without any filtering. Using only AutoPET, 100 position fix iterations were performed and the average 3D error in meters was computed. Within the same test case, the time for achieving a position fix was also recorded. The initial (0–30 seconds) position fix estimates are not very accurate. This is because only the first four acquired satellites are used for the position computation. As more satellites are acquired and tracked, their inclusion into the computation gradually improves the position accuracy to within 1 meter. The average TTFF was computed to 60.59 seconds.

Validity of C/N₀ Estimator. FIGURE 7 presents a comparison of C/N₀ measurements between the TUTGNSS receiver (extracted using dCAP) and a commercial receiver. The input power from the simulator was varied between -130 dBm and -151 dBm with steps of around 2 dB for 10 seconds each. The C/N₀ readings from the two receivers were measured at each power level and plotted on the same scale. The reference power level represents the C/N₀ readings of a hypothetical (ideal) receiver with zero radio front-end losses. As the figure shows, on average there is close conformance between the estimated values of C/N₀ in the two receivers. The difference between the two receivers and the reference is approximately 5 dB, which includes radio front-end noise and other losses. The TUTGNSS receiver displays lower C/N₀ estimation peak-to-peak inconsistency than the commercial receiver.

FIGURE 7. C/N₀ measurement using dCAP: Comparison between TUTGNSS, a commercial, and a hypothetical receiver.

Other Uses of dCAP. During initial prototype validation, we noticed that satellite tracking was inconsistent even under high C/N₀ conditions. dCAP was used to extract detailed baseband tracking information that helped to identify the source of the problem: signal anomalies due to insufficient clock buffering on an experimental RF front end, as shown in FIGURE 8. Such anomalies would have been impossible to detect with traditional black-box testing practices. Once the problem was rectified, dCAP was used once again to monitor the RF front-end signals and performance of the baseband tracking loops, where FIGURES 9 and 10 show a marked improvement in the receiver signal processing and satellite tracking performance.

FIGURE 8. Signal anomaly in the Q-branch signal due to insufficient clock buffering in the experimental RF front end: detected using dCAP.

FIGURE 9. Code Doppler extracted from one tracking loop.

FIGURE 10. Carrier Doppler extracted from one tracking loop using dCAP.

Conclusion

In this article, we have demonstrated the results of the TUTGNSS prototype receiver testing using AutoPET and dCAP. Results were presented, analyzed, and conclusions drawn for the GPS L1 performance of the receiver. Furthermore, the procedures can be easily replicated through software modifications for testing more advanced multi-frequency, multi-constellation modes of the receiver.

Added to the benefits of automation in terms of improved accuracy and personnel efficiency, the proposed AutoPET is a cost-effective solution to anyone working on GNSS receiver technology to understand its most important performance parameters. This tool is portable (software platform-independent), easy to install, and easy to execute on any computer with the basic scientific software. From an academic point of view, dCAP is useful for teaching the spectral characteristics of GNSS signals at every stage from deep inside the receiver to researchers or university students in laboratory exercises. Together, these tools have assisted in the complete characterization of the TUTGNSS receiver at TUT, and can be easily adapted, enhanced, and applied to other research-based receivers as well. In other words, the proposed research has an academic as well as practical appeal.

Acknowledgments

This research work received support from the Tampere Doctoral Programme in Information Science and Engineering (TISE), Nokia Foundation, and the Ulla Tuominen Foundation. It has also been partially supported by the Academy of Finland (under the projects: 251138 “Digitally-Enhanced RF for Cognitive Radio Devices”, and 256175 “Cognitive Approaches for Location in Mobile Environments”). We wish to gratefully acknowledge each of these institutions. This article is based on the paper “Automated Test-bench Infrastructure for GNSS Receivers – Case Study of the TUTGNSS Receiver” presented at the 26th International Technical Meeting of the Satellite Division of The Institute of Navigation held in Nashville, Tennessee, September 16–20, 2013.

Manufacturers

The tests described in this article used a Spirent Federal Systems STR4500 multi-channel GPS/SBAS simulator and a u-blox AG EVK-5P GNSS receiver evaluation kit with a LEA-5P receiver module.

SARANG THOMBRE is a GNSS research scientist in the Department of Navigation and Positioning at the Finnish Geodetic Institute (FGI), Helsinki.

JUSSI RAASAKKA is a GNSS R&D scientist at Honeywell International s.r.o. in the Czech Republic.

TOMMI PAAKKI is a teaching assistant and a doctoral student at the Department of Electronics and Communications Engineering, Tampere University of Technology (TUT).

FRANCESCANTONIO DELLA ROSA is the project manager of the Multitechnology Positioning Professionals (MULTI-POS) Marie Curie Initial Training Network and a research scientist at TUT.

MIKKO VALKAMA is a full professor and the head of the Department of Communications Engineering at TUT.

LAURA RUOTSALAINEN is the deputy head of the Department of Navigation and Positioning and aspecialist research scientist at FGI.

HEIDI KUUSNIEMI is a professor and the acting head of the Department of Navigation and Positioning at FGI.

JARI NURMI is a professor in the Department of Electronics and Communications Engineering at TUT.

FURTHER READING

• Authors’ Conference Paper

“Automated Test-bench Infrastructure for GNSS Receivers – Case Study of the TUTGNSS Receiver” by S. Thombre, J. Raasakka, T. Paakki, F. Della Rosa, M. Valkama, and J. Nurmi in Proceedings of ION GNSS+ 2013, the 26th International Technical Meeting of the Satellite Division of The Institute of Navigation, Nashville, Tennessee, September 16–20, 2013, pp. 1919–1930.

• TUTGNSS

“TUTGNSS – University Based Hardware/Software GNSS Receiver for Research Purposes” by T. Paakki, J. Raasakka, F. Della Rosa, H. Hurskainen, and J. Nurmi, in Proceedings of Ubiquitous Positioning Indoor Navigation and Location Based Service (UPINLBS), 2010, Helsinki, Finland, October 14–15, 2010, doi: 10.1109/UPINLBS.2010.5654337.

• Automated GNSS Receiver Testing

“GPS Interference Testing: Lab, Live, and LightSquared” by P. Boulton, R. Borsato, B. Butler, and K. Judge in InsideGNSS, Vol. 6, No. 4, July/August 2011, pp. 32–45.

“Software-based GNSS Signal Simulators: Past, Present and Possible Future” by S. Thombre, E.S. Lohan, J. Raquet, H. Hurskainen, and J. Nurmi, in Proceedings of ENC GNSS 2010, the European Navigation Conference 2010, Braunschweig, Germany, October 19–21, 2010.

• GNSS Receiver Testing in General

“Simulating GPS Signals: It Doesn’t Have to Be Expensive” by A. Brown, J. Redd, and M.-A. Hutton in GPS World, Vol. 23, No. 5, May 2012, pp. 44–50.

“Realistic Randomization: A New Way to Test GNSS Receivers” by A. Mitelman in GPS World, Vol. 22, No. 3, March 2011, pp. 43–48.

“Record, Replay, Rewind: Testing GNSS Receivers with Record and Playback Techniques” by D.A. Hall in GPS World, Vol. 21, No. 10, October 2010, pp.28–34.

• NMEA 0183

NMEA 0183, The Standard for Interfacing Marine Electronic Devices, Ver. 4.10, published by the National Marine Electronics Association, Severna Park, Maryland, June 2012.

“NMEA 0183: A GPS Receiver Interface Standard” by R.B. Langley in GPS World, Vol. 6, No. 7, July 1995, pp. 54–57.

Unofficial online NMEA 0183 descriptions: “NMEA data”; “NMEA Revealed” by E.S. Raymond, Ver. 2.13, November 2013.
March 1, 2014
Innovation: Cycle Slips

Detection and Correction Using Inertial Aiding

By Malek O. Karaim, Tashfeen B. Karamat, Aboelmagd Noureldin, Mohamed Tamazin, and Mohamed M. Atia

A team of university researchers has developed a technique combining GPS receivers with an inexpensive inertial measuring unit to detect and repair cycle slips with the potential to operate in real time.

INNOVATION INSIGHTS by Richard Langley

DRUM ROLL, PLEASE. The “Innovation” column and GPS World are celebrating a birthday. With this issue, we have started the 25th year of publication of the magazine and the column, which appeared in the very first issue and has been a regular feature ever since. Over the years, we have seen many developments in GPS positioning, navigation, and timing with a fair number documented in the pages of this column.

In January 1990, GPS and GLONASS receivers were still in their infancy. Or perhaps their toddler years. But significant advances in receiver design had already been made since the introduction around 1980 of the first commercially available GPS receiver, the STI-5010, built by Stanford Telecommunications, Inc. It was a dual-frequency, C/A- and P-code, slow-sequencing receiver. Cycling through four satellites took about five minutes, and the receiver unit alone required about 30 centimeters of rack space. By 1990, a number of manufacturers were offering single or dual frequency receivers for positioning, navigation, and timing applications. Already, the first handheld receiver was on the market, the Magellan NAV 1000. Its single sequencing channel could track four satellites. Receiver development has advanced significantly over the intervening 25 years with high-grade multiple frequency, multiple signal, multiple constellation GNSS receivers available from a number of manufacturers, which can record or stream measurements at data rates up to 100 Hz. Consumer-grade receivers have proliferated thanks, in part, to miniaturization of receiver chips and modules. With virtually every cell phone now equipped with GPS, there are over a billion GPS users worldwide. And the chips keep getting smaller. Complete receivers on a chip with an area of less than one centimeter squared are common place. Will the “GPS dot” be in our near future?

The algorithms and methods used to obtain GPS-based positions have evolved over the years, too. By 1990, we already had double-difference carrier-phase processing for precise positioning. But the technique was typically applied in post-processing of collected data. It is still often done that way today. But now, we also have the real-time kinematic (or RTK) technique to achieve similar positioning accuracies in real time and the non-differenced precise point positioning technique, which does not need base stations and which is also being developed for real-time operation. But in all this time, we have always had a “fly in the ointment” when using carrier-phase observations: cycle slips. These are discontinuities in the time series of carrier-phase measurements due to the receiver temporarily losing lock on the carrier of a GPS signal caused by signal blockage, for example. Unless cycle slips are repaired or otherwise dealt with, reduction in positioning accuracy ensues. Scientists and engineers have developed several ways of handling cycle slips not all of which are capable of working in real time. But now, a team of university researchers has developed a technique combining GPS receivers with an inexpensive inertial measuring unit to detect and repair cycle slips with the potential to operate in real time. They describe their system in this month’s column.

“Innovation” is a regular feature that discusses advances in GPS technology and its applications as well as the fundamentals of GPS positioning. The column is coordinated by Richard Langley of the Department of Geodesy and Geomatics Engineering, University of New Brunswick. He welcomes comments and topic ideas.

GPS carrier-phase measurements can be used to achieve very precise positioning solutions. Carrier-phase measurements are much more precise than pseudorange measurements, but they are ambiguous by an integer number of cycles. When these ambiguities are resolved, sub-centimeter levels of positioning can be achieved.

However, in real-time kinematic applications, GPS signals could be lost temporarily because of various disturbing factors such as blockage by trees, buildings, and bridges and by vehicle dynamics. Such signal loss causes a discontinuity of the integer number of cycles in the measured carrier phase, known as a cycle slip. Consequently, the integer counter is reinitialized, meaning that the integer ambiguities become unknown again. In this event, ambiguities need to be resolved once more to resume the precise positioning and navigation process. This is a computation-intensive and time-consuming task. Typically, it takes at least a few minutes to resolve the ambiguities.

The ambiguity resolution is even more challenging in real-time navigation due to receiver dynamics and the time-sensitive nature of the required kinematic solution. Therefore, it would save effort and time if we could detect and estimate the size of these cycle slips and correct the measurements accordingly instead of resorting to a new ambiguity resolution. In this article, we will briefly review the cause of cycle slips and present a procedure for detecting and correcting cycle slips using a tightly coupled GPS/inertial system, which could be used in real time. We will also discuss practical tests of the procedure.

Cycle Slips and Their Management

A cycle slip causes a jump in carrier-phase measurements when the receiver phase tracking loops experience a temporary loss of lock due to signal blockage or some other disturbing factor. On the other hand, pseudoranges remain unaffected. This is graphically depicted in FIGURE 1. When a cycle slip happens, the Doppler (cycle) counter in the receiver restarts, causing a jump in the instantaneous accumulated phase by an integer number of cycles. Thus, the integer counter is reinitialized, meaning that ambiguities are unknown again, producing a sudden change in the carrier-phase observations.

FIGURE 1. A cycle slip affecting phase measurements but not the pseudoranges.

Once a cycle slip is detected, it can be handled in two ways. One way is to repair the slip. The other way is to reinitialize the unknown ambiguity parameter in the phase measurements. The former technique requires an exact estimation of the size of the slip but could be done instantaneously. The latter solution is more secure, but it is time-consuming and computationally intensive. In our work, we follow the first approach, providing a real-time cycle-slip detection and correction algorithm based on a GPS/inertial integration scheme.

GPS/INS Integration

An inertial navigation system (INS) can provide a smoother and more continuous navigation solution at higher data rates than a GPS-only system, since it is autonomous and immune to the kinds of interference that can deteriorate GPS positioning quality. However, INS errors grow with time due to the inherent mathematical double integration in the mechanization process. Thus, both GPS and INS systems exhibit mutually complementary characteristics, and their integration provides a more accurate and robust navigation solution than either stand-alone system. GPS/INS integration is often implemented using a filtering technique. A Kalman filter is typically selected for its estimation optimality and time-recursion properties.

The two major approaches of GPS/INS integration are loosely coupled and tightly coupled. The former strategy is simpler and easier to implement because the inertial and GPS navigation solutions are generated independently before being weighted together by the Kalman filter. There are two main drawbacks with this approach: 1) signals from at least four satellites are needed for a navigation solution, which cannot always be guaranteed; and 2) the outputs of the GPS Kalman filter are time correlated, which has a negative impact upon the system performance. The latter strategy performs the INS/GPS integration in a single centralized Kalman filter. This architecture eliminates the problem of correlated measurements, which arises due to the cascaded Kalman filtering in the loosely coupled approach. Moreover, the restriction of visibility of at least four satellites is removed. We specifically use a tightly coupled GPS/reduced inertial sensor system approach.

Reduced Inertial Sensor System. Recently, microelectromechanical system or MEMS-grade inertial sensors have been introduced for low-cost navigation applications. However, these inexpensive sensors have complex error characteristics.

Therefore, current research is directed towards the utilization of fewer numbers of inertial sensors inside the inertial measurement unit (IMU) to obtain the navigation solution.

The advantage of this trend is twofold. The first is avoidance of the effect of inertial sensor errors. The second is reduction of the cost of the IMU in general. One such minimization approach, and the one used in our work, is known as the reduced inertial sensor system (RISS). The RISS configuration uses one gyroscope, two accelerometers, and a vehicle wheel-rotation sensor. The gyroscope is used to observe the changes in the vehicle’s orientation in the horizontal plane. The two accelerometers are used to obtain the pitch and roll angles. The wheel-rotation sensor readings provide the vehicle’s speed in the forward direction. FIGURE 2 shows a general view of the RISS configuration.

FIGURE 2. A general view of the RISS configuration.

A block diagram of the tightly coupled GPS/RISS used in our work is shown in FIGURE 3. At this stage, the system uses GPS pseudoranges together with the RISS observables to compute an integrated navigation solution. In this three-dimensional (3D) version of RISS, the system has a total of nine states. These states are the latitude, longitude, and altitude errors ( ; the east, north, and up velocity errors ; the azimuth error ; the error associated with odometer-driven acceleration ; and the gyroscope error .

The nine-state error vector x_k at time t_k is expressed as:
    (1)

FIGURE 3. Tightly coupled integration of GPS/RISS using differential pseudorange measurements.

Cycle Slip Detection and Correction

Cycle slip handling usually happens in two discrete steps: detection and fixing or correction. In the first step, using some testing quantity, the location (or time) of the slip is found. During the second step, the size of the slip is determined, which is needed along with its location to fix the cycle slip. Various techniques have been introduced by researchers to address the problem of cycle-slip detection and correction. Different measurements and their combinations are used including carrier phase minus code (using L1 or L2 measurements), carrier phase on L1 minus carrier phase on L2, Doppler (on L1 or L2), and time-differenced phases (using L1 or L2). In GPS/INS integration systems, the INS is used to predict the required variable to test for a cycle slip, which is usually the true receiver-to-satellite range in double-difference (DD) mode, differencing measurements between a reference receiver and the roving receiver and between satellites. In this article, we introduce a tightly coupled GPS/RISS approach for cycle-slip detection and correction, principally for land vehicle navigation using a relative-positioning technique.

Principle of the Algorithm. The proposed algorithm compares DD L1 carrier-phase measurements with estimated values derived from the output of the GPS/RISS system. In the case of a cycle slip, the measurements are corrected with the calculated difference. A general overview of the system is given in FIGURE 4.

FIGURE 4. The general flow diagram of the proposed algorithm.

The number of slipped cycles is given by
   (2)
where
is the DD carrier-phase measurement (in cycles)
is DD estimated carrier phase value (in cycles).
is compared to a pre-defined threshold μ . If the threshold is exceeded, it indicates that there is a cycle slip in the DD carrier-phase measurements.

Theoretically, would be an integer but because of the errors in the measured carrier phase as well as errors in the estimations coming from the INS system, will be a real or floating-point number.

The estimated carrier-phase term in Equation (2) is obtained as follows:
    (3)
where
λ is the wavelength of the signal carrier (in meters)
are the estimated ranges from the rover to satellites i and j respectively (in meters)
are known ranges from the base to satellites i and j respectively (in meters).
What we need to get from the integrated GPS/RISS system is the estimated range vector from the receiver to each available satellite ( ). Knowing our best position estimate, we can calculate ranges from the receiver to all available satellites through:
(4)
where
is the calculated range from the receiver to the mth satellite
x^KF is the receiver position obtained from GPS/RISS Kalman filter solution
x^m is the position of the mth satellite
M is the number of available satellites.
Then, the estimated DD carrier-phase term in Equation (3) can be calculated and the following test quantity in Equation (2) can be applied:
   (5)
If a cycle slip occurred in the ith DD carrier-phase set, the corresponding set is instantly corrected for that slip by:
   (6)
where s is the DD carrier-phase-set number in which the cycle slip has occurred.

Experimental Work

The performance of the proposed algorithm was examined on the data collected from several real land-vehicle trajectories. A high-end tactical grade IMU was integrated with a survey-grade GPS receiver to provide the reference solution. This IMU uses three ring-laser gyroscopes and three accelerometers mounted orthogonally to measure angular rate and linear acceleration. The GPS receiver and the IMU were integrated in a commercial package. For the GPS/RISS solution, the same GPS receiver and a MEMS-grade IMU were used. This IMU is a six-degree of freedom inertial system, but data from only the vertical gyroscope, the forward accelerometer, and the transversal accelerometer was used. TABLE 1 gives the main characteristics of both IMUs. The odometer data was collected using a commercial data logger through an On-Board Diagnostics version II (OBD-II) interface. Another GPS receiver of the same type was used for the base station measurements. The GPS data was logged at 1 Hz.

Table 1. Characteristics of the MEMS and tactical grade IMUs.

Several road trajectories were driven using the above-described configuration. We have selected one of the trajectories, which covers several real-life scenarios encountered in a typical road journey, to show the performance of the proposed algorithm. The test was carried out in the city of Kingston, Ontario, Canada. The starting and end point of the trajectory was near a well-surveyed point at Fort Henry National Historic Site where the base station receiver was located. The length of the trajectory was about 30 minutes, and the total distance traveled was about 33 kilometers with a maximum baseline length of about 15 kilometers. The trajectory incorporated a portion of Highway 401 with a maximum speed limit of 100 kilometers per hour and suburban areas with a maximum speed limit of 80 kilometers per hour. It also included different scenarios including sharp turns, high speeds, and slopes.

FIGURE 5 shows measured carrier phases at the rover for the different satellites. Some satellites show very poor presence whereas some others are consistently available. Satellites elevation angles can be seen in FIGURE 6.

FIGURE 5. Measured carrier phase at the rover.

FIGURE 6. Satellite elevation angles.

Results

We start by showing some results of carrier-phase estimation errors. Processing is done on what is considered to be a cycle-slip-free portion of the data set for some persistent satellites (usually with moderate to high elevation angles). Then we show results for the cycle-slip-detection process by artificially introducing cycle slips in different scenarios. In the ensuing discussion (including tables and figures), we show results indicating satellite numbers without any mention of reference satellites, which should be implicit as we are dealing with DD data.

FIGURE 7 shows DD carrier-phase estimation errors whereas FIGURE 8 shows DD measured carrier phases versus DD estimated carrier phases for sample satellite PRN 22.

FIGURE 7. DD-carrier-phase estimation error, reference satellite with PRN 22.

FIGURE 8. Measured versus estimated DD carrier phase, reference satellite with PRN 22.

As can be seen in TABLE 2, the root-mean-square (RMS) error varies from 0.93 to 3.58 cycles with standard deviations from 0.85 to 2.47 cycles. Estimated phases are approximately identical to the measured ones. Nevertheless, most of the DD carrier-phase estimates have bias and general drift trends, which need some elaboration. In fact, the bias error can be the result of more than one cause. The low-cost inertial sensors always have bias in their characteristics, which plays a major role in this. The drift is further affecting relatively lower elevation angle satellites which can also be attributed to more than one reason. Indeed, one reason for choosing this specific trajectory, which was conducted in 2011, was to test the algorithm with severe ionospheric conditions as the year 2011 was close to a solar maximum: a period of peak solar activity in the approximately 11-year sunspot cycle.

Table 2. Estimation error for DD carrier phases (in cycles).

Moreover, the time of the test was in the afternoon, which has the maximum ionospheric effects during the day. Thus, most part of the drift trend must be coming from ionospheric effects as the rover is moving away from the base receiver during this portion of the trajectory. Furthermore, satellite geometry could contribute to this error component. Most of the sudden jumps coincide with, or follow, sharp vehicle turns and rapid tilts. Table 2 shows the averaged RMS and standard deviation (std) DD carrier-phase estimation error for the sample satellite-pairs. We introduced cycle slips at different rates or intensities and different sizes to simulate real-life scenarios. Fortunately, cycle slips are usually big as mentioned earlier and this was corroborated by our observations from real trajectory data. Therefore, it is more important to detect and correct for bigger slips in general.

Introducing and Detecting Cycle Slips. To test the robustness of the algorithm, we started with an adequate cycle slip size. Cycle slips of size 10–1000 cycles were introduced with different intensities. These intensities are categorized as few (1 slip per 100 epochs), moderate (10 slips per 100 epochs), and severe (100 slips per 100 epochs). This was applied for all DD carrier-phase measurement sets simultaneously. The threshold was set to 1.9267 (average of RMS error for all satellite-pairs) cycles. Four metrics were used to describe the results. Mean square error (MSE); accuracy, the detected cycle slip size with respect to the introduced size; True detection (TD) ratio; and Mis-detection (MD) ratio. Due to space constraints and the similarity between results for different satellites, we only show results for the reference satellite with PRN 22. FIGURES 9–12 show introduced versus calculated cycle slips along with the corresponding detection error for sample satellites in the different scenarios. TABLES 3–5 summarize these results.

FIGURE 9. Introduced and calculated cycle slips (upper plot) and detection error (lower plot). Few cycle slips case, reference satellite with PRN 22.

FIGURE 10. Introduced and calculated cycle slips (upper plot) and detection error (lower plot). Moderate cycle slips case, reference satellite with PRN 22.

FIGURE 11. Introduced and calculated cycle slips (upper plot) and detection error (lower plot). Intensive cycle slips case, reference satellite with PRN 22.

FIGURE 12. Introduced and calculated cycle slips (upper plot) and detection error (lower plot). Small cycle slips case, reference satellite with PRN 22.

Table 3. Few slips (1 slip per 100 epochs).

Table 4. Moderate slips (10 slips per 100 epochs).

Table 5. Intensive slips (100 slips per 100 epochs).

All introduced cycle slips were successfully detected in all of the few, moderate, and severe cases with very high accuracy. A slight change in the accuracy (increasing with higher intensity) among the different scenarios shows that detection accuracy is not affected by cycle-slip intensity. Higher mis-detection ratios for smaller cycle-slip intensity comes from bigger error margins than the threshold for several satellite pairs. However, this is not affecting the overall accuracy strongly as all mis-detected slips are of comparably very small sizes. MD ratio is zero in the intensive cycle-slip case as all epochs contain slips is an indicator of performance compromise with slip intensity.

It is less likely to have very small cycle slips (such as 1 to 2 cycles) in the data and usually it will be hidden with the higher noise levels in kinematic navigation with low-cost equipment. However, we wanted to show the accuracy of detection in this case. We chose the moderate cycle slip intensity for this test. TABLE 6 summarizes results for all satellites.

Table 6. Small slips (1–2 cycles) at moderate intensity (10 slips per 100 epochs).

We get a moderate detection ratio and modest accuracy as the slips are of sizes close to the threshold. The MSE values are not far away from the case of big cycle slips but with higher mis-detection ratio.

Conclusions

The performance of the proposed algorithm was examined on several real-life land vehicle trajectories, which included various driving scenarios including high and slow speeds, sudden accelerations, sharp turns and steep slopes. The road testing was designed to demonstrate the effectiveness of the proposed algorithm in different scenarios such as intensive and variable-sized cycle slips.

Results of testing the proposed method showed competitive detection rates and accuracies comparable to existing algorithms that use full MEMS IMUs. Thus with a lower cost GPS/RISS integrated system, we were able to obtain a reliable phase-measurement-based navigation solution. Although the testing discussed in this article involved post-processing of the actual collected data at the reference station and the rover, the procedure has been designed to work in real time where the measurements made at the reference station are transmitted to the rover via a radio link. This research has a direct influence on navigation in real-time applications where frequent cycle slips occur and resolving integer ambiguities is not affordable because of time and computational reasons and where system cost is an important factor.

Acknowledgments

This article is based on the paper “Real-time Cycle-slip Detection and Correction for Land Vehicle Navigation using Inertial Aiding” presented at ION GNSS+ 2013, the 26th International Technical Meeting of the Satellite Division of The Institute of Navigation held in Nashville, Tennessee, September 16–20, 2013.

Manufacturers

The research reported in this article used a Honeywell Aerospace HG1700 AG11 tactical-grade IMU and a NovAtel OEM4 GPS receiver integrated in a NovAtel G2 Pro-Pack SPAN unit, a Crossbow Technology (now Moog Crossbow) IMU300CC MEMS-grade IMU, an additional NovAtel OEM4 receiver at the base station, a pair of NovAtel GPS-702L antennas, and a Davis Instruments CarChip E/X 8225 OBD-II data logger.

Malek Karaim is a Ph.D. student in the Department of Electrical and Computer Engineering of Queen’s University, Kingston, Ontario, Canada.

Tashfeen Karamat is a doctoral candidate in the Department of Electrical and Computer Engineering at Queen’s University.

Aboelmagd Noureldin is a cross-appointment professor in the Departments of Electrical and Computer Engineering at both Queen’s University and the Royal Military College (RMC) of Canada, also in Kingston.

Mohamed Tamazin is a Ph.D. student in the Department of Electrical and Computer Engineering at Queen’s University and a member of the Queen’s/RMC NavINST Laboratory.

Mohamed M. Atia is a research associate and deputy director of the Queen’s/RMC NavINST Laboratory.

FURTHER READING

• Cycle Slips

“Instantaneous Cycle-Slip Correction for Real-Time PPP Applications” by S. Banville and R.B. Langley in Navigation, Vol. 57, No. 4, Winter 2010–2011, pp. 325–334.

“GPS Cycle Slip Detection and Correction Based on High Order Difference and Lagrange Interpolation” by H. Hu and L. Fang in Proceedings of PEITS 2009, the 2nd International Conference on Power Electronics and Intelligent Transportation System, Shenzhen, China, December 19–20, 2009, Vol. 1, pp. 384–387, doi: 10.1109/PEITS.2009.5406991.

“Cycle Slip Detection and Fixing by MEMS-IMU/GPS Integration for Mobile Environment RTK-GPS” by T. Takasu and A. Yasuda in Proceedings of ION GNSS 2008, the 21st International Technical Meeting of the Satellite Division of The Institute of Navigation, Savannah, Georgia, September 16–19, 2008, pp. 64–71.

“Instantaneous Real-time Cycle-slip Correction of Dual-frequency GPS Data” by D. Kim and R. Langley in Proceedings of KIS 2001, the International Symposium on Kinematic Systems in Geodesy, Geomatics and Navigation, Banff, Alberta, June 5–8, 2001, pp. 255–264.

“Carrier-Phase Cycle Slips: A New Approach to an Old Problem” by S.B. Bisnath, D. Kim, and R.B. Langley in GPS World, Vol. 12, No. 5, May 2001, pp. 46-51.

“Cycle-Slip Detection and Repair in Integrated Navigation Systems” by A. Lipp and X. Gu in Proceedings of PLANS 1994, the IEEE Position Location and Navigation Symposium, Las Vegas, Nevada, April 11–15, 1994, pp. 681–688, doi: 10.1109/PLANS.1994.303377.

Short-Arc Orbit Improvement for GPS Satellites by D. Parrot, M.Sc.E. thesis, Department of Geodesy and Geomatics Engineering Technical Report No. 143, University of New Brunswick, Fredericton, New Brunswick, Canada, June 1989.

• Reduced Inertial Sensor Systems

“A Tightly-Coupled Reduced Multi-Sensor System for Urban Navigation” by T. Karamat, J. Georgy, U. Iqbal, and N. Aboelmagd in Proceedings of ION GNSS 2009, the 22nd International Technical Meeting of the Satellite Division of The Institute of Navigation, Savannah, Georgia, September 22–25, 2009, pp. 582–592.

“An Integrated Reduced Inertial Sensor System – RISS / GPS for Land Vehicle” by U. Iqbal, A. Okou, and N. Aboelmagd in Proceedings of PLANS 2008, the IEEE/ION Position Location and Navigation Symposium, Monterey, California, May 5–8, 2008, pp. 1014–1021, doi: 10.1109/PLANS.2008.4570075.

• Integrating GPS and Inertial Systems

Fundamentals of Inertial Navigation, Satellite-based Positioning and their Integration by N. Aboelmagd, T. B. Karmat, and J. Georgy. Published by Springer-Verlag, New York, New York, 2013.

Aided Navigation: GPS with High Rate Sensors by J. A. Farrell. Published by McGraw-Hill, New York, New York, 2008.

Global Positioning Systems, Inertial Navigation, and Integration, 2nd edition, by M.S. Grewal, L.R. Weill, and A.P. Andrews. Published by John Wiley & Sons, Inc., Hoboken, New Jersey, 2007.

January 1, 2014
Innovation: Hunting for GNSS Echoes

Analysis of Signal Tracking Techniques for Multipath Mitigation

By Antonio Fernández, Mariano Wis, Pau Closas, Carles Fernández-Prades, José A. García, Francesca Zanier, and Massimo Crisci

INNOVATION INSIGHTS by Richard Langley

GETTING RID OF A NUISANCE. No, I’m not talking about your neighbor’s barking dog or the IT guy when he shows up to fiddle, yet again, with your computer. I’m talking about multipath. What is multipath, you ask? Herewith, Multipath 101. When a radio signal travels from a transmitting antenna to a receiving antenna, it will follow a direct line-of-sight path. But the signal might also travel to the receiving antenna after being reflected off a nearby building, say, resulting in a delayed signal or echo along with the line-of-sight one. Those of us of a certain age will remember ghost images on the screens of TVs connected to “rabbit ears” or outdoor antennas. That was multipath. These days, with TV signals primarily delivered by cable and satellite, we don’t see multipath much anymore. But we do hear it in our cars, from time to time, while listening to FM radio. Although the FM “capture effect” provides some margin against multipath, it is not uncommon to lose stereo reception or to experience fading out of the signal while driving in built-up areas as a result of reflections.

This same multipath phenomenon also affects GNSS signals. Unlike satellite TV antennas, the antennas feeding our GNSS receivers are omnidirectional. So we have the possibility of not only receiving a direct, line-of-sight signal from a GNSS satellite but also any indirect signal from the satellite that gets reflected off nearby buildings or other objects or even the ground. The related phenomena of diffraction and scattering can also generate multipath signals.

In a GNSS receiver, the line-of-sight and multipath signals combine to corrupt tracking of the line-of-sight signal resulting in increased pseudorange and carrier-phase measurement errors.

GNSS antenna and receiver manufacturers have developed techniques to minimize some of the impact of multipath on the GNSS observables. And tracking of some of the newer GNSS signals is a bit more resistant to multipath. But multipath, at some level, is still a problem looking for a better solution.

This brings us to ARTEMISA, which stands for Advanced Receiver Techniques: Multiprocessing Algorithms. It’s a European initiative to develop techniques to minimize the effects of multipath in GNSS receivers. For those of you who are a little rusty on your Greek mythology, Artemis (or Artemisa in Spanish — after all, she was a woman) was the Greek goddess of the hunt. You might better know her Roman equivalent: Diana. Her parents were Zeus and Leto, and Apollo was her twin brother. She is often depicted carrying a bow and arrows. How appropriate a name for a project whose goal is to try to kill off the effects of multipath in GNSS receivers.

In this month’s column, the team of researchers involved with ARTEMISA describe their efforts to generate synthetic multipath for GPS L1 and Galileo E1 signals and to test different signal tracking techniques in a simulated receiver to see which techniques best minimize the effects of multipath on positioning solutions and which might be feasible candidates for incorporating in real receivers. The hunt is on.

GNSS navigation in urban environments is usually challenged by a number of effects such as multipath and weak signal conditions. In particular, the pernicious effects of multipath on signal tracking and system accuracy are widely known. To mitigate these effects, there is a series of techniques that range from modified antenna design to combining the GNSS receiver with other sensors or subsystems. Another possibility is to implement advanced tracking techniques specifically designed for these purposes. Such techniques usually impose computational load and implementation complexity, which make them hard to implement in an application-specific-integrated-circuit-based receiver. However, given the current advances in computer technology and the possibilities of field-programmable-gate-array- (FPGA-)based hardware, it is possible to implement these new techniques in an operational receiver.

We have studied this possibility as part of the ARTEMISA project, carried out by DEIMOS Space and the Centre Tecnològic de Telecomunicacions de Catalunya, and supervised by the European Space Agency’s European Space Research and Technology Centre. We have implemented and tested a series of innovative techniques that are able to cope with, and even to estimate, multipath (MP) parameters, using a simulated software receiver based on the GRANADA (Galileo Receiver Analysis and Design Application) GNSS blockset for MathWork’s Simulink graphical programming language tool. These techniques are based on the maximum likelihood principle (as implemented in the Multipath Estimating Delay Lock Loop) or on online Bayesian techniques for the estimation of multipath (as implemented in the Multipath Estimating Particle Filter), involving architectural modifications of the tracking loops (as in vector tracking loops), or even constituting a new paradigm in receiver design (direct position estimation).

Our effort in this project has focused on two main tasks. The first task is the design and development of the simulation platform, the techniques to be tested, and a multipath model representative of the urban environment. The second task involves a simulation campaign that has been carried out to test the different techniques and to contrast the results obtained against the legacy delay lock loop / phase lock loop (DLL/PLL) tracking loop schemes. This article describes these tasks and some of the results we have obtained so far.

Keep in mind that at the time of writing, ARTEMISA is still ongoing. Therefore, more results are expected up until the end of the project.

Simulation Platform

The simulation platform has been developed in Matlab/Simulink with DEIMOS Engenharia’s GRANADA GNSS Blockset. This blockset is a collection of Simulink models and blocks that can be used to design and simulate any kind of GNSS receiver. The main block is the Factor Correlator Model (FCM), which implements the set of correlators through an analytical (set of equations) model. Carrier phase, code misalignment, autocorrelation function, and even the correlated noise and the multipath at the output of every virtual correlator are simulated for a given input trajectory. On the other hand, the tracking loops are implemented as independent modules representative of an actual receiver. This semi-analytical approach has the advantage of performing a fast simulation of the correlator output without the need for implementing the baseband correlation operation. In addition, its implementation in Simulink allows for the development of innovative tracking schemes. This approach also matches with the ARTEMISA project concept, where a series of innovative tracking loops has been implemented with the aim of replacing or improving conventional PLL/DLL schemes.

The main architecture of the simulation platform is shown in FIGURE 1. We use an external trajectory file to generate the reference data that will be used with the FCM block to generate the correlator output that will feed the tracking loop blocks. These trajectory files are also used to feed the multipath scenario generators, to test each technique under a number of defined scenarios so that we can assess their performances and find their limitations under multipath.

Figure 1. ARTEMISA simulation platform architecture.

We used two statistical models for the description of the signal propagation: the Controlled Stochastic Channel Model (CSCM), which is a modification of the Land Mobile Satellite channel developed by Pérez-Fontán, and the well-known Land Mobile Channel Model, developed by the German Aerospace Center (Deutsches Zentrum für Luft- und Raumfahrt or DLR); see Further Reading. These models generate a series of scenario files, which can be loaded into the FCM to introduce the multipath effects in the correlator output.

These two statistical models are complementary. The CSCM model allows the user to set the multipath channel characteristics to be able to stress the tracking technique, while the DLR model permits evaluation of the technique’s response in realistic conditions.

A deterministic user-defined multipath generator was implemented to check the response of the tracking techniques under well-defined multipath conditions. The Multipath Error Envelope (MPEE) was also computed to evaluate the response of the technique under one echo with varying delay conditions.

The purpose of this article is not to explain all the developments performed with the platform. It focuses on the results obtained with the deterministic and statistical channel model. For this reason, the development of the CSCM generator will be explained in detail.

Controlled Stochastic Channel Model

The CSCM module was specifically created for this project. Its purpose is to generate a stochastic channel, but with the capability to control the multipath power levels and the number of echoes generated in the scenario, thus creating a set of realistic multipath signals, but with the capability of being easily controlled by the user. Time series for multipath echoes are generated, following a Rice or Rayleigh stochastic model, but the mean power levels, the amplification K factors, and the power switching times are chosen by the user.

The model allows the selection of the number of satellites (channels) that are generated in the model, the sampling period (related to the loop integration time), the length of the simulation, the receiver speed, the signal carrier frequency, and channel specific parameters such as the number of echoes, the average delay of the echoes and the decay slope (echo power loss ratio with signal delay).

One of the parameters the user can set implicitly is the multipath echo lifetime. What the user really sets is the channel transition time, or the time during which the channel keeps the multipath configuration. Every time the transition time is changed, a series of multipath echoes are canceled and other ones appear. This set of disappearing/appearing echoes is performed in pairs in such a way that the transition between echoes is smooth. FIGURE 2 illustrates this mechanism.

Figure 2. CSCM echoes smooth transition.

When an echo is disappearing (a red one in the figure), its associated echo is at its maximum value (a blue scatterer). In the next interval, a new echo appears in a different delay position (a green one) and the associated scatterer begins to decrease its power. This mechanism allows the user to easily associate the transition time to half the multipath echo lifetime.

Simulation Plan

The simulation plan is structured into different stages. The first stage of the simulation plan is based on the controlled multipath environment, with specific tests for each technique. The purpose of this stage is the tuning up of the techniques. As an example, for the Multipath Estimating Delay Lock Loop (MEDLL) technique, parameters such as the precorrelation bandwidth, the number of MEDLL iterations or the number of assumed multipath echoes are parameters that are adjusted after these tests are carried out. Another purpose of this first stage is to check the performance of the techniques under deterministic, fully controlled multipath. Parameters like signal-to-multipath ratio, multipath delay, and the number of multipath echoes can be controlled at any time.

Another test is the generation of the multipath envelope error plot. The utility of this test is to find the optimal configuration of the correlator position for each kind of signal, which is a trade-off between the obtained root-mean-square error (RMSE) and the number of correlators. This procedure is repeated for each signal considered in the plan.

The next stage of the simulation plan is the test with the CSCM model. This test consists of the generation of a series of scenarios characterized by the stochastic model, the number of echoes, and the user receiver dynamics. Scenarios presented in this article are shown in TABLE 1. Two types of user environments are defined: one for pedestrian and one for vehicular receivers. The dynamics (user speed) define the integration (sampling time) and the multipath Doppler spread. The stochastic model parameters include the type of stochastic model for the line-of-sight signal (LOSS) and multipath echoes, mean power level, and amplification K factor. These scenarios represent a moderate multipath level with four echoes.

Table 1. CSCM scenarios in the simulation plan.

Each scenario was run with different settings of carrier-to-noise-density ratio (C/N₀) and with all the signals that were planned for the project: GPS L1 C/A, GPS L5, Galileo E1 and Galileo E5. In this article, we focus on L1 band signals, analyzing the results for GPS L1 and Galileo E1.

Table 2. GNSS signals considered in the analysis.

The final stage in the simulation plan is testing each technique under realistic channel conditions with the DLR model. In this case, an urban scenario is set, assuming two types of receiver dynamics (a pedestrian user and a vehicular user). No more details about this stage are given because these tests are currently ongoing.

Technique Descriptions

After an initial evaluation of the candidate techniques, the following techniques were selected for implementation and testing in the project.

Multipath Estimating Delay Lock Loop. MEDLL was proposed by Van Nee (see Further Reading). It is a robust statistical approach to the multipath problem, where the maximum likelihood principle is applied to a signal model consisting of LOSS and M-1 multipath rays or echoes. The key idea is to perform the estimation of the whole set of parameters of the incoming signals (that is, amplitude, delay, and phase) in an iterative manner. When their amplitudes, time delays, and carrier phases are estimated, the effect of the reflections in the correlation can be removed. Applying standard assumptions, the maximization of the likelihood function yields a set of interrelated equations from which one can estimate iteratively the LOSS and multipath parameters. The implementation of MEDLL considers a similar architecture as conventional DLLs, although in practice more than three correlators per loop are required for effective multipath mitigation. Despite its demanding requirements in terms of a large number of correlators and computational load, MEDLL was successfully implemented in NovAtel receivers because of its excellent performance in multipath mitigation.

The contributions of all signals (that is, LOSS and an unknown number of echoes) are not calculated simultaneously from the outset. First, the contribution of only one signal is calculated, then the contribution of a second signal is added with the contributions of both signals being optimized, then the contribution of a third signal is added and the contributions of all three signals are optimized, and so on. The process is repeated until a suitable stop criterion is fulfilled. A possible approach for deciding when to stop adding more rays to the signal model is to detect when the error increases; that is, observing that the signal-to-residual ratio when considering an extra path decreases with respect to the case of not considering it, or reaching a maximum number of considered paths, which is a design parameter of this technique.

Multipath Estimating Particle Filter. The MEPF shares with MEDLL the philosophy of estimating the parameters of multipath to mitigate its effect. The main difference is that in this case the statistical principle is not the ML (as in MEDLL) but Bayesian filtering. Here the term Bayesian means that the algorithm is using some sort of a priori information regarding these parameters (such as interdependencies and time evolution models). Therefore, instead of assuming that each integration period is independent of the others, a first-order Markov process is assumed for the unknown parameters (that is, amplitude, delay, and phase). The resulting problem is formulated as a nonlinear state-space model and can be solved by means of a Rao-Blackwellized particle filtering method.

The MEPF has a relatively high computational load, which is a function of the number of particles (N_p), the number of correlation points, and the number of rays to be estimated (M). In all cases, the larger the values the more computationally demanding the algorithm is. According to the simulations performed, configurations with N_p larger than 2000 are not worthwhile, because it implies a high computational cost and the results do not improve significantly the accuracy of a GNSS receiver. Also notice that the dimension of the state-space to be estimated is 3 × M ([amplitude, phase, delay] × M), and thus estimating an additional multipath ray implies augmenting the dimensionality by three. Theoretically, the convergence results of particle filters are independent of the dimensionality of the problem. However, it is agreed that in practice these methods fail when the problem increases in dimension. Therefore, setting M larger than 5 is not reasonable since the number of particles required for convergence would be too large to consider its implementation in a receiver.

Vector Tracking Loops. A conventional GNSS receiver consists of several parallel scalar DLLs, each of which independently estimates the individual pseudoranges. The parallel set of measured pseudoranges (plus Doppler or accumulated delta range measurements on the carrier) are then fed to a Kalman filter estimator, and thus each DLL effectively produces an independent estimate for each of the N pseudoranges for each of the N satellites. However, not all of their measurements are truly independent, although they are treated as such by the DLL, and if there are more than four measurements being made and four or fewer unknowns, the system is overdetermined. Furthermore, the geometry of the satellite-user paths generally prevents the measurements from being truly independent.

The concept of vector tracking loops firstly appeared in the early 1980s. Recently, the method has attracted the attention of many researchers (see Further Reading) due to its good performance in weak signal scenarios.

In this article, we present the implementation of a vector tracking loop that works with pseudorange and pseudorange-rate measurements and provides estimations for position, velocity, receiver clock error, and receiver clock drift. The vector loop has been integrated in a basic receiver based on the classical DLL and PLL/frequency lock loop (FLL)-assisted architecture, acting as an overlay procedure, which activates automatically once the basic receiver has obtained the first position fix. Then, the position/velocity/time solution is used to jointly estimate the synchronization parameters (time delay and carrier phase) for each of the received GNSS signals by means of an extended Kalman filter, and those values are re-injected into the tracking loop. By using position for deriving such synchronization parameters, the algorithm exploits the problem’s inherent geometric constraints, processing all the channels jointly and providing robustness in scenarios with weak receiving power, high multipath, or fast fading.

Direct Position Estimation. Although the conventional two-step position determination is the approach taken traditionally, it is seen to have a number of drawbacks. In contrast, direct position estimation (DPE for short) proposes an alternative where the estimation of a user’s position is performed directly from the received and sampled signal, thus avoiding intermediate steps and jointly considering signals from all satellites when estimating the position solution. By merging the two-step approach into a single estimation problem, DPE addresses some of the inherent drawbacks of the conventional approach where the dependencies between channels are efficiently exploited, in the sense that signals from visible satellites are jointly processed to obtain the user’s position. At the time of writing, this technique is being implemented and its analysis is left for future publications.

Results

Not all the results that we have obtained have been included in this article; only the most relevant ones for L1/E1 CBOC and BPSK signals are given.

MEDLL. Optimal Correlator Configuration. As stated previously, the test consisted of executing different correlator configurations. TABLE 3 shows the correlator configuration ID, the number of correlation samples, and the location of the late correlators with respect to the prompt one and normalized to the chip time (Tc). The corresponding early samples are defined analogously. For all these configurations, it is assumed that there is a correlator at zero chips.

Table 3. MEDLL correlator configurations tested.

These configurations were selected in order to test different possibilities as regular spacing between correlators (MP31, MP61, or MP91), setting the correlators to inflexion points of the autocorrelation function (ACF) (NC37 or NC43 for CBOC modulation), or variable spacing (as configurations AR01 to AR05 and GE01) that optimize the number of correlators. The results of the tests can be observed in FIGURE 3 for a BPSK(1) signal and for a CBOC(6,1,1/11) signal.

Figure 3. MPEE for MEDLL with BPSK and CBOC signals for different correlator configurations (Tc=chip time).

In general, it is observed that the greater the number of correlators, the smaller is the spacing and therefore, the area covered by the multipath error envelope is smaller. However, the greater the number of correlators, the greater is the processing time. Therefore, it is necessary to find an optimal configuration that best fits the multipath variation.

After having analyzed the tests for these configurations, we found that the optimal configurations are AR01 for the BPSK signal and NC37 for the CBOC(6,1,1/11) signal. These configurations are shown in FIGUR 4, and were used for the remaining tests.

Figure 4. Correlator configurations for CBOC and BPSK signals.

CSCM. As mentioned in the simulation plan description, the tests for a pedestrian and a vehicular user in a moderate multipath environment were performed with a scenario file generated with the CSCM tool. These scenarios are characterized by a series of multipath echo levels, number of echoes, and specific stochastic models that are detailed in Table 1. In this table, the pedestrian scenario is known as SP1 and the vehicular scenario is shown as SV1. The results with these scenarios are presented in this article.

The RMSE for range estimates in these scenarios is shown in FIGURE 5 for SP1 and SV1. For comparison, these plots show also the theoretical lower bound (Nunes bound) computed for each technique.

Figure 5. Range RMSE for MEDLL in pedestrian and vehicular scenarios.

The plots show that MEDLL outperforms the conventional DLL results above a minimum C/N₀ value. This happens beyond 32 dB-Hz for a CBOC signal and 35 dB-Hz for a BPSK(1) signal for low dynamics (pedestrian) environments.

In the case of vehicle scenarios such as SV1, it is observed that it requires significantly higher values of C/N₀ in order to outperform the DLL RMSE. This result makes the technique impractical for vehicular applications.

Time Performance. A collateral result that has been obtained with CSCM scenarios is the time performance of the technique, compared to the legacy DLL/PLL scheme. The ratios of the execution times needed for the techniques have been computed. This is an indicative measure of the computational load. The time performance of MEDLL with the selected correlator configuration against DLL is 10:1. That is, 10 times more time is required for the MEDLL technique than the DLL to run the same scenario. It is also observed that the ratio for BPSK modulation is slightly larger than the ratio for the CBOC signal, because more correlators are used in that case.

MEPF. Performance Under Controlled Channel. FIGURE 6 shows an example of the performance of the MEPF technique tested in a controlled channel environment.

Figure 6. Results of the MEPF in the controlled multipath environment with 2 and 3 estimated rays including the line-of-sight. For comparison purposes, DLL/PLL results are also included (blue line in upper plot).

It can be observed how the DLL/PLL technique has an error, which grows as the number of rays is increased. However, when the MEPF is applied, it can be observed how the technique is capable of dealing with a multipath-changing environment despite the fact that it is varying in time with the number of rays increasing. It can also be noticed that the response of the technique is practically the same independent of the number of estimated rays (M). These results were achieved with a BPSK signal and 500 particles.

Covariance Matrix Adjustment. Before starting the specific test for the evaluation of the scenario performance, it is necessary to calibrate the particle filter. The calibration procedure consists of the adjustment of the process covariance matrix. The observation covariance can be adjusted, simply by analyzing the raw observables, or it can even be done automatically.

The critical point is the adjustment of the states’ covariance. It has been observed that an improper adjustment of the covariance may cause an increment in the range RMSE or even the divergence of the filter. For that reason, a systematic procedure for the adjustment of the covariance matrix has been followed. This procedure evaluates the performance using different configurations of the covariance matrices of the line-of-sight and multipath parameters in two stages, and allows us to find a set of optimal values for each scenario.

Optimal Correlator Configuration. We also analyzed the optimal correlator configuration for the MEPF technique. A number of configurations in Table 3 were used under CSCM scenarios. It was found that the correlator configuration does not have a strong influence on the performance of the MEPF technique. It is worth mentioning that in cases where fewer than five correlators were used, the performance degraded. Despite the weak influence of the range RMSE with the correlator configuration, it was observed that there is a minimum for the BPSK signal that can be chosen for the MEPF technique (MP31). For the CBOC signal, the same configuration used for MEDLL is finally chosen (NC37), provided that it is the optimal configuration that balances range RMSE and the number of correlators.

Number of Particles. It is important to notice that an initial set of tests was performed with 500 particles. However, it was found that by increasing the number of particles to 2000, the performance of the technique with CSCM was remarkably better. Because of this, for the remaining tests using the CSCM, this number of particles was the default value. Using a higher number is not worthwhile since results are not much improved and the computational load becomes too high.

CSCM Results. The range RMSE results for the pedestrian scenario SP1 with the MEPF technique are shown in FIGURE 7. In this case, it is observed that for the CBOC signal, MEPF outperforms the DLL/PLL technique for low C/N₀ values. This result could not be reproduced for the BPSK signal because of the adjustment of the covariance matrix, which in this test was optimized for CBOC.

Figure 7. Range RMSE for MEPF and pedestrian scenario.

These promising results for CBOC open the possibility of using this technique in applications where the LOSS is very weak or where the multipath signals are very strong. More tests with this technique are currently being executed for a better adjustment of covariance settings for BPSK.

Time Performance. The previous time performance analysis was performed with the MEPF technique. The performance ranges among 180:1 and 340:1, when compared to DLL/PLL schemes using three correlation samples. The extremely long time required to simulate the scenario makes this technique inadvisable for implementation within an operational receiver using current technology.

VTL. CSCM Results. The same CSCM scenarios were performed for the VTL technique, but using a multi-channel receiver. Results for pedestrian SP1 and vehicular SV1 scenarios are shown in FIGURE 8.

Figure 8. Position RMSE for VTL with pedestrian and vehicular scenarios.

It must be noted that, in order to perform fair comparisons, the results of ordinary DLL/PLL tracking used a conventional KF for the computation of the position instead of the least squares module available in the GRANADA GNSS Blockset.

In our numerical simulations, a significant improvement in the performance of VTL with respect to the DLL/PLL+KF scheme was not observed in pedestrian environments, due to the low dynamics. Under those dynamic conditions, both systems exhibited similar (statistically equivalent) behavior.

In the case of scenario SV1, the velocity applied to the receiver was higher than in the pedestrian case, and the VTL exhibited a remarkable improvement over the DLL/PLL+KF-based receiver. For the BPSK modulation, it was observed that for low values of C/N₀ , the VTL performs better than DLL/PLL+KF. However, for stronger signals, both techniques have the same behavior as shown for the pedestrian scenario results. On the other hand, for the CBOC modulation, it is observed that VTL performs better than DLL/PLL+KF for the whole range of C/N₀ values. The precorrelation bandwidth was set to 8 MHz in the case of BPSK, while for CBOC it was set to 14 MHz. That wider bandwidth of the CBOC receiver has an impact in the estimation of the measurements covariance matrix. In can be observed that in such higher dynamic stress conditions, the VTL outperformed DLL/PLL+KF in our numerical simulations.

It must be mentioned that for all the simulations, we assumed the same C/N₀ for all satellites. However, we plan to run tests with LOSS fading (variation of C/N₀). It is expected that the VTL technique will show its advantage in those simulations.

It was also observed that settings of the KF covariance have an important effect on the results. This covariance must be adjusted according to the multipath signal level. Therefore, a calibration operation, which adapts the KF to a particular scenario should be performed. While the measurement covariance matrix can be adaptively estimated from the output of the DLLs and the PLLs, the process covariance matrix should be adjusted depending on the trajectory characteristics.

Time Performance. The time performance analysis also shows that the time performances of VTL and DLL/PLL+KF are very similar. Only a small increment of 15% of the computational cost for vehicular scenarios has been observed. This makes this technique a good candidate to be implemented in an operational software-based receiver.

Results Overview

After analyzing the current results from the different techniques, some general conclusions on their performance can be made.

MEDLL can be useful to mitigate and improve the pseudorange measurement provided that the C/N₀ value is greater than a threshold value. An intuitive reasoning for this is that the estimation of multipath rays is easier at high C/N₀ values. Below this value, the MEDLL technique does not present an important advantage over ordinary DLL/PLL given the time performance it has (a ratio of 10:1). MEDLL is especially well-suited for static and pedestrian environments.

MEPF is a technique that still needs research before it can be used operatively. Results show that it works quite well for low C/N₀ values when compared to DLL/PLL. Results show that for the CBOC signal, at least 2000 particles are needed to give good results for low C/N₀ values. However, its time performance is very poor (around 200:1 with respect to DLL with 2000 particles)

VTL does not present an advantage in the pseudorange measurement domain, but it clearly improves the position solution with respect to a classical DLL/PLL+KF tracker. This improvement is remarkable under dynamic conditions. It is also observed that the time performance is very similar to the DLL/PLL+KF one. Therefore this technique is a good candidate for implementation in a receiver prototype based on embedded hardware, an FPGA implementation, or software-based radio.

For the sake of completeness, a performance comparison in the range domain of the different techniques with the BPSK and CBOC signals for the pedestrian SP1 scenario are presented in FIGURE 9. The metric used is the range RMSE.

Figure 9. Compared range RMSE for different techniques: pedestrian scenario with BPSK(1) and CBOC(6,1,1/11) signals.

Conclusions and Future Work

This article has presented a series of innovative multipath-estimating techniques using non-conventional approaches in the tracking algorithms. These non-conventional approaches are based on maximum likelihood (MEDLL) or non-linear online filtering algorithms (MEPF). Alternative approaches in the positioning algorithms have also been analyzed (VTL and DPE).

To check the performance of these techniques, we have developed a simulation platform, able to carry out deterministic multipath simulations, in which the multipath environment can be controlled deterministically, and stochastic simulations based on tested multipath models (CSCM and DLR). The CSCM model is capable of simulating realistic multipath environments but with the capability to control the multipath ray parameters and the number of these rays. The DLR model allows us to perform simulations based on the conditions in realistic urban environments.

This article has focused on the CSCM model. It shows simulations for a pedestrian and a vehicular scenario that represent typical dynamic conditions for each kind of user.

These results show that the MEDLL technique performs very well in a static multipath environment under low dynamics with good visibility conditions.

Concerning the MEPF, it has been found that the adjustment of the covariance for the observables is very important for achieving good results for the range RMSE. If this adjustment is done well, the results for low values of C/N₀ outperform the ordinary DLL/PLL technique. This adjustment has been successfully achieved for a CBOC signal, but a BPSK signal still requires additional work. This may open the possibility of using this technique in applications in which the LOSS is very weak.

It has also been observed that the VTL technique is very effective in high dynamics applications and noisy environments, provided that the internal KF process noise covariance has been properly estimated. VTL also shows a performance very similar to the DLL/PLL+KF scheme in mild-condition scenarios. That makes this technique a good candidate for implementation in a real-time software-based receiver.

Finally, it is necessary to remark that ARTEMISA is still a work in progress. An extensive simulation campaign in a realistic urban environment under different conditions using the DLR multipath model is ongoing. In addition to the techniques presented in this article, other advanced techniques such as direct position estimation are under evaluation.

Acknowledgments

The ARTEMISA project, funded by the European Space Agency (ESA) is being carried out by DEIMOS Space, with the Centre Tecnològic de Telecomunicacions de Catalunya as subcontractor. The content of the present article reflects solely the authors’ views and by no means represents official ESA policy.

The authors of this article would like to thank Tiago Peres, Joao Silva, and Pedro Silva from DEIMOS Engenharia; José Antonio Pulido from DEIMOS Space; and Roberto Prieto-Cerdeira from ESA’s European Space Research and Technology Centre for their support in the adaptation of the GRANADA GNSS Blockset and the simulation platform to the requirements of the techniques and multipath environments tested in the project.

ANTONIO FERNANDEZ co-founded DEIMOS Space with headquarters in Madrid, in 2001, where he is currently in charge of the GNSS Business Unit.

MARIANO WIS is currently working for DEIMOS Space as a project engineer in the GNSS Business Unit. He is also a Ph.D. candidate in the Aerospace Science and Technology Program of Universitat Politècnica de Catalunya in Barcelona.

PAU CLOSAS is a senior research associate and head of the Statistical Interference Department in the Communications Systems Division of the Centre Tecnològic de Telecomunicacions de Catalunya (CTTC) in Barcelona.

CARLES FERNANDEZ–PRADES is serving as head of the Communications Systems Division at CTTC, where he holds a position as senior researcher.

JOSE A. GARCIA is with the Radio Navigation Systems and Techniques Section at the European Space Agency’s European Space Research and Technology Centre (ESA/ESTEC) in Noordwijk, The Netherlands.

FRANCESCA ZANIER is also with the ESA/ESTEC Radio Navigation Systems and Techniques Section.

MASSIMO CRISCI is the head of the ESA/ESTEC Radio Navigation Systems and Techniques Section.

FURTHER READING

• ARTEMISA

“ARTEMISA: New GNSS Receiver Processing Techniques for Positioning and Multipath Mitigation” by A.J. Fernandez, J.A. Pulido, M. Wis, F. Zanier, R. Prieto-Cerdeira, M. Crisci, P. Closas, and C. Fernández-Prades in Proceedings of Navitec 2012, the 6th ESA Workshop on Satellite Navigation Technologies, and the European Workshop on GNSS Signals and Signal Processing, Noordwijk, The Netherlands, December 5–7, 2012, doi: 10.1109/NAVITEC.2012.6423092.

• GRANADA GNSS Blockset

“Factored Correlator Model: A Solution for Fast, Flexible, and Realistic GNSS Receiver Simulations” by J.S. Silva, P.F. Silva, A. Fernández, J. Diez, and J.F.M. Lorga in Proceedings of ION GNSS 2007, the 20th International Technical Meeting of the Satellite Division of The Institute of Navigation, Fort Worth, Texas, September 25–28 2007, pp. 2676-2686.

• Signal Propagation Statistical Models

“A Location and Movement Dependent GNSS Multipath Error Model for Pedestrian Applications” by A. Steingass, A. Lehner, and F. Schubert in Proceedings of ION GNSS 2009, the 22nd International Technical Meeting of The Satellite Division of the Institute of Navigation, Savannah, Georgia, September 22–25, 2009, pp. 2284-2296.

“Statistical Modeling of the LMS Channel” by F.P. Fontan, M. Vazquez-Castro, C.E. Cabado, J.P. Garcia, and E. Kubista in IEEE Transactions on Vehicular Technology, Vol. 50, No. 6, November 2001, pp. 1549–1567, doi: 10.1109/25.966585.

• Multipath Estimating Delay Lock Loop

“The Multipath Estimating Delay Lock Loop: Approaching Theoretical Accuracy Limits” by R.D.J. Van Nee, J. Siereveld, P. C. Fenton, and B. R. Townsend in Proceedings of PLANS 1994, the Institute of Electrical and Electronics Engineers Position, Location and Navigation Symposium, Las Vegas, Nevada, April 11–15, 1994, pp. 246–251, doi: 10.1109/PLANS.1994.303320.

• Multipath Estimating Particle Filter

“Nonlinear Bayesian Tracking Loops for Multipath Mitigation” by P. Closas, C. Fernández-Prades, J. Diez, and D. de Castro in International Journal of Navigation and Observation, Vol. 2012, Article ID 359128, 15 pages, 2012, doi:10.1155/2012/359128.

• Vector Tracking Loops

Modeling and Performance Analysis of GPS Vector Tracking Algorithms by M. Lashley, Ph.D. dissertation, Auburn University, Auburn, Alabama, December 2009.

“A VDLL Approach to GNSS Cell Positioning for Indoor Scenarios” by F.D. Nunes, F.M.G. Sousa, and N. Blanco-Delgado in Proceedings of ION GNSS 2009, the 22nd International Technical Meeting of the Satellite Division of The Institute of Navigation, Savannah, Georgia, September 22–25, 2009, pp. 1690–1699.

• Direct Position Estimation

“Maximum Likelihood Estimation of Position in GNSS” by P. Closas, C. Fernández-Prades, and J.A. Fernández-Rubio in IEEE Signal Processing Letters, Vol. 14, No. 15, May 2007, pp. 359-362, doi: 10.1109/LSP.2006.888360.

• Some Previous Innovation Columns on Multipath Mitigation

“Under Cover: Synthetic-Aperture GNSS Signal Processing” by T. Pany, N. Falk, B. Riedl, C. Stöber, J.O. Winkel, and F.-J. Schimpl in GPS World, Vol. 24, No. 9, September 2013, pp. 42–50.

“Multipath Minimization Method: Mitigation Through Adaptive Filtering for Machine Automation Applications” by L. Serrano, D. Kim, and R.B. Langley in GPS World, Vol. 22, No. 7, July 2011, pp. 42–48.

“Multipath Mitigation: How Good Can it Get With the New Signals?” by L.R. Weill, in GPS World, Vol. 14, No. 6, June 2003, pp. 106–113.

“GPS Signal Multipath: A Software Simulator” by S.H. Byun, G.A. Hajj, and L.W. Young in GPS World, Vol. 13, No. 7, July 2002, pp. 40–49.

“Conquering Multipath: The GPS Accuracy Battle” by L.R. Weill in GPS World, Vol. 8, No. 4, April 1997, pp. 59–66.

November 1, 2013
Innovation: Getting Closer to Everywhere

Accurately Tracking Smartphones Indoors

By Ramsey Faragher and Robert Harle

If we wish to obtain consistently usable positions indoors using a mobile phone, we can augment its GPS or GNSS receiver with other unfettered sensing technologies such as gyroscopes and accelerometers supplemented by radio signals of opportunity. But is all of this actually feasible? The authors have conducted tests of a multi-system approach to positioning indoors with favorable results.

INNOVATION INSIGHTS by Richard Langley

IS GPS REALY A GLOBAL POSITIONING SYSTEM? Well, that depends on your definition of “global.” If it means that GPS operates well all over the world in environments where it was designed to work, then, yes, it is a global system. But, if you define global as meaning that GPS operates well everywhere not only outdoors with a clear view of the sky but also indoors and in other restricted environments, then (as some have argued), GPS is not truly global.

So why doesn’t GPS work (for the most part) indoors? Our mobile phones do and they use similar bits of the electromagnetic spectrum. The basic problem is that the signals from GPS (and other GNSS) satellites are just too weak to easily penetrate buildings. They are more than strong enough to yield excellent positioning, navigation, and timing (or PNT) results if the antenna connected to the receiver can “see” the satellites unobstructed. But even outdoors, trees, buildings, and mountains can block the signals from one or more satellites at a time. And indoors, the signals are usually attenuated by walls, floors, and ceilings so much that a conventional receiver cannot lock onto them.

Receiver manufacturers have developed more sensitive receivers that can operate, at least to some degree, indoors but with a good antenna. And receiver chips or modules with this more sensitive technology are often found in modern mobile phones. But they don’t typically provide reliable indoor positioning because they are being used with inexpensive, suboptimal antennas. Some potential improvement in indoor positioning capability is possible by supplying the receiver with satellite orbit and clock information through the mobile network rather than having the receiver acquire this information directly from the satellite signals. This assisted-GNSS technique allows a receiver to work with weaker signals. But it is not a panacea. Gaps or holes still exist for positioning indoors or in other obstructed environments, prompting one industry wag to liken GNSS coverage to Swiss cheese.

So, what are we to do if we wish to obtain consistently usable positions indoors using a mobile phone? As we will see in this month’s column, we can augment or bypass its GPS or GNSS receiver with other unfettered sensing technologies such as gyroscopes and accelerometers. These devices can be made very small using microelectromechanical technology and are already included in some mobile phones.

However, there are some issues with these devices for positioning, not the least of which is rapid position drift. We can restrain the drift by using magnetometers, for example – also present in some mobile phones. We can also use radio signals of opportunity to help in the positioning – signals available in the phone such as multi-generation mobile signals, Bluetooth, and Wi-Fi through their signal strength “fingerprints.” But is all of this actually feasible?

The authors of the article in this month’s column have conducted tests of such a multi-system approach to positioning indoors with quite favorable results. Are we at the stage of accurate positioning (and tracking) everywhere? Not quite, but we are getting closer.

“Innovation” is a regular feature that discusses advances in GPS technology and its applications as well as the fundamentals of GPS positioning. The column is coordinated by Richard Langley of the Department of Geodesy and Geomatics Engineering, University of New Brunswick. He welcomes comments and topic ideas.

In recent years, there has been increasing interest in ubiquitous positioning — accurate location fixes in any environment, outdoors and indoors. We have all become used to the availability and performance of global navigation satellite systems (GNSS) for accurate outdoor radio positioning with a reasonable degree of reliability and availability. However, indoor radio positioning is more challenging because GNSS signals do not penetrate buildings well, and we must instead rely on local infrastructure and other available inputs to aid the user.

Indoor radio positioning is, however, available to the general public today through the use of signal strength fingerprint databases managed and provided by third-party providers such as Skyhook. These typically use Wi-Fi and cellular signals because of their ubiquity and the prevalence of appropriate receiver circuits in consumer devices. The user can also access the fingerprint database through these media. These systems, therefore, have two clear constraints: the database must have been previously built via some form of survey process, and the user must have a data connection available to obtain it. A more scalable system would not rely on such constraints, and would instead develop its own database during operation.

The benefits of such a system are significant: it can provide location-based services, situational awareness, and asset tracking in new and unknown environments for consumers, emergency services, the military, lone workers, security personnel, and autonomous vehicles. There is no requirement for a data link to function, nor any prior surveying of the radio environment, nor any other prior knowledge such as a floor plan or map. However, the system can also be used to quickly and easily generate maps of the radio environment or floor plans, which can be beneficial for organizations wishing to provide positioning services to the public using a simpler positioning method; that is, this method can be used to rapidly survey an area and generate a signal fingerprint database for other users to exploit. Best of all, all of this can be achieved today in real time using an app for a consumer smartphone.

The Digital Swiss Army Knife

The last couple of decades have seen steady improvements in a variety of sectors that have led to new and flexible navigation capabilities — and all of these improvements can now be found in the little chunks of silicon, plastic, and glass in our pockets and handbags. Moore’s Law and the miniaturization of electronics have enabled us all to carry handheld programmable supercomputers around with us every day. Microelectromechanical systems technologies and the demand for better gaming and augmented reality experiences on our smartphones mean that any new phone contains the same types of sensors for enhancing user experiences that cruise missiles and smartbombs use to ensure they hit their targets precisely.

Finally, your smartphone contains more radios than you probably realize. GPS (or GNSS); 2G, 3G, and 4G network radios; near field communications, like RFID; Bluetooth; Wi-Fi; and even a VHF FM chip might be tucked away in there somewhere. The near future is likely to bring a “whitespace” radio (using re-assigned vacated spectrum) along with a 60-GHz wireless USB transceiver. We are bathed in a phenomenal number of radio signals as we go about our daily lives, completely oblivious to the rich tapestry we are walking through — an invisible, permanent, detailed map just waiting to be sensed by our smartphones and annotated for our navigation purposes.

So, just what is possible with a commodity smartphone and its arsenal of features?

Pedestrian Motion Modeling

We can begin with the accelerometers, magnetometers, gyroscopes and barometers found in recent smartphones. These sensors collectively form an inertial measurement unit (IMU) that can be used to track the motion of a user through any environment, regardless of the availability of GNSS (at least in theory).

Unfortunately, there are many stumbling blocks in the way for any new navigator starting down this road. The standard approach for inertial navigation involves using the gyroscopes to maintain an estimate of the orientation of the device relative to the Earth, and to integrate the accelerometer measurements to calculate the system velocity and subsequently the change in position with each measurement update. A key aspect of this process is the removal of the effect of gravity, which requires us to estimate the value of the local gravity field strength (which varies with location across the globe) and its direction (which we do based on the estimated orientation of the device according to the gyroscopes). There are inevitably some errors associated with the estimates of both of these quantities.

In addition, the sensors themselves suffer noise, biases, instabilities, non-linearities, and other effects that only decrease the system performance further. These errors accumulate over time because the position and orientation estimates at any moment depend on the cumulative sum of all measurements since the start of the journey. The result is rapid and unbounded growth in position and orientation error. The cost of the sensors is, of course, tightly correlated with their quality, and so the rate at which the navigation performance degrades. The quality of the sensors in smartphones is so low that this approach is rendered useless within the first few seconds of use. To make progress we must apply regular position corrections to the system by applying external constraints or incorporating external sensor measurements.

Alternative. GNSS measurements provide constraints and corrections for inertial navigation systems, but here we are considering operating indoors where these are unavailable or severely degraded. An alternative solution for most smartphone users is to use the inertial sensors in a different manner, within a so-called pedestrian dead- reckoning (PDR) approach. Here, it is assumed that the device being tracked is held by (or attached to) someone walking in a manner that can be modeled. The inertial sensors are not now used to reproduce the full 3D motion of the device at the update rate of the sensors, but instead used simply to detect stepping motions and to infer that the user has moved some number of steps. Looking for patterns in the accelerometer data where minimum and maximum thresholds are exceeded within a certain time window is a surprisingly robust step counter when the user walks “normally” (more complicated actions such as side steps and stumbles require more complex algorithms). The smartphone can estimate its orientation by fusing together its gyroscope (which offers good short-term orientation-tracking) and its magnetic compass (good long-term orientation-tracking with periodic fluctuations from local magnetic anomalies). The step length of the user (a surprisingly consistent quantity) and any bias in the gyro-smoothed compass heading can both be measured and modeled during periods of GNSS availability such that the best possible estimates are available when GNSS is lost.

FIGURE 1 shows the functional flow diagrams for a strapdown inertial navigation system (top) and a PDR system (bottom). Note that the PDR scheme accumulates error more slowly than the INS scheme (involves fewer integrations over lower-rate data) but is heavily dependent on the performance of the gait recognition, floor-change detection, and step-length-estimation algorithms.

FIGURE 1. Functional flow diagrams for a strapdown inertial navigation system (top) and a pedestrian dead-reckoning system (bottom).

However, PDR techniques still accumulate error, resulting in gradual position drift, but with much higher performance than would be achieved by integrating the raw data in the traditional INS manner. Typical PDR schemes can track the user with an accuracy of a few percent of the distance walked, although this performance degrades with any un-modeled motions that confuse the step detector, such as infrequent backward or sidesteps. So how do we deal with this issue?

Machine Learning

The accuracy of PDR schemes is dependent on the validity of the pedestrian motion model. Any un-modeled action has the potential to generate false positive events in the step detector and hence contribute to position error. Users may stoop, crawl, jump, hop, or shake their device while static — motions that are all very difficult to unambiguously discriminate in raw sensor data.

There are many approaches to solving this problem of gait recognition, and most exploit machine learning techniques. The basic principle of supervised machine learning is that a large set of labeled training data (that is, lots of manually categorized data of each type) is analyzed by a computer in order to extract patterns, statistics, or certain measurement sequences from the inertial sensor measurements that reveal the type of step that was taken. In unsupervised learning, the clusters and categories within the data must be found by the algorithms themselves.

The outputs from such algorithms are typically thresholds, signatures, and other learned metrics that can be installed in a smartphone and used to dynamically classify movements. It is also possible to deploy the learning algorithms on the device itself so that it can learn what the particular user’s signatures are to permit better step and gait detection (like training a speech-recognition program to understand your accent). A simple example of this is running an error-state Kalman filter while GNSS signals are available to determine the user step length and to detect any background compass bias that is corrupting the system.

A problem yet to be resolved for PDR schemes is a basic physical one: the laws of physics are the same for an object at rest as for one moving at constant speed. This means that it is theoretically possible for a suitably skilled person to simulate the “already moving at constant velocity” version of any of these motions while static by moving the device in just the right manner, effectively spoofing as many steps or motions as they like. The opening and closing phases of a journey (that is, the very first and last steps) are critical in distinguishing real and spoofed motion if only inertial sensing is used to disambiguate real and spoofed motion through an environment. We will, however, return to this problem in a moment.

Simultaneous Localization and Mapping

The application of machine learning can be extended to the entire indoor navigation problem using a technique called Simultaneous Localization and Mapping (SLAM). A key aspect here is the hypothesis that there are some measurements that can be taken within an indoor environment that vary rapidly on the spatial scale but only slowly on a temporal scale. These opportunistic measurements are typically of radio signal strength (Wi-Fi, cellular, television, VHF FM, and so on) and magnetic field strength, although in principle many other metrics could be used such as light level and temperature. They are deemed to be opportunistic because they already exist in the environment and have not been generated specifically for this positioning system. Moving along a corridor is expected to result in a particular sequence of measurements that is repeatable on the next visit to that corridor with a confidence based on the time since the last visit. Tight agreement is expected within the next few minutes, close agreement within the next few days, and so on. It is not expected that these fingerprints will necessarily be valid for months or years, as objects may move around the environment; for example, large items may be relocated and Wi-Fi access points may be moved. The ability to exploit the expectation of high repeatability over short time periods of a few hours is the key to developing a system that can learn about its environment and improve its performance during use.

As the device moves through the indoor environment (with position estimate driven by the PDR estimation), the opportunistic fingerprints are captured and stored. If the device returns to a region it has been in before, then it will record a sequence of measurements that will agree closely with the previous sequence that was recorded in the past. This provides a constraint to the system: whatever path was taken in between, it has converged with a section of its historical path and “closed a loop.” Any offset in these two path sections at this point reveals the inertial error that has accumulated during this loop. The system can therefore correct its own inertial error growth, allowing extended operations in GNSS-denied areas.

Fingerprint Maps. The gathered opportunistic measurements can also be used to generate fingerprint maps of the areas that can be shared with other users to allow them to accurately position themselves within those areas in the future, reducing everyone’s reliance on PDR schemes and removing the need for environments to be manually surveyed for their environmental maps. The maps are automatically calibrated and corrected by the SLAM process. As more users operate in the environment and more data accumulate it is easier to identify and remove erroneous data that does not fit into the consensus being formed by the “intelligence of crowds.” This opportunistic navigation scheme can also feed back into the PDR scheme to aid with motion detection — as fingerprints are expected to vary on a fine spatial scale as users move through an environment. They can be used to detect when a PDR device is in reality static, but being moved in a manner that is erroneously triggering the step-detection routine.

FIGURE 2 shows a plot of the magnetic-field-strength variations recorded during four walks down the same corridor of a building at four different times of day on four different days. The traces have been manually aligned by the clear drop in field strength at step number 40. A fixed step length was assumed, and the relative stretching evident across the traces is due to small differences in walking speeds across the tests. Step-length changes can be estimated using changes in the stepping frequency, and the typical step length can be observed and calibrated during periods of GNSS availability.

FIGURE 2. Repeatability tests of the magnetic field strength from four walks along an indoor corridor at four different times during the day on four different days.

There are two distinct classes of SLAM algorithm for PDR. The most common class involves an iterative batch process applied after the data have been collected (that is, offline). This process (which might be least-squares fitting or maximum likelihood estimation, for example) identify loop closure points and provide an optimal joint estimation of the path taken by the user that satisfies these constraints and the raw odometry data as much as possible. The
Wi-Fi SLAM approaches, Gaussian Processes Latent Variables and GraphSLAM, both use such schemes. The results are typically robust, but the offline processing stage can be lengthy.

SLAM can, however, be performed in real time, even on a smartphone, by exploiting an efficient multi-hypothesis scheme. As the user moves, we retain multiple hypotheses for their position and, crucially, record the history of each hypothesis. This is typically done using a particle filter, where each particle represents a unique hypothesis. In this context, we must store the tree of ancestors for each particle at each epoch. When we detect a loop closure, we prune the history to remove all hypotheses that did not result in a loop closure at that point and therefore dynamically correct our errors. Note that each particle can even be assigned different parameter values, such as step length or heading bias, and if a gait detection scheme cannot confidently identify the type of step taken, new particles representing every possible user motion at that epoch can be generated.

Occupancy Grid. Rather than running a specific loop closure algorithm, an occupancy grid is used, whereby the environment is defined by a grid of small cells, for example, one meter by one meter squares. As each particle propagates, representing a hypothesis of the user path, it posts its identity and the current step number into the occupancy grid. As the user continues to move, the particles check the grid cells they move through for any previous visits. If a particle has visited a cell before, the current sensor measurements are compared to those recorded at the time of the last visit. If there is close agreement (typically scored using metrics such as the Euclidean or Mahalanobis distances) then that particular particle is given a high weight. Conversely, poor agreement results in a low weighting.

The entire particle cloud can be reweighted accordingly with low-scoring particles being killed and high-scoring particles being duplicated. The result is the particle cloud collapsing towards the region of close agreement between old and new sensor measurements. Because the occupancy grid contains the historical path of each particle stored via their IDs and step-number sequence, when a reweighting of particles occurs, the historical path of the user is updated and improved accordingly along with the current estimate of the user’s location.

The SLAM estimate can be improved by many types of observations, not just loop closures. If the user moves outside and confident GNSS locations become available, these can also be used to reweight the particle cloud. If the user moves into a region where the floor plan of the building is available to the positioning engine, particles can be pruned whenever they try to cross walls. If desired, even direct user interaction such as manually tapping the map on the smartphone display could be used to provide a position estimate and so constrain the particle cloud.

FIGURE 3 shows six stages from a walk around the corridors of a building using an indoor positioning smartphone app to track the user. The red dashed line shows the trace using just the PDR scheme, which exhibits gradual degradation in positioning accuracy. The green solid line shows the trace using SLAM to constrain the PDR error growth using magnetic anomalies and Wi-Fi signal strengths.

FIGURE 3A.

FIGURE 3B.

FIGURE 3C.

FIGURE 3D.

FIGURE 3E.

FIGURE 3F.

Visual Odometry

A further modern advance is in computer vision: the use of cameras and algorithms to monitor and interpret features in the environment. The movement of features within the field of view from frame to frame can be used to determine the motion of the camera if it is assumed that the majority of the objects tracked through the view are actually static in the environment. Consistency checks between features allow those corresponding to other moving objects to be filtered out.

The result of this visual odometry scheme is the ability to determine the speed and heading changes of the camera by observing the optical flow of the environment. As with PDR approaches, integrating over visual odometry measurements results in motion tracking with much slower reduction in accuracy over time and distance than for systems built upon traditional IMU integration (accelerometers and gyroscopes) alone. If specific objects or features can be uniquely identified and recognized when seen again in the future, then SLAM techniques can also be applied. At the moment, smartphones are powerful enough to apply computer vision techniques and calculations at moderate update rates of a few frames per second. As smartphones become more powerful, or if mobile operating systems will, in future, permit these computer vision algorithms to be deployed on the dedicated graphical processing units, or even perhaps if devices such as Google Glass result in the deployment of dedicated computer vision chips within devices, we will see computer vision coupled with augmented reality move to the forefront of smartphone navigation.

The Future

Our desire for accurate positioning and tracking anywhere will never go away. The availability of cheap, accurate GPS over the last decade has resulted in accurate positioning, navigation, and timing not only being something we take for granted, but something society has come to depend upon. The positioning capabilities of our smartphones will continue to improve, not only because of the new developments and capabilities described above, but because of new infrastructure developments.

The In-Location Alliance is a large consortium of companies including big names like Nokia and CSR who are defining standards for Bluetooth and other beacon-based positioning technologies for dedicated deployments in indoor environments such as shopping centers, airports, and museums. The new 4G LTE signal structure also contains a dedicated ranging signal to permit traditional timing-based positioning schemes to be easily deployed using these new cellular standards. All infrastructure-based schemes incur costs associated with deployment and maintenance that ultimately limit their scope of deployment; opportunistic schemes are the key to truly ubiquitous positioning.

While billions of dollars are being spent worldwide on deploying and maintaining new GNSS, there will always be scenarios and environments where these weak signals are blocked or severely corrupted. In these cases, opportunistic sensing powered by smart algorithms running on consumer devices costing a few hundred dollars will be there to fill those gaps.

Ramsey Faragher is a senior research associate at the University of Cambridge and an associate editor for the journal of the Royal Institute of Navigation. Previously he was a principal scientist at the BAE Systems Advanced Technology Centre, near Chelmsford in the United Kingdom, where he developed the NAVSOP GNSS-denied positioning system. His research interests include opportunistic positioning, sensor fusion, and machine learning.

Robert Harle is a senior lecturer at the University of Cambridge with research interests in positioning, sensor fusion, and wireless sensor networks. He has worked on indoor positioning since 2000, developing a series of infrastructure-based and infrastructure-free solutions.

FURTHER READING

• Simultaneous Localization and Mapping

“SmartSLAM – An Efficient Smartphone Indoor Positioning System Exploiting Machine Learning and Opportunistic Sensing” by R.M. Faragher and R.K. Harle in Proceedings of ION GNSS+ 2013, the 26th International Technical Meeting of the Satellite Division of The Institute of Navigation, Nashville, Tennessee, September 16–20, 2013 (in press).

“Opportunistic Radio SLAM for Indoor Navigation Using Smartphone Sensors,” by R. Faragher, C. Sarno, and M. Newman in Proceedings of PLANS 2012, Institute of Electrical and Electronics Engineers / Institute of Navigation Position, Location and Navigation Symposium, Myrtle Beach, South Carolina, April 23–26, 2012, pp. 120-128.

“Efficient, Generalized Indoor WiFi GraphSLAM” by J. Huang, D. Millman, M. Quigley, D. Stavens, S. Thrun, and A. Aggarwal in Proceedings of 2011 IEEE International Conference on Robotics and Automation, Shanghai, May 9–13, 2011, pp. 1038–1043, doi: 10.1109/ICRA.2011.5979643.

“WiFi-SLAM Using Gaussian Process Latent Variable Models” by B. Ferris, D. Fox, and N. Lawrence in Proceedings of IJCAI-07, the 20th International Joint Conference on Artificial Intelligence, Hyderabad, India, January 6–12, 2007, R. Sangal, H. Mehta, and R. K. Bagga (Eds.), published by Morgan Kaufmann Publishers Inc., San Francisco, California, pp. 2480–2485.

“Simultaneous Map Building and Localization for an Autonomous Mobile Robot” by J.J. Leonard and H.F. Durrant-Whyte in Proceedings of IROS’91, Institute of Electrical and Electronics Engineers / Robotics Society of Japan International Workshop on Intelligence for Mechanical Systems, Osaka, Japan, November 3–5, 1991, pp. 1442–1447, doi: 10.1109/IROS.1991.174711.

• Integrated Indoor Navigation

“A Survey of Indoor Inertial Positioning Systems for Pedestrians” by R. Harle in IEEE Communications Surveys & Tutorials, Vol. 15, No. 3, 2013, pp. 1281–1293, doi: 10.1109/SURV.2012.121912.00075.

Principles of GNSS, Inertial, and Multisensor Integrated Navigation Systems, Second Edition, by P.D. Groves, published by Artech House, Boston, Massachusetts, 2013.

• Wi-Fi Positioning

“Wi-Fi Azimuth and Position Tracking Using Directional Received Signal Strength Measurements” by J. Seitz, T. Vaupel, S. Haimerl, J.G. Boronat, and J. Thielecke in Proceedings of 2012 Workshop on Sensor Data Fusion: Trends, Solutions, Applications, Bonn, September 4–6, 2012, pp. 72–77, doi: 10.1109/SDF.2012.6327911.

“Comparison of WiFi Positioning on Two Mobile Devices” by P.A. Zandbergen in Journal of Location Based Services, Vol. 6, No. 1, 2012, pp. 35–50, doi: 10.1080/17489725.2011.630038.

• Step Length and Pedestrian Navigation

“Step Length Estimation Using Handheld Inertial Sensors” by V. Renaudin, M. Susi, and G. Lachapelle in Sensors, Vol. 12, No. 7, 2012, pp. 8507–8525, doi: 10.3390/s120708507.

• Computer Vision and Navigation

“Improving the Accuracy of EKF-Based Visual-Inertial Odometry” by L. Mingyang and A.I. Mourikis in Proceedings of 2012 IEEE International Conference on Robotics and Automation, Saint Paul, Minnesota, May 14–18, 2012, pp. 828–835, doi: 10.1109/ICRA.2012.6225229.

• Machine Learning

Information Theory, Inference and Learning Algorithms by D.J.C. MacKay, published by Cambridge University Press, Cambridge, U.K., 2003.

• Mobile Phone GPS Antenna Performance

“Mobile-Phone GPS Antennas: Can They Be Better?” by T. Haddrell, M. Phocas, and N. Ricquier in GPS World, Vol. 21, No. 2, February 2010, pp. 29–35.

October 1, 2013
Innovation: Under Cover
Synthetic-Aperture GNSS Signal Processing

By Thomas Pany, Nico Falk, Bernhard Riedl, Carsten Stöber, Jón O. Winkel, and Franz-Josef Schimpl

INNOVATION INSIGHTS by Richard Langley

A SYNTHETIC APERTURE? WHAT’S THAT? Well, an aperture in optics is just a hole or opening through which light travels. Those of us into photography know that the amount of light reaching the camera’s imaging sensor is controlled by the shutter speed and the size of the lens opening or aperture (called the f-stop). And a correct combination of the aperture setting and shutter speed results in a correct exposure. For an optical telescope, its aperture is the diameter of its main, light-gathering lens or mirror. A larger aperture gives a sharper and brighter view or image.

In the radio part of the electromagnetic spectrum, the term aperture refers to the effective collecting (or transmitting) area of an antenna. The gain of the antenna is proportional to its aperture and its beamwidth or resolution is inversely proportional to it.

Astronomers, whether using optical or radio telescopes, often seek higher and higher resolutions to see more detail in the objects they are investigating. Conventionally, that means larger and larger telescopes. However, there are limits to how large a single telescope can be constructed. But by combining the light or radio signals from two or more individual telescopes, one can synthesize a telescope with a diameter equal to the baseline(s) connecting those telescopes. The approach is known as interferometry. It was first tried in the optical domain by the American physicist Albert Michelson who used the technique to measure the diameter of the star Betelgeuse. Radio astronomers developed cable- and microwave-connected interferometers and subsequently they invented the technique of very long baseline interferometry (VLBI) where atomic-clock-stabilized radio signals are recorded on magnetic tape and played back through specially designed correlators to form an image. (VLBI has also been used by geodesists to precisely determine the baselines between pairs of radio telescopes even if they are on separate continents.)

A similar approach is used in synthetic-aperture radar (SAR). Mounted on an aircraft or satellite, the SAR beam-forming antenna emits pulses of radio waves that are reflected from a target and then coherently combined. The different positions of the SAR, as it moves, synthesize an elongated aperture resulting in finer spatial resolution than would be obtained by a conventional antenna.

But what has all of this got to do with GNSS? In this month’s column, we take a look at a novel GNSS signal-processing technique, which uses the principles of SAR to improve code and carrier-phase observations in degraded environments such as under forest canopy. The technique can simultaneously reject multipath signals while maximizing the direct line-of-sight signal power from a satellite. Along with a specially programmed software receiver, it uses either a single conventional antenna mounted, say, on a pedestrian’s backpack for GIS applications or a special rotating antenna for high-accuracy surveying. Want to learn more? Read on.

“Innovation” is a regular feature that discusses advances in GPS technology andits applications as well as the fundamentals of GPS positioning. The column is coordinated by Richard Langley of the Department of Geodesy and Geomatics Engineering, University of New Brunswick. He welcomes comments and topic ideas.

Over the past few years, we have been developing new GNSS receivers and antennas based on an innovative signal-processing scheme to significantly improve GNSS tracking reliability and accuracy under degraded signal conditions. It is based on the principles of synthetic-aperture radar. Like in a multi-antenna phased-array receiver, GNSS signals from different spatial locations are combined coherently forming an optimized synthetic antenna-gain pattern. Thereby, multipath signals can be rejected and the line-of-sight received signal power is maximized. This is especially beneficial in forests and in other degraded environments.

The method is implemented in a real-time PC-based software receiver and works with GPS, GLONASS, and Galileo signals. Multiple frequencies are generally supported.

The idea of synthetic-aperture processing is realized as a coherent summation of correlation values of each satellite over the so-called beam-forming interval. Each correlation value is multiplied with a phase factor. For example, the phase factor can be chosen to compensate for the relative antenna motion over the beam-forming interval and the resulting sum of the scaled correlation values represents a coherent correlation value maximizing the line-of-sight signal power. Simultaneously, signals arriving from other directions are partly eliminated.

Two main difficulties arise in the synthetic-aperture processing. First, the clock jitter during the beam-forming interval must be precisely known. It can either be estimated based on data from all signals, or a stable oscillator can be used. In one of our setups, a modern oven-controlled crystal oscillator with an Allan variance of 0.5 × 10-13 at an averaging period of 1 second is used. Second, the precise relative motion of the antenna during the beam-forming interval must be known. Again it can be estimated if enough sufficiently clean signals are tracked. The antenna trajectory is estimated directly from the correlator values as shown later in this article. In more severely degraded environments, the antenna may be moved along a known trajectory. We are developing a rotating antenna displacement unit. (see FIGURE 1). The rotational unit targets forestry and indoor surveying applications. The relative motion of the antenna is measured with sub-millimeter accuracy.

FIGURE 1. Artist’s impression of the synthetic-aperture GNSS system for surveying in a forest.

After beam-forming, the code pseudoranges and the carrier phases are extracted and used in a conventional way. That is, they are written into Receiver Independent Exchange (RINEX) format files and standard geodetic software can be used to evaluate them. In the case where the artificial movement antenna is used, the GNSS signal processing removes the known part of the movement from the observations, and the observations are then like those from a static antenna. As a result, common static positioning algorithms, including carrier-phase ambiguity fixing, can be applied. The presented method therefore prepares the path for GNSS surveying applications in new areas. An important point is the mechanical realization of the antenna movement. This has to be done in a cost-efficient and reliable way. Lubrication-free actuators are used together with magnetic displacement sensors. The sensors are synchronized to the software receiver front end with better than 1 millisecond accuracy. The rotating antenna uses slip rings to connect the antenna elements. The rotating antenna can also be used to map the received signal power as a function of elevation and azimuth angles. This is beneficial for researchers. For example, it could be used to estimate the direction of arrival of a spoofing signal or to determine which object causes multipath in an indoor environment. For the latter purpose, the rotating antenna can be equipped with left-hand and right-hand circularly polarized antennas on both ends of the rotating bar. The rotating antenna is mounted on a geodetic tripod. See Further Reading for reports of initial studies of the rotating antenna.

Tracking Modes

The synthetic-aperture tracking scheme can be extended to different user-motion schemes or sensor-aiding schemes allowing a wide range of applications. This is reflected in the algorithm implementation within the modular structure of the software receiver. The base module “µ-trajectory & Clock Estimator” in Figure 2 prepares the synthetic-aperture tracking scheme. Different implementations derive from this base class. Each derived module is used for a different user motion scheme and makes use of a different sensor.

FIGURE 2. Different µ-trajectory motion estimators used by the synthetic-aperture processing.

Basically, the modules differ in the way they estimate the relative antenna motion over the beam-forming interval. This relative motion is called the µ-trajectory. Usually the µ-trajectory covers time spans from a few hundreds of milliseconds to a few seconds.

The µ-trajectories have the following characteristics:
- The pedestrian motion estimator does not rely on any sensor measurements and fits a second-order polynomial into the user µ-trajectory of a walking pedestrian. A second-order polynomial is good for representing the motion for up to a quarter of a second.
- The sensor input to the rotating antenna estimator is the relative angular displacement of the rotating antenna. The estimator estimates the absolute direction, which is stable in time. Thus the number of µ-trajectory parameters equals one.
- The vertical antenna motion estimator retrieves the vertical position of the antenna and does not estimate any µ-trajectory parameters. Only clock parameters are estimated.
- Finally, the inertial navigation estimator uses accelerometer and gyro measurements and estimates the 3D user motion. The µ-trajectory parameters consist of accelerometer biases, the gyro biases, attitude errors, and velocity errors. The estimation process is much more complex and exploits the timely correlation of the parameters.
Signal Processing Algorithm

Two kinds of (related) carrier-phase values occur in a GNSS receiver: the numerically controlled oscillator (NCO) internal carrier phase    and the carrier phase pseudorange , which is actually the output of the receiver in, for example, RINEX format files. Both are a function of time t and when expressed in radians are related via Equation (1):

    (1)

Here, f_o denotes the receiver internal nominal intermediate frequency (IF) at which all signal processing takes place. The output carrier-phase pseudorange is an estimate of the true carrier-phase pseudorange , which, in turn, relates to the geometric distance to the satellite by the following standard model:

   (2)

This model applies to each signal propagation path separately; that is, a separate model can be set up for the line-of-sight signal and for each multipath signal. In Equation (2), λ denotes the nominal carrier wavelength in meters, ρ(t) is the geometric distance in meters between transmitting and receiving antennas, f_RF is the nominal carrier frequency in hertz, dt_sat(t) and dt_rec(t) are the satellite and receiver clock errors in seconds, N is the carrier-phase ambiguity, and T(t) contains atmospheric delays as well as any hardware delays in meters. Here, no measurement errors are included, because we are considering the relationship between true values.

Defining now a reference epoch t₀, we will describe a procedure to obtain an improved carrier-phase estimate for this epoch using data from an interval [t₀ – T_BF, t₀]. The beam-forming interval T_BF can be chosen to be, for example, 0.2–2 seconds but should be significantly longer than the employed predetection integration time (the primary one, without beam forming).

Correlator Modeling. In this sub-section, the relationships between phase, correlator values, and geometric distances will be established. These relationships apply for each propagation path individually. In the next section these relationships will be applied to the total received signal, which is the sum of all propagation paths plus thermal noise. To model the correlator output we assume that any effect of code or Doppler-frequency-shift misalignment on carrier-phase tracking can be neglected. This is reasonable if the antenna motion can be reasonably well predicted and this prediction is fed into the tracking loops as aiding information. Then the prompt correlator output is given as

.   (3)

Again, any noise contribution is not considered for the moment. Here a(t) denotes the signal amplitude and d(t) a possibly present navigation data bit. The carrier phase difference Δφ is given as

(4)

where φ(t) is the true carrier phase and φ_NCO(t) is the NCO carrier phase used for correlation.

We now split the geometric line-of-sight distance into an absolute distance, the satellite movement and a relative distance:

(5)

For the example of the rotating antenna, t₀ might be the epoch when the antenna is pointing in the north direction. The term ρ₀(t₀) is the conventional satellite-to-reference-point distance (for example, to the rotation center) and ρ_sat(t₀,t) accounts for the satellite movement during the beam-forming interval.

The term Δρ_µ(t) is the rotational movement and may depend on the parameter µ. The parameter µ represents, for the rotating antenna, the absolute heading but may represent more complex motion parameters. The absolute term ρ₀(t₀) is constant but unknown in the beam-forming interval. We assume that approximate coordinates are available and thus Δρ_µ(t) can be computed for a given set of µ (that is, the line-of-sight projection of the relative motion is assumed to be well predicted even with only approximate absolute coordinates). The same applies also to ρ_sat(t₀,t).

Let’s assume that the NCOs are controlled in a way that the satellite movement is captured as well as the satellite clock drift and the atmospheric delays:

. (6)

Then

(7)

and

.(8)

Thus the correlator output depends on the absolute distance of the reference point to the satellite at t₀, the relative motion of the antenna, the receiver clock error, the received amplitude and the broadcast navigation data bits. Satellite movement and satellite clock drift are absent.

Let us now denote m as the index for the different satellites under consideration. The index k denotes correlation values obtained during the beam-forming interval at the epoch t_k. Then:

.(9)

If multiple signal reflections are received and if they are denoted by the indices m₁, m₂, … , then the correlator output is the sum of those:

.(10)

For the following, m or m₁ denotes the line-of-sight signal and mn with n > 1 denoting multipath signals.

Estimation Principle. It seems natural to choose receiver clock parameters dt_rec and trajectory parameters µ in a way that they optimally represent the receiver correlation values. This approach mimics the maximum likelihood principle. The estimated parameters are:

.(11)

Data bits are also estimated in Equation (11). Once this minimization has been carried out, the parameters µ and dt_rec are known as well as the data bits. The real-time implementation of Equation (11) is tricky. It is the optimization of a multi-dimensional function. Our implementation consists of several analytical simplifications as well as a highly efficient implementation in C code. The pedestrian estimator has been ported to a Compute-Unified-Device-Architecture-capable graphics processing unit exploiting its high parallelism.

Equation (11) realizes a carrier-phase-based vector tracking approach and the whole µ-trajectory (not only positions or velocity values) is estimated at once from the correlation values. This optimally combines the signals from all satellites and frequencies. The method focuses on the line-of-sight signals as only line-of-sight signals coherently add up for the true set of µ-trajectory and clock parameters. On the other hand, multipath signals from different satellites are uncorrelated and don’t show a coherent maximum.

Purified Correlator Values. The line-of-sight relative distance change Δρ_µ^m(t) due to the antenna motion is basically the projection of the µ-trajectory onto the line-of-sight. Multipath signals may arrive from different directions, and   is the antenna motion projected onto the respective direction of arrival.

Let the vector denote the phase signature of the nth multipath signal of satellite m based on the assumed µ-trajectory parameters µ:

.(12)

Projecting the correlator values that have been corrected by data bits and receiver clock error onto the line-of-sight direction yields:

. (13)

The correlator values Q are called purified values as they are mostly free of multipath, provided a suitable antenna movement has been chosen. This is true if we assume a sufficient orthogonality of the line-of-sight signal to the multipath signals, and we can write:

.(14)

where K is the number of primary correlation values within the beam-forming interval. The projection onto the line-of-sight phase signature is then

.(15)

Thus the purified correlator values represent the unknown line-of-sight distance from the reference point to the satellite. Those values are used to compute the carrier pseudorange. The procedure can similarly also be applied for early and late correlators. The purified and projected correlation values represent the correlation function of the line-of-sight signal and are used to compute the code pseudorange.

Block Diagram

This section outlines the block diagram shown in Figure 3 to realize the synthetic-aperture processing. The signal processing is based on the code/Doppler vector-tracking mode of the software receiver.

FIGURE 3. Synthetic-aperture signal processing.

The scheme has not only to include the algorithms of the previous section but it has also to remove the known part of the motion (for the rotating antenna, say) from the output observations. In that case, the output RINEX observation files should refer to a certain static reference point. This is achieved by a two-step process.

First, the known and predictable part of the motion is added to the NCO values. By doing that, the correlation process follows the antenna motion to a good approximation, and the antenna motion does not stress the tracking loop dynamics of the receiver. Furthermore, discriminator values are small and in the linear region of the discriminator. Second, the difference between the current antenna position and the reference point is projected onto the line-of-sight and is removed from the output pseudoranges and Doppler values. For further details on the processing steps of the block diagram, see the conference paper on which this article is based, listed in Further Reading.

Pedestrian Estimator

We tested the synthetic-aperture processing for pedestrians on a dedicated test trial and report the positing results in this section. These results are not final and are expected to improve as more GNSSs are included and general parameter tuning is performed.

Test Area. To test the pedestrian estimator, we collected GPS L1 C/A-code and GLONASS G1 signals while walking through a dense coniferous forest. The trees were up to 30–40 meters high and are being harvested by a strong local lumber industry. The test was carried out in May 2012. We staked out a test course inside the forest and used terrestrial surveying techniques to get precise (centimeter accuracy) coordinates of the reference points. Figure 4 shows a triangular part of the test course.

FIGURE 4. Triangular test course in a forest.

Measurement data was collected with a geodetic-quality GNSS antenna fixed to a backpack. This is a well-known style of surveying. We used a GNSS signal splitter and a commercial application-specific-integrated-circuit- (ASIC-) based high-sensitivity GNSS receiver to track the signals and to have some kind of benchmark. The algorithms of this ASIC-based receiver are not publicly known, but the performance is similar to other ASIC-based GNSS receivers inside forests.

We came from the west, walked the triangular path five times, left to the north, came back from the north, walked the triangular path again five times clockwise, and left to the west. We note that the ASIC-based receiver shows a 3–5 meter-level accuracy with some outliers of more than 10 meters. We further note that the use of the geodetic antenna was critical to achieve this rather high accuracy inside the forest.

µ-trajectory Estimation. As mentioned before, the pedestrian estimator uses a second-order polynomial to model the user motion over an interval of 0.2 seconds. If we stack the estimated µ-trajectories over multiple intervals, we get the relative motion of the user. An example of the estimated user motion outside (but near) the forest is shown in Figure 5.

FIGURE 5. Estimated relative user trajectory over 5 seconds outside the forest; user walking horizontally.

The figure clearly shows that the walking pattern is quite well estimated. An up/down movement of ~10 cm linked to the walking pattern is visible. Inside the forest, the walking pattern is visible but with less accuracy.

Synthetic-Aperture Antenna Pattern. It is possible to estimate the synthetic antenna gain pattern for a given antenna movement (see “Synthetic Phased Array Antenna for Carrier/Code Multipath Mitigation” in Further Reading). The gain pattern is the sensitivity of the receiver/antenna system to signals coming from a certain direction. It depends on the known direction of the line-of-sight signal and is computed for each satellite individually. It adds to the normal pattern of the used antenna element.

We assume that the system simply maximizes the line-of-sight signal power for an assumed satellite elevation of 45° and an azimuth of 135°. We model the pedestrian movement as horizontal with a constant speed of 1 meter per second, and an up/down movement of ± 7.5 centimeters with a period of 0.7 seconds. Employing a beam-forming interval of 2 seconds yields the synthetic antenna gain pattern of Figure 6.The pattern is symmetric to the walking direction. It shows that ground multipath is suppressed.

FIGURE 6. Synthetic antenna aperture diagram for a walking user and beam-forming interval of 2 seconds.

Positioning Results. Our receiver implements a positioning filter based on stacking the estimated µ-trajectory segments. As already mentioned, the stacked µ-trajectory segments represent the relative movement of the user. GNSS code pseudorange observations are then used to get absolute coordinates. Basically, an extended Kalman filter is used to estimate a timely variable position offset to the stacked µ-trajectory segments. The Kalman filter employs a number of data-quality checks to eliminate coarse outliers. They are quite frequent in this hilly forested environment.

The positioning results obtained are shown in Figure 7. They correspond to the same received GPS+GLONASS signal but three different beam-forming intervals (0.2, 1, and 2 seconds) have been used. The position output rate corresponds to the beam-forming interval. Blue markers correspond to the surveyed reference positions, and the yellow markers are estimates when the user is at those reference markers. For each marker, there are ten observations.

FIGURE 7. Estimated user trajectory with 0.2, 1, and 2 seconds beam-forming interval (blue: surveyed reference markers).

The triangular walking path is clearly visible. We observe a bias of around 3 meters and a distance-root-mean-square of 1.2 meters if accounting for this bias (the values refer to the 2-second case). The reason for the bias has not yet been investigated. It could be due to ephemeris or ionospheric errors, but also possibly multipath reflections.

For the short beam-forming interval of 0.2 seconds, we observe noisier walking paths, and we would also expect less accurate code observations. However, the code observation rate is highest in this case (5 Hz), and multipath errors tend to average out inside the Kalman filter. In contrast, the walking paths for the 1-second or 2-second case are straighter. The beam-forming seems to eliminate the multipath, and there are fewer but more precise observations.

Artificial Motion Antennas

The rotating antenna targets surveying applications. It fits standard geodetic equipment. The antenna is controlled by the software receiver, and the rotational information is synchronized to the received GNSS signal.

Synthetic-Aperture Antenna Pattern. With the same methodology as referenced previously, it is possible to estimate the synthetic antenna gain pattern. We assume that the pattern simply maximizes the line-of-sight signal power for an assumed satellite elevation angle of 45° and an azimuth of 135°. We use a rotation radius of 50 cm. The antenna has a really high directivity, eliminating scattered signals from trees. The gain pattern is symmetric with respect to the horizon and ground multipath of perfectly flat ground would not be mitigated by the synthetic aperture. Ground multipath is only mitigated by the antenna element itself (for example, a small ground plane can be used). However, mostly the ground is not flat, and in that case the rotating antenna also mitigates the ground multipath.

Results with a Simulator. The rotating antenna has been tested with simulated GNSS signals using an RF signal generator. The signal generator was configured to start with the antenna at rest, and at some point the antenna starts rotating with a speed of 15 revolutions per minute. Six GPS L1 C/A-code signals have been simulated.

The signal-processing unit has to estimate the antenna state (static or rotating) and the north direction. The quality of the estimation can be visualized by comparing the complex argument of the prompt correlator values to the modeled correlator values. Two examples are shown in FIGURES 8 and 9. In Figure 8, the differences are at the millimeter level corresponding to the carrier-phase thermal noise. This indicates that the absolute heading and receiver clock parameters have been estimated to a high precision.

FIGURE 8. Carrier-phase residuals for all satellites observed with the rotating antenna without multipath. Time is in seconds and all data contributing to the RINEX observation record has been considered.

FIGURE 9. Carrier-phase residuals for all satellites observed with the rotating antenna with multipath. Time is in seconds and all data contributing to the RINEX observation record has been considered.

If multipath from a reflection plane is present (see Figure 9), the phase residuals show the multipath reflection. For example, around t = -0.65 seconds in the figure, the antenna is moving parallel to the reflection plane and the phase residuals are constant over a short time span. As the distance of the antenna to the reflection plane changes, the phase residuals start to oscillate. Generally, the estimation of the absolute heading and of the receiver clock parameters works even with strong multipath signals, but the parameters are not as stable as in the multipath-free case.

In the case when the antenna is rotating, signal processing has to remove the rotation from the code and carrier observations. To check if this elimination of the artificial motion is done correctly, we use carrier-smoothed code observations to compute a single-point-positioning solution. Only if the antenna is rotating can the system estimate the absolute heading and refer the observations to the rotation center. Before that point, the observations refer to the antenna position. The antenna position and the rotation center differ by the radius of 0.5 meters. Since the position is stable for t > 100 seconds, we conclude that the elimination of the artificial motion has been done correctly.

Conclusion

We are in the process of developing positioning solutions for degraded environments based on principles of synthetic-aperture processing. The tools target operational use as an end goal, supporting standard geodetic form factors (tripods) and the software receiver running on standard laptops, and producing data in standardized formats (such as RINEX or the National Marine Electronics Association (NMEA) standards).
Acknowledgments

The research leading to the results reported in this article received funding from the European Community’s Seventh Framework Programme (FP7/2007-2013) under grant agreement No. 287226. This support is gratefully acknowledged. It also received funding from the Upper Bavarian Administration Aerospace Support Program under the contract number 20-8-3410.2-14-2012 (FAUSST), which is also thankfully acknowledged. This article is based on the paper “Concept of Synthetic Aperture GNSS Signal Processing Under Canopy” presented at the European Navigation Conference 2013, held in Vienna, Austria, April 23–25, 2013.

Manufacturer

The research described in this article used an IFEN SX-NSR GNSS software receiver and an IFEN NavX-NCS RF signal generator. The rotating antenna displacement unit was designed and manufactured by Blickwinkel Design & Development.

THOMAS PANY works for IFEN GmbH in Munich, Germany, as a senior research engineer in the GNSS receiver department. He also works as a lecturer (Priv.-Doz.) at the University of the Federal Armed Forces (FAF) Munich and for the University of Applied Science in Graz, Austria. His research interests include GNSS receivers, GNSS/INS integration, signal processing and GNSS science.

NICO FALK received his diploma in electrical engineering from the University of Applied Sciences in Offenburg, Germany. Since then, he has worked for IFEN GmbH in the receiver technology department, focusing on signal processing, hardware, and field-programmable-gate-array development.

BERNHARD RIEDL received his diploma in electrical engineering and information technology from the Technical University of Munich. Since 1994, he has been concerned with research in the field of real-time GNSS applications at the University FAF Munich, where he also received his Ph.D. In 2006, he joined IFEN GmbH, where he is working as the SX-NSR product manager.

JON O. WINKEL is head of receiver technology at IFEN GmbH since 2001. He studied physics at the universities in Hamburg and Regensburg, Germany. He received a Ph.D. (Dr.-Ing.) from the University FAF Munich in 2003 on GNSS modeling and simulations.

FRANZ-JOSEF SCHIMPL started his career as a mechanical engineer and designer at Wigl-Design while studying mechanical engineering. In 2002, he founded Blickwinkel Design & Development with a focus on prototyping and graphic design.

FURTHER READING

• Authors’ Conference Paper

“Concept of Synthetic Aperture GNSS Signal Processing Under Canopy” by T. Pany, N. Falk, B. Riedl, C. Stöber, J. Winkel, and F.-J. Schimpl, Proceedings of ENC-GNSS 2013, the European Navigation Conference 2013, Vienna, Austria, April 23–25, 2013.

• Other Publications on Synthetic-Aperture GNSS Signal Processing

“Synthetic Aperture GPS Signal Processing: Concept and Feasibility Demonstration” by A. Soloviev, F. van Graas, S. Gunawardena, and M. Miller in Inside GNSS, Vol. 4, No. 3, May/June 2009, pp. 37–46. An extended version of the article is available online: http://www.insidegnss.com/node/1453

“Demonstration of a Synthetic Phased Array Antenna for Carrier/Code Multipath Mitigation” by T. Pany and B. Eissfeller in Proceedings of ION GNSS 2008, the 21st International Technical Meeting of The Institute of Navigation, Savannah, Georgia, September 16–19, 2008, pp. 663-668.

“Synthetic Phased Array Antenna for Carrier/Code Multipath Mitigation” by T Pany, M. Paonni, and B. Eissfeller in Proceedings of ENC-GNSS 2008, the European Navigation Conference 2013, Toulouse, France, April 23–25, 2008.

• Software Receiver

“Software GNSS Receiver: An Answer for Precise Positioning Research” by T. Pany, N. Falk, B. Riedl, T. Hartmann, G. Stangl, and C. Stöber in GPS World, Vol. 23, No. 9, September 2012, pp. 60–66.
September 1, 2013
Indian Regional Navigation Satellite Starts Signal Transmissions

By Richard B. Langley

Update (July 29, 2013): The spectrum recorded by the German Aerospace Center researchers appears to be consistent with a combination of BPSK(1) and BOC(5,2) modulation. This is the signal structure that ISRO announced would be used for IRNSS transmissions in the L-band:

“The IRNSS signals consist of two special services namely Standard Positioning Service (SPS) and a Restricted Service (RS) [that] will be carried on L5 and S bands. The SPS will be modulated by a 1 MHz BPSK signal and RS will use BOC(5,2) modulation.”

(“Spectral Compatibility of BOC(5,2) Modulation with Existing GNSS Signals” by S.B. Sekar, S. Sengupta, and K.Bandyopadhyay in Proceedings of IEEE/ION Position Location and Navigation Symposium (PLANS) 2012, Myrtle Beach, SC, April 23–26, 2012, pp.886–890, doi: 10.1109/PLANS.2012.6236831.)

Scientists from the German Aerospace Center’s Institute of Communications and Navigation in Oberpfaffenhofen, Germany, have received signals from IRNSS-1A, the first satellite in the Indian Regional Navigation Satellite System.

Launched on July 1, 2013, the satellite reached its designated inclined geosynchronous orbit by July 18 with an inclination of 27 degrees and an equator crossing of 55 degrees east longitude. Indian Space Research Organisation (ISRO) chairperson Dr. K. Radhakrishnan announced on July 18 that testing of the satellite’s navigation payload would begin within a week.

On July 23, the German Aerospace Center scientists pointed their 30-meter dish antenna at Weilheim towards the satellite and found that it was already transmitting a signal in the L5 frequency band.

FIGURE 1. Spectrum of IRNSS-1A L5 signal. Source: Richard B. Langley

Figure 1 shows the spectrum of the received signal. Centered at 1176.45 MHz, the signal has a single symmetrical main lobe and a number of side lobes characteristic of a spread-spectrum signal. The corresponding IQ constellation diagram is shown in Figure 2. The signal structure appears to be unlike those used by the GPS, GLONASS, Galileo, or BeiDou constellations. Further analysis will be required to sleuth the signal details as ISRO, so far, has not publicly released an IRNSS interface control document (ICD). ICDs characteristically describe a satellite system’s signal structure in detail.

FIGURE 2. IQ constellation diagram of IRNSS-1A L5 signal. Source: Richard B. Langley

The German scientists caution that “this is a very early snapshot of the current signal transmission and probably both the signal power and the signal quality will change and possibly improve during the in-orbit-testing phase of the satellite’s operation.”

July 25, 2013
Innovation: Getting a Grip on Multi-GNSS
The International GNSS Service MGEX Campaign

INNOVATION INSIGHTS by Richard Langley

By Oliver Montenbruck, Chris Rizos, Robert Weber, Georg Weber, Ruth Neilan, and Urs Hugentobler

GPS IS ALMOST 40 YEARS OLD. While mass consumer use of GPS began only within the past decade or so, GPS was “born” during the Labor Day weekend of 1973, when about a dozen military officers and industry analysts under the leadership of Brad Parkinson met to consolidate the concept for a single satellite-based navigation system for the U.S. Department of Defense. Their proposal for NAVSTAR GPS was approved on December 22, 1973. The first satellite to be launched under the GPS program, on July 14, 1974, was the Naval Research Laboratory’s Navigation Technology Satellite (NTS) 1. NTS-2 followed, with a launch on June 23, 1977. These satellites carried payload components similar to those to be used on the subsequent GPS Block I or Navigation Technology Satellites. The first Block I satellite was launched on February 22, 1978, and was followed by nine others. With the launch of the Block II and IIA Operational Satellites and with 24 satellites on orbit, Initial Operational Capability was declared on December 8, 1993. Following testing, Full Operational Capability (FOC) was announced on July 17, 1995.

Whether in reaction to the development of GPS or simply to fulfill the requirement for a system with similar capabilities for its armed forces, the former Soviet Union developed the Global’naya Navigatsionnaya Sputnikovaya Sistema or GLONASS. The first GLONASS satellite was launched on October 12, 1982. By early 1996, a fully populated FOC constellation of 24 satellites was in orbit, but the number of operational satellites dwindled to a handful due to lack of financial support. Eventually the needed funds started flowing again and on December 8, 2011, FOC was again achieved and subsequently maintained.

With the announcement of FOC and the removal of the accuracy-limiting policy of Selective Availability on May 2, 2000, widespread consumer use of GPS took off.

And now GPS is approaching middle age. And like for some humans approaching that milestone desirous of change, a GPS renewal or modernization is under way. New civil and military signals are being transmitted by the Block IIR-M and IIF satellites along with the legacy signals pioneered by the Block I satellites. And new GNSS signals are now coming from the satellites orbited for the Chinese BeiDou Navigation Satellite System, the Japanese Quasi-Zenith Satellite System, and Europe’s Galileo, as well as those from satellite-based augmentation systems. Although it will be some years before full constellations will be transmitting these signals, scientists and engineers are already monitoring and analyzing the new signals to learn how best to use them and how to integrate subsets of them for a wide variety of applications in positioning, navigation, and timing.

In this month’s column, we learn the details of the effort established by the International GNSS Service to support the study of these new signals: the Multi-GNSS Experiment.

“Innovation” is a regular feature that discusses advances in GPS technology andits applications as well as the fundamentals of GPS positioning. The column is coordinated by Richard Langley of the Department of Geodesy and Geomatics Engineering, University of New Brunswick. He welcomes comments and topic ideas. To contact him, see the “Contributing Editors” section on page 4.

Over the past four decades, GPS has evolved from a primarily military navigation system into an indispensable tool not only for society at large, but also for geodetic research and global monitoring of the Earth. And within the past decade, the world of satellite navigation has experienced dramatic changes. With GLONASS, a second GNSS has achieved full operational status; GPS is introducing modernized civil and encrypted navigation signals; and a variety of new navigation constellations are being built up in Asia and Europe.

As of early 2013, Europe has successfully launched a total of four Galileo In-Orbit Validation (IOV) satellites, which are undergoing testing in parallel to the build up and verification of the ground segment. The satellites routinely transmit signals at four frequencies (E1, E5a, E5b, and E6) and offer a variety of publicly accessible pilot and data signals. As a unique feature, Galileo enables tracking of the alternative binary-offset-carrier (AltBOC) signal in the combined E5a+E5b band, which offers superior noise and multipath performance.

Meanwhile, the Chinese BeiDou Satellite Navigation System (BDS; formerly known as Compass) has completed the first stage of its system deployment and declared a regional navigation service for the Asia-Pacific region operational. A total of 14 functioning satellites have been launched so far, which includes five satellites in geostationary orbit (GEO), five satellites in inclined geosynchronous orbit (IGSO), and four in medium-altitude Earth orbit (MEO). These satellites transmit signals in three frequency bands (B1, B2, B3), and tracking of the corresponding open service (OS) signals is already supported by a variety of GNSS receivers. With the release of a B1 OS Interface Control Document (ICD) at the end of 2012, the BeiDou navigation message has become publicly accessible, and users throughout the Asia-Pacific region can now benefit from BeiDou as a supplementary or stand-alone navigation system.

The Japanese Quasi-Zenith Satellite System (QZSS) has, so far, only launched a single satellite but recent political decisions have paved the way for the build up of a mini-constellation of IGSO and GEO satellites. Aside from a high level of compatibility with GPS, QZSS has introduced new signals such as the modernized L1 Civil (L1C) signal and the L-band Experiment (LEX) signal (also known as L6) for high-precision point positioning in the E6 band. Along with this unique set of navigation signals, QZSS provides innovative service features such as the L1 Sub-meter-class Augmentation with Integrity Function (L1-SAIF or L1S) message. Also, QZSS precedes GPS in offering the new Civil Navigation (CNAV) message on L2C and L5, as well as the CNAV2 message on L1C. Long before their planned use in GPS, these messages are now broadcast on a routine basis and contain novel information such as inter-frequency corrections and Earth-orientation parameters.

Last, but not least, GPS has now a total of four Block IIF satellites in orbit that transmit an operational L5 signal for aviation users (and others) and which fly a new generation of highly stable rubidium clocks. While neither L2C nor L5 are transmitted by a full constellation, users and investigators can gradually familiarize themselves with these new signals that will enable encryption-free dual-frequency navigation services for aeronautical and other civil applications.

Within the International GNSS Service (IGS), more than 200 worldwide agencies have, for many years, pooled resources and permanent GNSS station data to generate precise GNSS products in support of Earth science research, multidisciplinary applications, and education. So far, this service has been restricted to two systems — namely, GPS and GLONASS. In recognition of the rapidly evolving GNSS landscape, the IGS has set up the Multi-GNSS Experiment (MGEX) to explore and promote the use of new navigation signals and constellations. It will enable an early familiarization with new GNSS, identify and overcome relevant challenges, and prepare use of emerging navigation systems in routine IGS products. MGEX comprises the build-up of a new network of sensor stations, the characterization of the user equipment and space segment, the development of new concepts and data processing tools, and the generation of early data products for Galileo, QZSS, and BeiDou. MGEX is coordinated by the IGS Multi-GNSS Working Group (MGWG), which interacts closely with other IGS entities, such as the RINEX WG, the Antenna WG, the Data Center WG, and the Infra-structure Committee.

The article starts out with a description of the MGEX network that formed the starting point and initial focus of the overall MGEX project. Following a description of system characterization activities, the current status of multi-GNSS data products and ongoing efforts for the development of new standards for multi-GNSS-related work within the IGS are presented.

Network

Following a call-for-participation released in the summer of 2011, the build up of a new international network of multi-GNSS sensor stations was initiated and has grown substantionally in a short time. By the end of 2012, the MGEX network had comprised approximately 50 stations supporting at least one of the new navigation systems (Galileo, BeiDou, and QZSS) in addition to the legacy GPS, GLONASS, and SBAS systems. At last count, the network now includes 75 stations.

The bulk of the stations is provided by IGS partners such as Bundesamt für Kartographie und Geodäsie (BKG), Centre National d’Etudes Spatiales (CNES), Deutsches GeoForschungsZentrum (GFZ), Deutsches Zentrum für Luft- und Raumfahrt (DLR), Geoscience Australia (GA), the Geospatial Information Authority of Japan (GSI), Institut National de l’Information Géographique et Forestière (IGN), and the Swedish National Land Survey (Lantmaeteriverket, LMV), that have upgraded existing sites with new, multi-GNSS-capable receivers and antennas or started to deploy new multi-GNSS networks (such as the COperative Network for GNSS Observation (CONGO) or CNES’s REseau GNSS pour l’IGS et la Navigation (REGINA) network).

As shown in FIGURE 1, the present set of MGEX stations exhibits almost global coverage, even though a concentration in Europe and a reduced coverage in the Americas and the western Pacific are obvious. However, this situation is expected to improve soon with announced contributions from Geoscience Australia, the Multi-GNSS Monitoring Network (MGM-net) of the Japan Aerospace Exploration Agency (JAXA), and other stations. While most MGEX sites support tracking of Galileo satellites, only a subset of stations provides data for QZSS and BeiDou. In particular, the regional BeiDou constellation (that is, the GEO and IGSO) satellites are not well covered by the current network.

Figure 1. Distribution of MGEX stations supporting tracking of QZSS (blue), Galileo (red), and BeiDou (yellow) as of June 2013.

Further MGEX sites are encouraged and the nomination of sites is still possible through the MGEX submission form under the provision of relevant improvements to the capabilities, coverage, and homogeneity of the overall network.

In terms of equipment, five basic receiver types and seven basic antenna types are employed at the MGEX stations (see TABLES 1 and 2). Observation types provided by the individual receivers have been compiled from summary reports generated by the Astronomical Institute of the University of Bern (AIUB) as part of their routine monitoring of RINEX 3 observation files from MGEX stations.

Table 1. Receiver types in use within the MGEX network (status as of June 2013). Observation types for Galileo (E), BeiDou (C), and QZSS (J) are based on RINEX 3 observation codes as reported in the submitted data files (frequency bands: 1=L1/E1, 2=L2/B1, 5=L5/E5a, 6=E6/B3, 7=E5b/B2, 8=E5ab; signals: C = C/A-code, I = data, Q = pilot, X = data+pilot). They do not necessarily indicate the full tracking capabilities supported by the receivers but rather the observations made available to MGEX users from the respective stations.

Table 2. Antenna types employed within the MGEX network (as of June 2013).

Note that no common standard has yet evolved in terms of supported signals and observation types. This causes certain restrictions for data analysis and product generation. As an example, Galileo orbit and clock products will (at least initially) be based on E1/E5a observations due to a limited coverage of E5b and E5ab tracking.

Selected sites (such as UNB and USNO) offer multiple receivers in short- or zero-baseline configurations to facilitate equipment characterization. Further such installations will be added to the MGEX network at the time of the proposed extensions.

While all stations contribute data to offline archives hosted by the Crustal Dynamics Data Information Service (CDDIS), IGN, and BKG for the MGEX project, a selected subset also supports real-time analyses (see FIGURE 2). All real-time streams utilize the Networked Transport of RTCM via Internet Protocol (NTRIP), which has emerged as a standard for real-time GNSS data exchange. A dedicated MGEX caster is hosted by BKG in Frankfurt, where native raw data streams received from the individual sites are converted and encoded in the RTCM3 Multiple-Signal-Message (MSM) format.

Figure 2. Distribution of MGEX real-time stations supporting tracking of QZSS (blue), Galileo (red), and BeiDou (yellow) as of June 2013.

RTCM3-MSM will enable a harmonized framework for multi-GNSS real-time operations and ensure a seamless conversion to the RINEX3 offline data format. The new MGEX NTRIP caster provides a basis to gain early experience with the new MSM format and facilitates a timely adaptation of user software. This is further supported through freeware software modules for data conversion provided by BKG.

System Characterization

While a systematic quality control of the MGEX data has not yet started, first performance assessments of both the ground and space segment have been reported in the literature (see Further Reading). Overall, the measurement quality of the employed multi-GNSS receivers is found to be comparable or even superior to established GPS reference stations. A high performance is, in particular, obtained for unencrypted signals with high chipping rates and bandwidths such as the GPS/QZSS L5 and Galileo AltBOC.

By way of example, FIGURE 3 illustrates the elevation-angle-dependency of pseudorange errors for BeiDou tracking with a Trimble NetR9 receiver as used at numerous MGEX stations. Aside from the expected variation of receiver noise, the analysis reveals a systematic code bias that varies by 0.4-0.6 meters from horizon to zenith and can best be attributed to spacecraft internal multipath.

Figure 3. Code noise and elevation-dependent biases for BeiDou tracking.

An interesting opportunity for system characterization is provided by triple-frequency observations (GPS+QZSS L1/L2/L5, BeiDou B1/B2/B3, Galileo E1/E5a/E5b) made available by a subset of the MGEX network. A thermal variation of inter-frequency biases has earlier been identified for the GPS Block IIF satellites, but a high level of consistency is demonstrated for QZSS, BeiDou, and Galileo (see FIGURE 4).

Figure 4. Triple-frequency combination of Galileo IOV-3 observations.

Products

While the newly established MGEX network forms a mandatory prerequisite for multi-GNSS work within the IGS, the MGEX campaign supports a wider range of activities, which are now being established. Foremost, the generation of orbit and clock products for the new constellations is promoted in coordination with new and established IGS analysis centers.

Initial Galileo IOV products have been provided by CNES/CLS, CODE, and GFZ since mid-2012, and a combined Galileo+QZSS product has been added by Technische Universität München (TUM). Aside from the MGEX network, some of these solutions make complementary use of proprietary multi-GNSS networks to compensate existing coverage limitations and achieve an improved product quality.

TABLE 3 compares selected MGEX orbit products for the Galileo IOV satellites, while FIGURE 5 shows a time series of the difference between the TUM and CODE orbit products for IOV-1 (E11).

Table 3. Inter-comparison of selected MGEX orbit products for the Galileo IOV satellites. (IOV-1 (E11), DOY 323-329, 2012). Orbit differences (mean ± standard deviation) in radial (R), along-track (T) and cross-track (N) directions are provided in the upper right cells, 3D RMS position differences in the lower left. All values are in units of meters.

Figure 5. Difference of MGEX Galileo IOV-1 (E11) orbit products from TUM and CODE for DOY 232-239, 2012 (R: radial direction, T: along-track direction, N: cross-track direction).

TABLE 4 shows the residuals of Galileo IOV-1/2 satellite laser ranging measurements (mean ± standard deviation) relative to the GNSS-based orbit products (days of year (DOY) 323-329, 2012).

Table 4. Satellite laser ranging residuals (mean ± standard deviation) for Galileo IOV-1/2 orbit products (DOY 323-329, 2012). Units are centimeters.

FIGURE 6 shows a time series of the differences between TUM orbit products for QZS-1 and precise ephemerides computed by the Japan Aerospace Exploration Agency (JAXA) for DOY 027-033, 2013. It demonstrates consistency at the 0.5-meter level, which already represents an important accomplishment, given the sparse subset of QZSS-capable MGEX stations available at the time. Further improvements are expected as the MGEX network continues to grow.

Figure 6. Comparison of TUM MGEX orbit products of QZS-1 with precise ephemerides of JAXA for DOY 027-033, 2013.

Concerning the Chinese BeiDou system, which has now reached initial operational capability for a regional service, early orbit and clock determination results have been reported by various Chinese and European researchers using data from dedicated regional sensor station networks. An effort will be made to promote the extension of the BeiDou tracking capabilities within the MGEX network and to make MGEX-only or mixed network-based orbit and clock products for BeiDou accessible to a wider user community through the MGEX data centers.

Given the late public availability of Galileo and BeiDou broadcast navigation messages, the MGEX orbit and clock products constitute a significant promotion for the early use of all available navigation systems. Aside from initial positioning experiments, they provide a basis for the in-depth characterization of both the space and user segment, and, it is hoped, will facilitate an improved interaction with system providers. Early applications of MGEX multi-GNSS products and observations have, for example, been reported at conferences and in journals (see Further Reading).

Standardization

In support of MGEX, the Multi-GNSS WG interacts closely with other IGS working groups to coordinate data formats, processing standards, and applicable models for use in multi-GNSS work. Examples include necessary RINEX 3 and RTCM3 extensions for full support of new GNSS signals and tracking modes as well as the rapidly growing set of diverse broadcast navigation data.

Another focus of current work addresses the proper modeling of antenna offsets and phase patterns for receiver and satellite antennas, along with documentation of constellation-specific spacecraft coordinate systems and attitude modes. This work is performed in close coordination with the IGS Antenna WG. Among others, conventional antenna phase center offsets (see TABLE 5) have been agreed upon, which will enable consistent processing of MGEX observations until the release of official information by the Galileo and BeiDou program offices.

Table 5. Conventional antenna offsets from the spacecraft center of mass for processing Galileo IOV and BeiDou observations. All values refer to the spacecraft coordinate system. Units are meters.

Public Outreach

As a central point for exchange of MGEX-related information with the user community, a dedicated website has been established at the IGS Central Bureau (see FIGURE 7). The new website provides an overview of available MGEX data and products with direct links to the respective archives at IGS data and product centers. Furthermore, users are provided with up-to-date information on the status of the emerging navigation satellite systems as well as recommended parameters (such as antenna offsets) for harmonized and consistent processing of MGEX observations.

Figure 7. Homepage of the IGS Multi-GNSS Experiment at http://igs.org/mgex.

Through individual members, the Multi-GNSS WG is represented on other boards and bodies such as the International Committee on GNSS, the International Association of Geodesy, and the Multi-GNSS Asia project.

Summary and Conclusions

As part of its continued effort to provide the highest quality data and products for all satellite navigation systems, the IGS has initiated the Multi-GNSS Experiment. MGEX supports early work with new signals and constellations. It enables a timely preparation of IGS analysis centers as well as the user community to expand from GPS/GLONASS towards full multi-GNSS processing.

Within the first year of MGEX, substantial progress has already been made. In particular, a global network of multi-GNSS receivers has been established in parallel with the existing core IGS network and by upgrading existing sites with new multi-GNSS equipment. The MGEX network forms the backbone for all other activities, such as system characterization and product generation. Aside from offline data provisions, a substantial fraction of MGEX stations are already offering real-time data streams, which paves the way for a rapid extension of the upcoming IGS real-time pilot service to Galileo and possibly other constellations. A limited number of analysis centers have already started to provide orbit and clock products for Galileo and/or QZSS as a basis for precise positioning applications. In addition to initial positioning experiments, they will provide a basis for the in-depth characterization of both the space and user segment, and help facilitate an improved interaction with system providers.

Upcoming activities will focus on the systematic incorporation of the BeiDou navigation satellite system. While BeiDou is the third constellation to reach an operational system status after GPS and GLONASS, it is not well covered by the current MGEX tracking network. Along with the targeted incorporation of BeiDou-capable reference stations (particularly in the Asia-Pacific region), the generation and provision of related orbit and clock products will be promoted to facilitate a timely use of BeiDou by the geodetic community.

Subject to active participation by a sufficient number of analysis centers, the MGEX project will eventually transition into an IGS pilot project offering operational data products of Galileo, QZSS, and BeiDou within the next few years.

OLIVER MONTENBRUCK is the head of the GNSS Technology and Navigation Group at DLR’s German Space Operations Center. He chairs the IGS Multi-GNSS Working Group and coordinates the MGEX Multi-GNSS Experiment.

CHRIS RIZOS is a professor at the University of New South Wales, Sydney, Australia; president of the International Association of Geodesy; co-chair of the Multi-GNSS Asia Steering Committee; and a member of the executive and governing board of the IGS.

ROBERT WEBER is an associate professor at the Department of Geodesy and Geoinformation, TU Vienna, and former chair of the IGS GNSS Working Group.

GEORG WEBER is the scientific director in the Department of Geodesy at the German Federal Agency for Cartography and Geodesy (BKG). He is a member of the IGS Real-time Working Group and the Radio Technical Commission for Maritime (RTCM) Services Special Committee (SC) 104 on Differential Global Navigation Satellite Systems (DGNSS).

RUTH NEILAN is the director of the Central Bureau of the IGS, vice-chair of the Global Geodetic Observing System Coordinating Board, and co-chair of Working Group D, Reference Frames, Timing and Applications, of the United Nation’s International Committee on GNSS.

URS HUGENTOBLER is a professor at the Institute of Astronomical and Physical Geodesy, TUM, Munich. He is the chair of the IGS Governing Board.

FURTHER READING

• The IGS MGEX Campaign
“The IGS MGEX Experiment as a Milestone for a Comprehensive Multi-GNSS Service” by C. Rizos, O. Montenbruck, R. Weber, G. Weber, R. Neilan, and U. Hugentobler in Proceedings of The Institute of Navigation 2013 Pacific PNT Meeting, Honolulu, Hawaii, April 23 –25, 2013, pp. 289–295.

“Multi–GNSS Working Group” by O. Montenbruck in International GNSS Service Technical Report 2012, edited by R. Dach and Y. Jean and published by the IGS Central Bureau, April 2013. Pp. 163–170.

“IGS M-GEX – The IGS Multi-GNSS Global Experiment” by R. Weber, U. Hugentobler, and R. Neilan in Proceedings of the 3rd International Colloquium on Scientific and Fundamental Aspects of the Galileo Programme, Copenhagen, 31 August – 2 September, 2011.

• Initial Monitoring and Analysis Results from MGEX
“CNES Computes Real-Time Decimeter-Accuracy Orbits with Galileo” on GPS World website, May 30, 2013.

“Initial Assessment of the COMPASS/BeiDou-2 Regional Navigation Satellite System” by O. Montenbruck, A. Hauschild, P. Steigenberger, U. Hugentobler, P. Teunissen, and S. Nakamura in GPS Solutions, Vol. 17, No. 2, April 2013, pp. 211–222, doi: 10.1007/s10291-012-0272-x.

“Orbit and Clock Determination of QZS-1 Based on the CONGO Network” by P. Steigenberger, A. Hauschild, O. Montenbruck, C. Rodriguez-Solano, and U. Hugentobler in Navigation – Journal of The Institute of Navigation, Vol. 60, No. 1, Spring 2013, pp. 31–40.

“Galileo IOV-3 Broadcasts E1, E5, E6 Signals” by O. Montenbruck and R. Langley in GPS World, Vol. 24, No. 1, January 2013, pp. 18, 27.

“Precise Positioning with Galileo Prototype Satellites: First Results” by R.B. Langley, S. Banville, and P. Steigenberger in GPS World, Vol. 23, No. 9, September 2012, pp. 45–49.

Oral and Poster Presentations at the International GNSS Service Analysis Center Workshop 2012, Olsztyn, Poland, July 23–27, 2012:
- “Curtin and Delft Multi-GNSS M-GEX Stations: Infrastructure and Analysis Tools” by N. Raziq, A. El-Mowafy, P.J.G. Teunissen, A. Nardo, and H. van der Marel. Poster presentation.
- “A COMPASS for Asia: First Experience with the BeiDou-2 Regional Navigation System” by O. Montenbruck, A. Hauschild, P. Steigenberger, U. Hugentobler, and S. Riley. Poster presentation.
- “MGEX Activities at CNES-CLS Analysis Center” by S. Loyer, F. Perosanz, F. Mercier, and H. Capdeville. Poster presentation.
- “MGEX Data Analysis at CODE – First Experiences” by L. Prange, R. Dach, S. Lutz, S. Schaer, M. Meindl, and A. Jäggi. Oral presentation.
- “GFZ’s Global Multi-GNSS Network and First Data Processing Results” by M. Uhlemann, M. Ramatschi, and G. Gendt. Poster presentation.
- “Galileo IOV Orbit Determination and Validation” by S. Hackel, P. Steigenberger, O. Montenbruck, A. Hauschild, and U. Hugentobler. Poster presentation.
• GNSS Signal Structures
Quasi-Zenith Satellite System Navigation Service: Interface Specification for QZSS (IS-QZSS), V1.5, Japan Aerospace Exploration Agency, March 27, 2013.

BeiDou Navigation Satellite System Signal In Space Interface Control Document – Open Service Signal B1I, Version 1.0, China Satellite Navigation Office, December 2012.

European GNSS (Galileo) Open Service: Signal In Space Interface Control Document, Ref : OS SIS ICD, Issue 1.1, September 2010.

• RTCM and NTRIP Formats
Differential GNSS (Global Navigation Satellite Systems) Services, Version 3, RTCM 10403.2, published by Radio Technical Commission for Maritime Services, Arlington, Virginia, February 1, 2013.

“The RTCM Multiple Signal Messages: A New Step in GNSS Data Standardization” by A. Boriskin, D. Kozlov, and G. Zyryanov in Proceedings of ION GNSS 2012, the 25th International Technical Meeting of The Satellite Division of the Institute of Navigation, Nashville, Tennessee, September 17–21, 2012, pp. 2947–2955.

“Networked Transport of RTCM via Internet Protocol (Ntrip) – IP-Streaming for Real-time GNSS Applications” by G. Weber, D. Dettmering, H. Gebhard, and R. Kalafus in Proceedings of ION GPS 2005, the 18th International Technical Meeting of the Satellite Division of The Institute of Navigation, Long Beach, California, September 13–16, 2005, pp. 2243–2247.
July 1, 2013
Innovation: GNSS Spoofing Detection

Correlating Carrier Phase with Rapid Antenna Motion

By Mark L. Psiaki with Steven P. Powell and Brady W. O’Hanlon

INNOVATION INSIGHTS by Richard Langley

IT’S A HOSTILE (ELECTRONIC) WORLD OUT THERE, PEOPLE. Our wired and radio-based communication systems are constantly under attack from evil doers. We are all familiar with computer viruses and worms hiding in malicious software or malware distributed over the Internet or by infected USB flash drives. Trojan horses are particularly insidious. These are programs concealing harmful code that can lead to many undesirable effects such as deleting a user’s files or installing additional harmful software. Such programs pass themselves off as benign, just like the “gift” the Greeks delivered to the Trojans as reported in Virgil’s Aeneid. This was a very early example of spoofing. Spoofing of Internet Protocol (IP) datagrams is particularly prevalent. They contain forged source IP addresses with the purpose of concealing the identity of the sender or impersonating another computing system.

To spoof someone or something is to deceive or hoax, passing off a deliberately fabricated falsehood made to masquerade as truth. The word “spoof” was introduced by the English stage comedian Arthur Roberts in the late 19th century. He invented a game of that name, which involved trickery and nonsense. Now, the most common use of the word is as a synonym for parody or satirize — rather benign actions. But it is the malicious use of spoofing that concerns users of electronic communications.

And it is not just wired communications that are susceptible to spoofing. Communications and other services using radio waves are, in principle, also spoofable. One of the first uses of radio-signal spoofing was in World War I when British naval shore stations sent transmissions using German ship call signs. In World War II, spoofing became an established military tactic and was extended to radar and navigation signals. For example, German bomber aircraft navigated using radio signals transmitted from ground stations in occupied Europe, which the British spoofed by transmitting similar signals on the same frequencies. They coined the term “meaconing” for the interception and rebroadcast of navigation signals (meacon = m(islead)+(b)eacon).

Fast forward to today. GPS and other GNSS are also susceptible to meaconing. From the outset, the GPS P code, intended for use by military and other so-called authorized users, was designed to be encrypted to prevent straightforward spoofing. The anti-spoofing is implemented using a secret “W” encryption code, resulting in the P(Y) code. The C/A code and the newer L2C and L5 codes do not have such protection; nor, for the most part, do the civil codes of other GNSS. But, it turns out, even the P(Y) code is not fully protected from sophisticated meaconing attacks.

So, is there anything that military or civil GNSS users can do, then, to guard against their receivers being spoofed by sophisticated false signals? In this month’s column, we take a look at a novel, yet relatively easily implemented technique that enables users to detect and sequester spoofed signals. It just might help make it a safer world for GNSS positioning, navigation, and timing.

“Innovation” is a regular feature that discusses advances in GPS technology andits applications as well as the fundamentals of GPS positioning. The column is coordinated by Richard Langley of the Department of Geodesy and Geomatics Engineering, University of New Brunswick. He welcomes comments and topic ideas. To contact him, see the “Contributing Editors” section on page 4.

The radionavigation community has known about the dangers of GNSS spoofing for a long time, as highlighted in the 2001 Volpe Report (see Further Reading). Traditional receiver autonomous integrity monitoring (RAIM) had been considered a good spoofing defense. It assumes a dumb spoofer whose false signal produces a random pseudorange and large navigation solution residuals. The large errors are easy to detect, and given enough authentic signals, the spoofed signal(s) can be identified and ignored.

That spoofing model became obsolete at The Institute of Navigation’s GNSS 2008 meeting. Dr. Todd Humphreys introduced a new receiver/spoofer that could simultaneously spoof all signals in a self-consistent way undetectable to standard RAIM techniques. Furthermore, it could use its GNSS reception capabilities and its known geometry relative to the victim to overlay the false signals initially on top of the true ones. Slowly it could capture the receiver tracking loops by raising the spoofer power to be slightly larger than that of the true signals, and then it could drag the victim receiver off to false, but believable, estimates of its position, time, or both.

Two of the authors of this article contributed to Humphreys’ initial developments. There was no intention to help bad actors deceive GNSS user equipment (UE). Rather, our goal was to field a formidable “Red Team” as part of a “Red Team/Blue Team” (foe/friend) strategy for developing advanced “Blue Team” spoofing defenses.

This seemed like a fun academic game until mid-December 2011, when news broke that the Iranians had captured a highly classified Central Intelligence Agency drone, a stealth Lockheed Martin RQ-170 Sentinel, purportedly by spoofing its GPS equipment. Given our work in spoofing and detection, this event caused quite a stir in our Cornell University research group, in Humphreys’ University of Texas at Austin group, and in other places. The editor of this column even got involved in our extensive e-mail correspondence. Two key questions were: Wouldn’t a classified spy drone be equipped with a Selective Availability Anti-Spoofing Module (SAASM) receiver and, therefore, not be spoofable? Isn’t it difficult to knit together a whole sequence of false GPS position fixes that will guide a drone to land in a wrong location? These issues, when coupled with apparent inconsistencies in the Iranians’ story and visible damage to the drone, led us to discount the spoofing claim.

Developing a New Spoofing Defense

My views about the Iranian claims changed abruptly in mid-April 2012. Todd Humphreys phoned me about an upcoming test of GPS jammers, slated for June 2012 at White Sands Missile Range (WSMR), New Mexico. The Department of Homeland Security (DHS) had already spent months arranging these tests, but Todd revealed something new in that call: He had convinced the DHS to include a spoofing test that would use his latest “Red Team” device. The goal would be to induce a small GPS-guided unmanned aerial vehicle (UAV), in this case a helicopter, to land when it was trying to hover. “Wow”, I thought. “This will be a mini-replication of what the Iranians claimed to have done to our spy drone, and I’m sure that Todd will pull it off. I want to be there and see it.” Cornell already had plans to attend to test jammer tracking and geolocation, but we would have to come a day early to see the spoofing “fun” — if we could get permission from U.S. Air Force 746th Test Squadron personnel at White Sands.

The implications of the UAV test bounced around in my head that evening and the next morning on my seven-mile bike commute to work. During that ride, I thought of a scenario in which the Iranians might have mounted a meaconing attack against a SAASM-equipped drone. That is, they might possibly have received and re-broadcast the wide-band P(Y) code in a clever way that could have nudged the drone off course and into a relatively soft landing on Iranian territory.

In almost the next moment, I conceived a defense against such an attack. It involves small antenna motions at a high frequency, the measurement of corresponding carrier-phase oscillations, and the evaluation of whether the motions and phase oscillations are more consistent with spoofed signals or true signals. This approach would yield a good defense for civilian and military receivers against both spoofing and meaconing attacks. The remainder of this article describes this defense and our efforts to develop and test it.

It is one thing to conceive an idea, maybe a good idea. It is quite another thing to bring it to fruition. This idea seemed good enough and important enough to “birth” the conception. The needed follow-up efforts included two parts, one theoretical and the other experimental.

The theoretical work involved the development of signal models, hypothesis tests, analyses, and software. It culminated in analysis and truth-model simulation results, which showed that the system could be very practical, using only centimeters of motion and a fraction of a second of data to reliably differentiate between spoofing attacks and normal GNSS operation.

Theories and analyses can contain fundamental errors, or overlooked real-world effects can swamp the main theoretical effect. Therefore, an experimental prototype was quickly conceived, developed, and tested. It consisted of a very simple antenna-motion system, an RF data-recording device, and after-the-fact signal processing. The signal processing used Matlab to perform the spoofing detection calculations after using a C-language software radio to perform standard GPS acquisition and tracking.

Tests of the non-spoofed case could be conducted anywhere outdoors. Our initial tests occurred on a Cornell rooftop in Ithaca, New York. Tests of the spoofed case are harder. One cannot transmit live spoofing signals except with special permission at special times and in special places, for example, at WSMR in the upcoming June tests. Fortunately, the important geometric properties of spoofed signals can be simulated by using GPS signal reception at an outdoor antenna and re-radiation in an anechoic chamber from a single antenna. Such a system was made available to us by the NASA facility at Wallops Island, Virginia, and our simulated spoofed-case testing occurred in late April of last year. All of our data were processed before mid-May, and they provided experimental confirmation of our system’s efficacy. The final results were available exactly three busy weeks after the initial conception.

Although we were convinced about our new system, we felt that the wider GNSS community would like to see successful tests against live-signal attacks by a real spoofer. Therefore, we wanted very much to bring our system to WSMR for the June 2012 spoofing attack on the drone. We could set up our system near the drone so that it would be subject to the same malicious signals, but without the need to mount our clumsy prototype on a compact UAV helicopter. We were concerned, however, about the possibility of revealing our technology before we had been able to apply for patent protection. After some hesitation and discussions with our licensing and technology experts, we decided to bring our system to the WSMR test, but with a physical cover to keep it secret. The cover consisted of a large cardboard box, large enough to accommodate the needed antenna motions. The WSMR data were successfully collected using this method. Post-processing of the data demonstrated very reliable differentiation between spoofed and non-spoofed cases under live-signal conditions, as will be described in subsequent sections of this article.

System Architecture and Prototype

The components and geometry of one possible version of this system are shown in FIGURE 1. The figure shows three of the GNSS satellites whose signals would be tracked in the non-spoofed case: satellites j-1, j, and j+1. It also shows the potential location of a spoofer that could send false versions of the signals from these same satellites. The spoofer has a single transmission antenna. Satellites j-1, j, and j+1 are visible to the receiver antenna, but the spoofer could “hijack” the receiver’s tracking loops for these signals so that only the false spoofed versions of these signals would be tracked by the receiver.

Figure 1. Spoofing detection antenna articulation system geometry relative to base mount, GNSS satellites, and potential spoofer. Photo: Mark L. Psiaki with Steven P. Powell and Brady W. O’Hanlon

The receiver antenna mount enables its phase center to be moved with respect to the mounting base. In Figure 1, this motion system is depicted as an open kinematic chain consisting of three links with ball joints. This is just one example of how a system can be configured to allow antenna motion. Spoofing detection can work well with just one translational degree of freedom, such as a piston-like up-and-down motion that could be provided by a solenoid operating along the z_a articulation axis. It would be wise to cover the motion system with an optically opaque radome, if possible, to prevent a spoofer from defeating this system by sensing the high-frequency antenna motions and spoofing their effects on carrier phase.

Suppose that the antenna articulation time history in its local body-fixed (x_a, y_a, z_a) coordinate system is b_a(t). Then the received carrier phases are sensitive to the projections of this motion onto the line-of-sight (LOS) directions of the received signals. These projections are along , , and in the non-spoofed case, with being the known unit direction vector from the jth GNSS satellite to the nominal antenna location. In the spoofed case, the projections are all along , regardless of which signal is being spoofed, with being the unknown unit direction vector from the spoofer to the victim antenna. Thus, there will be differences between the carrier-phase responses of the different satellites in the non-spoofed case, but these differences will vanish in the spoofed case. This distinction lies at the heart of the new spoofing detection method. Given that a good GNSS receiver can easily distinguish quarter-cycle carrier-phase variations, it is expected that this system will be able to detect spoofing using antenna motions as small as 4.8 centimeters, that is, a quarter wavelength of the GPS L1 signal.

The UE receiver and spoofing detection block in Figure 1 consists of a standard GNSS receiver, a means of inputting the antenna motion sensor data, and additional signal processing downstream of the standard GNSS receiver operations. The latter algorithms use as inputs the beat carrier-phase measurements from a standard phase-locked loop (PLL).

It may be necessary to articulate the antenna at a frequency nearly equal to the bandwidth of the PLL (say, at 1 Hz or higher). In this case, special post-processing calculations might be required to reconstruct the high-frequency phase variations accurately before they can be used to detect spoofing. The needed post-processing uses the in-phase and quadrature accumulations of a phase discriminator to reconstruct the noisy phase differences between the true signal and the PLL numerically controlled oscillator (NCO) signal. These differences are added to the NCO phases to yield the full high-bandwidth variations.

We implemented the first prototype of this system with one-dimensional antenna motion by mounting its patch antenna on a cantilevered beam. It is shown in FIGURE 2. Motion is initiated by pulling on the string shown in the upper left-hand part of the figure. Release of the string gives rise to decaying sinusoidal oscillations that have a frequency of about 2 Hz.

Figure 2. Antenna articulation system for first prototype spoofing detector tests: a cantilevered beam that allows single-degree-of-freedom antenna phase-center vibration along a horizontal axis. Photo: Mark L. Psiaki with Steven P. Powell and Brady W. O’Hanlon

The remainder of the prototype system consisted of a commercial-off-the-shelf RF data recording device, off-line software receiver code, and off-line spoofing detection software. The prototype system lacked an antenna motion sensor. We compensated for this omission by implementing additional signal-processing calculations. They included off-line parameter identification of the decaying sinusoidal motions coupled with estimation of the oscillations’ initial amplitude and phase for any given detection.

This spoofing detection system is not the first to propose the use of antenna motion to uncover spoofing, and it is related to techniques that rely on multiple antennas. The present system makes three new contributions to the art of spoofing detection: First, it clearly explains why the measured carrier phases from a rapidly oscillating antenna provide a good means to detect spoofing. Second, it develops a precise spoofing detection hypothesis test for a moving-antenna system. Third, it demonstrates successful spoofing detection against live-signal attacks by a “Humphreys-class” spoofer.

Signal Model Theory and Verification

The spoofing detection test relies on mathematical models of the response of beat carrier phase to antenna motion. Reasonable models for the non-spoofed and spoofed cases are, respectively:

(1a)

(1b)

where is the received (negative) beat carrier phase of the authentic or spoofed satellite-j signal at the kth sample time . The three-by-three direction cosines matrix A is the transformation from the reference system, in which the direction vectors   and are defined, to the local body-axis system, in which the antenna motion b_a(t) is defined. λ is the nominal carrier wavelength. The terms involving the unknown polynomial coefficients , , and model other low-frequency effects on carrier phase, including satellite motion, UE motion if its antenna articulation system is mounted on a vehicle, and receiver clock drift. The term is the receiver phase noise. It is assumed to be a zero-mean, Gaussian, white-noise process whose variance depends on the receiver carrier-to-noise-density ratio and the sample/accumulation frequency.

If the motion of the antenna is one-dimensional, then b_a(t) takes the form , with being the articulation direction in body-axis coordinates and r_a(t) being a known scalar antenna deflection amplitude time history. If one defines the articulation direction in reference coordinates as , then the carrier-phase models in Equations (1a) and (1b) become

   (2a)

(2b)

There is one important feature of these models for purposes of spoofing detection. In the non-spoofed case, the term that models the effects of antenna motion varies between GPS satellites because the direction vector varies with j. The spoofed case lacks variation between the satellites because the one spoofer direction replaces for all of the spoofed satellites. This becomes clear when one compares the first terms on the right-hand sides of Eqsuations (1a) and (1b) for the 3-D motion case and on the right-hand sides of Equations (2a) and (2b) for the 1-D case.

The carrier-phase time histories in FIGURES 3 and 4 illustrate this principle. These data were collected at WSMR using the prototype antenna motion system of Figure 2. The carrier-phase time histories have been detrended by estimating the , , and coefficients in Equations (2a) and (2b) and subtracting off their effects prior to plotting. In Figure 3, all eight satellite signals exhibit similar decaying sinusoid time histories, but with differing amplitudes and some of them with sign changes. This is exactly what is predicted by the 1-D non-spoofed model in Equation (2a). All seven spoofed signals in Figure 4, however, exhibit identical decaying sinusoidal oscillations because the term in Equation (2b) is the same for all of them.

Figure 3. Detrended carrier-phase data from multiple satellites for a typical non-spoofed case using a 1-D antenna articulation system.

Figure 4. Multiple satellites’ detrended carrier-phase data for a typical spoofed case using a 1-D antenna articulation system.

As an aside, an interesting feature of Figure 3 is its evidence of the workings of the prototype system. The ramping phases of all the signals from t = 0.4 seconds to t = 1.4 seconds correspond to the initial pull on the string shown in Figure 2, and the steady portion from t = 1.4 seconds to t = 2.25 seconds represents a period when the string was held fixed prior to release.

Spoofing Detection Hypothesis Test

A hypothesis test can precisely answer the question of which model best fits the observed data: Does carrier-phase sameness describe the data, as in Figure 4? Then the receiver is being spoofed. Alternatively, is carrier-phase differentness more reasonable, as per Figure 3? Then the signals are trustworthy.

A hypothesis test can be developed for any batch of carrier-phase data that spans a sufficiently rich antenna motion profile b_a(t) or ρ_a(t). The profile must include high-frequency motions that cannot be modeled by the , , and quadratic polynomial terms in Equations (1a)-(2b); otherwise the detection test will lose all of its power. A motion profile equal to one complete period of a sine wave has the needed richness.

Suppose one starts with a data batch that is comprised of carrier-phase time histories for L different GNSS satellites: for samples k = 1, …, M_j and for satellites j = 1,…, L. A standard hypothesis test develops two probability density functions for these data, one conditioned on the null hypothesis of no spoofing, H₀, and the other conditioned on the hypothesis of spoofing, H₁. The Neyman-Pearson lemma (see Further Reading) proves that the optimal hypothesis test statistic equals the ratio of these two probability densities. Unfortunately, the required probability densities depend on additional unknown quantities. In the 1-D motion case, these unknowns include the , , and coefficients, the dot product , and the direction   if one assumes that the UE attitude is unknown. A true Neyman-Pearson test would hypothesize a priori distributions for these unknown quantities and integrate their dependencies out of the two joint probability distributions. Our sub-optimum test optimally estimates relevant unknowns for each hypothesis based on the carrier-phase data, and it uses these estimates in the Neyman-Pearson probability density ratio. Although sub-optimal as a hypothesis test, this approach is usually effective, and it is easier to implement than the integration approach in the present case.

Consider the case of 1-D antenna articulation and unknown UE attitude. Maximum-likelihood calculations optimally estimate the nuisance parameters  , , and for j = 1, …, L for both hypotheses along with the unit vector for the non-spoofed hypothesis, or the scalar dot product for the spoofed hypothesis. The estimation calculations for each hypothesis minimize the negative natural logarithm of the corresponding conditional probability density. Because , , and enter the resulting cost functions quadratically, their optimized values can be computed as functions of the other unknowns, and they can be substituted back into the costs. This part of the calculation amounts to a batch high-pass filter of both the antenna motion and the carrier-phase response.

The remaining optimization problems take, under the non-spoofed hypothesis, the form:

find:          (3a)

to minimize:         (3b)

subject to:                (3c)

and, under the spoofed hypothesis, the form:

find:      η    (4a)

to minimize:         (4b)

subject to:      .   (4c)

The coefficient is a function of the deflections for k = 1, …, M_j, and the non-homogenous term is derived from the jth phase time history for k = 1, …, M_j. These two quantities are calculated during the  , , optimization. The constraint in Equation (3c) forces the estimate of the antenna articulation direction to be unit-normalized. The constraint in Eq. (4c) ensures that η is a physically reasonable dot product.

The optimization problems in Equations (3a)-(3c) and (4a)-(4c) can be solved in closed form using techniques from the literature on constrained optimization, linear algebra, and matrix factorization. The optimal estimates of and η can be used to define a spoofing detection statistic that equals the natural logarithm of the Neyman-Pearson ratio:

(5)

It is readily apparent that γ constitutes a reasonable test statistic: If the signal is being spoofed so that carrier-phase sameness is the best model, then η_opt will produce a small value of  because the spoofed-case cost function in Equation (4b) is consistent with carrier-phase sameness. The value of , however, will not be small because the plurality of directions in Equation (3b) precludes the possibility that any estimate will yield a small non-spoofed cost. Therefore, γ will tend to be a large negative number in the event of spoofing because >> is likely. In the non-spoofed case, the opposite holds true:   will yield a small value of , but no estimate of η will yield a small , and γ will be a large positive number because << .

Therefore, a sensible spoofing detection test employs a detection threshold γ_th somewhere in the neighborhood of zero. The detection test computes a γ value based on the carrier-phase data, the antenna articulation time history, and the calculations in Equations (3a)-(5). It compares this γ to γ_th. If γ ≥ γ_th, then the test indicates that there is no spoofing. If γ < γ_th, then a spoofing alert is issued.

The exact choice of γ_th is guided by an analysis of the probability of false alarm. A false alarm occurs if a spoofing attack is declared when there is no spoofing. The false-alarm probability is determined as a function of γ_th by developing a γ probability density function under the null hypothesis of no spoofing p(γ|H₀). The probability of false alarm equals the integral of p(γ|H₀) from γ = to γ = γ_th. This integral relationship can be inverted to determine the γ_th threshold that yields a given prescribed false-alarm probability

A complication arises because p(γ|H₀) depends on unknown parameters,   in the case of an unknown UE attitude and 1-D antenna motion. Although sub-optimal, a reasonable way to deal with the dependence of p(γ|,H₀) on is to use the worst-case for a given γ_th. The worst-case articulation direction maximizes the p(γ|,H₀) false-alarm integral. It can be calculated by solving an optimization problem. This analysis can be inverted to pick γ_th so that the worst-case probability of false alarm equals some prescribed value. For most actual values, the probability of false alarm will be lower than the prescribed worst case.

Given γ_th, the final needed analysis is to determine the probability of missed detection. This analysis uses the probability density function of g under the spoofed hypothesis, p(γ|η,H₁). The probability of missed detection is the integral of this function from γ = γ_th to γ = +. The dependence of p(γ|η,H₁) on the unknown dot product η can be handled effectively, though sub-optimally, by determining the worst-case probability of false alarm. This involves an optimization calculation, which finds the worst-case dot product η_wc that maximizes the missed-detection probability integral. Again, most actual η values will yield lower probabilities of missed detection.

Note that the above-described analyses rely on approximations of the probability density functions p(γ|,H₀) and p(γ|η,H₁). The best approximations include dominant Gaussian terms plus small chi-squared or non-central chi-squared terms. It is difficult to analyze the chi-squared terms rigorously. Their smallness, however, makes the use of Gaussian approximations reasonable.

We have developed and evaluated several alternative formulations of this spoofing detection method. One is the case of full 3-D b_a(t) antenna motion with unknown UE attitude. The full direction cosines matrix A is estimated in the modified version of the non-spoofed optimal fit calculations of Equations (3a)-(3c), and the full spoofing direction vector is estimated in the modified version of Equations (4a)-(4c). A different alternative allows the 1-D motion time history ρ_a(t) to have an unknown amplitude-scaling factor that must be estimated. This might be appropriate for a UAV drone with a wing-tip-mounted antenna if it induced antenna motions by dithering its ailerons. In fixed-based applications, as might be used by a financial institution, a cell-phone tower, or a power-grid monitor, the attitude would be known, which would eliminate the need to estimate or A for the non-spoofed case.

Test Results

The initial tests of our concept involved generation of simulated truth-model carrier-phase data using simulated , , and polynomial coefficients, simulated satellite LOS direction vectors for the non-spoofed cases, a simulated true spoofer LOS direction for the spoofed cases, and simulated antenna motions parameterized by and ρ_a(t). Monte-Carlo analysis was used to generate many different batches of phase data with different random phase noise realizations in order to produce simulated histograms of the p(γ|, H₀) and p(γ|η,H₁) probability density functions that are used in false-alarm and missed-detection analyses.

The truth-model simulations verified that the system is practical. A representative calculation used one cycle of an 8-Hz 1-D sinusoidal antenna oscillation with a peak-to-peak amplitude of 4.76 centimeters (exactly 1/4 of the L1 wavelength). The accumulation frequency was 1 kHz so that there were M_j = 125 carrier-phase measurements per satellite per data batch. The number of satellites was L = 6, their LOS vectors were distributed to yield a geometrical dilution of precision of 3.5, and their carrier-to-noise-density ratios spanned the range 38.2 to 44.0 dB-Hz. The worst-case probability of a spoofing false alarm was set at 10^-5 and the corresponding worst-case probability of missed detection was 1.2 ´ 10^-5. Representative non-worst-case probabilities of false alarm and missed detection were, respectively, 1.7 ´ 10^-9 and 1.1 ´ 10^-6. These small numbers indicate that this is a very powerful test. Ten-thousand run Monte-Carlo simulations of the spoofed and non-spoofed cases verified the reasonableness of these probabilities and the reasonableness of the p(γ|, H₀) and p(γ|η,H₁) Gaussian approximations that had been used to derive them.

The live-signal tests bore out the truth-model simulation results. The only surprise in the live-signal tests was the presence of significant multipath, which was evidenced by received carrier amplitude oscillations that correlated with the antenna oscillations and whose amplitudes and phases varied among the different received GPS signals. As a verification that these oscillations were caused by multipath, the only live-signal data set without such amplitude oscillations was the one taken in the NASA Wallops anechoic chamber, where one would not expect to find multipath. The multipath, however, seems to have negligible impact on the efficacy of this spoofing detection system.

FIGURES 5 and 6 show the results of typical non-spoofed and spoofed cases from WSMR live-signal tests that took place on the evening of June 19–20, 2012. Each plot shows the spoofing detection statistic γ on the horizontal axis and various related probability density functions on the vertical axis. This statistic has been calculated using a modified test that includes the estimation of two additional unknowns: an antenna articulation scale factor f and a timing bias t₀ for the decaying sinusoidal oscillation . The damping ratio ζ and the undamped natural frequency w_n are known from prior system identification tests.

Figure 5. Spoofing detection statistic, threshold, and related probability density functions for a typical non-spoofed case with live data.

Figure 6. Performance of a typical spoofed case with live data: spoofing detection statistic, threshold, and related probability density functions.

The vertical dashed black line in each plot shows the actual value of γ as computed from the GPS data. There are three vertical dash-dotted magenta lines that lie almost on top of each other. They show the worst-case threshold values γ_th as computed for the optimal and ±2σ estimates of t₀: t_0opt, t_0opt+2σ_{t_0opt}, and t_0opt-2σ_{t_0opt}. They have been calculated for a worst-case probability of false alarm equal to 10^-6. An ad hoc method of compensating for the prototype system’s t₀ uncertainty is to use the left-most vertical magenta line as the detection threshold γ_th. The vertical dashed black line lies very far to the right of all three vertical dash-dotted magenta lines in Figure 5, which indicates a successful determination that the signals are not being spoofed. In Figure 6, the situation is reversed. The vertical dashed black line lies well to the left of the three vertical dash-dotted magenta lines, and spoofing is correctly and convincingly detected.

These two figures also plot various relevant probability density functions. Consistent with the consideration of three possible values of the t₀ motion timing estimate, these are plotted in triplets. The three dotted cyan probability density functions represent the worst-case non-spoofed situation, and the dash-dotted red probability functions represent the corresponding worst-case spoofed situations. Obviously, there is sufficient separation between these sets of probability density functions to yield a powerful detection test, as evidenced by the ability to draw the dash-dotted magenta detection thresholds in a way that clearly separates the red and cyan distributions. Further confirmation of good detection power is provided by the low worst-case probabilities of false alarm and missed detection, the latter metric being 1.6 ´ 10^-6 for the test in Figure 5 and 7 ´ 10^-8 for Figure 6.

The solid-blue distributions on the two plots correspond to the η_opt estimate and the spoofed assumption, which is somewhat meaningless for Figure 5, but meaningful for Figure 6. The dashed-green distributions are for the estimate under the non-spoofed assumption. The wide separations between the blue distributions and the green distributions in both figures clearly indicate that the worst-case false-alarm and missed-detection probabilities can be very conservative.

The detection test results in Figures 5 and 6 have been generated using the last full oscillation of the respective carrier-phase data, as in Figures 3 and 4, but applied to different data sets. In Figure 3, the last full oscillation starts at t = 3.43 seconds, and it starts at t = 2.11 seconds in Figure 4. The peak-to-peak amplitude of each last full oscillation ranged from 4-6 centimeters, and their periods were shorter than 0.5 seconds. It would have been possible to perform the detections using even shorter data spans had the mechanical oscillation frequency of the cantilevered antenna been higher.

Conclusions

In this article, we have presented a new method to detect spoofing of GNSS signals. It exploits the effects of intentional high-frequency antenna motion on the measured beat carrier phases of multiple GNSS signals. After detrending using a high-pass filter, the beat carrier-phase variations can be matched to models of the expected effects of the motion. The non-spoofed model predicts differing effects of the antenna motion for the different satellites, but the spoofed case yields identical effects due to a geometry in which all of the false signals originate from a single spoofer transmission antenna. Precise spoofing detection hypothesis tests have been developed by comparing the two models’ ability to fit the measured data.

This new GNSS spoofing detection technique has been evaluated using both Monte-Carlo simulation and live data. Its hypothesis test yields theoretical false-alarm probabilities and missed-detection probabilities on the order of 10^-5 or lower when working with typical numbers and geometries of available GPS signals and typical patch-antenna signal strengths. The required antenna articulation deflections are modest, on the order of 4-6 centimeters peak-to-peak, and detection intervals less than 0.5 seconds can suffice.

A set of live-signal tests at WSMR evaluated the new technique against a sophisticated receiver/spoofer, one that mimics all visible signals in a way that foils standard RAIM techniques. The new system correctly detected all of the attacks. These are the first known practical detections of live-signal attacks mounted against a civilian GNSS receiver by a dangerous new generation of spoofers.

Future Directions

This work represents one step in an on-going “Blue Team” effort to develop better defenses against new classes of GNSS spoofers. Planned future improvements include 1) the ability to use electronically synthesized antenna motion that eliminates the need for moving parts, 2) the re-acquisition of true signals after detection of spoofing, 3) the implementation of real-time prototypes using software radio techniques, and 4) the consideration of “Red-Team” counter-measures to this defense and how the “Blue Team” could combat them; counter-measures such as high-frequency phase dithering of the spoofed signals or coordinated spoofing transmissions from multiple locations.

Acknowledgments

The authors thank the following people and organizations for their contributions to this effort: The NASA Wallops Flight Facility provided access to their anechoic chamber. Robert Miceli, a Cornell graduate student, helped with data collection at that facility. Dr. John Merrill and the Department of Homeland Security arranged the live-signal spoofing tests. The U.S. Air Force 746th Test Squadron hosted the live-signal spoofing tests at White Sands Missile Range. Prof. Todd Humphreys and members of his University of Texas at Austin Radionavigation Laboratory provided live-signal spoofing broadcasts from their latest receiver/spoofer.

Manufacturers

The prototype spoofing detection data capture system used an Antcom Corp. (www.antcom.com) 2G1215A L1/L2 GPS antenna. It was connected to an Ettus Research (www.ettus.com) USRP (Universal Software Radio Peripheral) N200 that was equipped with the DBSRX2 daughterboard.

MARK L. PSIAKI is a professor in the Sibley School of Mechanical and Aerospace Engineering at Cornell University, Ithaca, New York. He received a B.A. in physics and M.A. and Ph.D. degrees in mechanical and aerospace engineering from Princeton University, Princeton, New Jersey. His research interests are in the areas of GNSS technology, applications, and integrity, spacecraft attitude and orbit determination, and general estimation, filtering, and detection.

STEVEN P. POWELL is a senior engineer with the GPS and Ionospheric Studies Research Group in the Department of Electrical and Computer Engineering at Cornell University. He has M.S. and B.S. degrees in electrical engineering from Cornell University. He has been involved with the design, fabrication, testing, and launch activities of many scientific experiments that have flown on high altitude balloons, sounding rockets, and small satellites. He has designed ground-based and space-based custom GPS receiving systems primarily for scientific applications.

BRADY W. O’HANLON is a graduate student in the School of Electrical and Computer Engineering at Cornell University. He received a B.S. in electrical and computer engineering from Cornell University. His interests are in the areas of GNSS technology and applications, GNSS security, and GNSS as a tool for space weather research.

VIDEO

Here is a video of Cornell University’s antenna articulation system for the team’s first prototype spoofing detector tests.

FURTHER READING

• The Spoofing Threat and RAIM-Resistant Spoofers

“Status of Signal Authentication Activities within the GNSS Authentication and User Protection System Simulator (GAUPSS) Project” by O. Pozzobon, C. Sarto, A. Dalla Chiara, A. Pozzobon, G. Gamba, M. Crisci, and R.T. Ioannides, in Proceedings of ION GNSS 2012, the 25th International Technical Meeting of The Institute of Navigation, Nashville, Tennessee, September 18–21, 2012, pp. 2894-2900.

“Assessing the Spoofing Threat” by T.E. Humphreys, P.M. Kintner, Jr., M.L. Psiaki, B.M. Ledvina, and B.W. O’Hanlon in GPS World, Vol. 20, No. 1, January 2009, pp. 28-38.

Vulnerability Assessment of the Transportation Infrastructure Relying on the Global Positioning System – Final Report. John A. Volpe National Transportation Systems Center, Cambridge, Massachusetts, August 29, 2001.

• Moving-Antenna and Multi-Antenna Spoofing Detection

“Robust Joint Multi-Antenna Spoofing Detection and Attitude Estimation by Direction Assisted Multiple Hypotheses RAIM” by M. Meurer, A. Konovaltsev, M. Cuntz, and C. Hattich, in Proceedings of ION GNSS 2012, the 25th International Technical Meeting of The Institute of Navigation, Nashville, Tennessee, September 18–21, 2012, pp. 3007-3016.

“GNSS Spoofing Detection for Single Antenna Handheld Receivers” by J. Nielsen, A. Broumandan, and G. Lachapelle in Navigation, Vol. 58, No. 4, Winter 2011, pp. 335-344.

• Alternate Spoofing Detection Strategies

“Who’s Afraid of the Spoofer? GPS/GNSS Spoofing Detection via Automatic Gain Control (AGC)” by D.M. Akos, in Navigation, Vol. 59, No. 4, Winter 2012-2013, pp. 281-290.

“Civilian GPS Spoofing Detection based on Dual-Receiver Correlation of Military Signals” by M.L. Psiaki, B.W. O’Hanlon, J.A. Bhatti, D.P. Shepard, and T.E. Humphreys in Proceedings of ION GNSS 2011, the 24th International Technical Meeting of The Institute of Navigation, Portland, Oregon, September 19–23, 2011, pp. 2619-2645.

• Statistical Hypothesis Testing

Fundamentals of Statistical Signal Processing, Volume II: Detection Theory by S. Kay, published by Prentice Hall, Upper Saddle River, New Jersey,1998.

An Introduction to Signal Detection and Estimation by H.V. Poor, 2nd edition, published by Springer-Verlag, New York, 1994.

June 1, 2013