Unicore has launched its next-generation quad-system GNSS module, the UM482.
The UM482 is a multi-frequency high-precision heading module with a small footprint, supporting the satellite signals BDS B1/B2, GPS L1/L2, GLONASS L1/L2, Galileo E1/ E5b and SBAS.
The module is designed for applications such as robotics, drones, intelligent drives and mechanical control.
1-cm RTK positioning accuracy and 0.2-degree heading accuracy with 1-m baseline
Dual antenna input with support of antenna signal detection
Supporting simultaneous output of heading and positioning, 20-Hz data output rate
Adaptive recognition of RTCM input data format
On-board micro-electro-mechanical system (MEMS) integrated navigation
The UM482 GNSS RTK module adopts Unicore’s new-generation Nebulas II chip and UGypsophila real-time kinematic (RTK) algorithm.
Based on high performance data-sharing technology and the simplified operation system of the Nebulas II chip, the UGypsophila RTK algorithm dramatically optimizes matrix processing, the company said. It can involve all satellites from GPS, BDS, GLONASS and Galileo in RTK and heading processing, shorten RTK and heading initialization time to 5 seconds and significantly improve the reliability and accuracy of RTK and heading.
Furthermore, the UM482 integrates the onboard MEMS chip and U-Fusion integrated navigation algorithm, resulting in optimized continuity and reliability of accurate heading and positioning output in tough environments such as city canyons, tunnels and overpasses. Inputs of odometer and external higher performance inertial components are supported.
The UM482, along with all the UM and UB family of receivers, will be on display at booth B4018 for the duration of the Intergeo 2017 trade show, which takes place Sept. 26-28 at Berlin Exhibition Center, Berlin, Germany.
Measure, a U.S. provider of drone services to enterprise customers, has added turnkey wind farm inspection capabilities to its portfolio of aerial data collection solutions.
Wind farm operators can outsource preventive maintenance inspections to Measure’s drone pilots and data analysts for fast, accurate, safe and timely problem identification. The service helps avert critical turbine failures and efficiency losses while reducing repair downtime and its associated revenue impact.
The company’s drone inspection solution has already been used to successfully examine more than 400 MW of wind farms. The package spans all inspection and reporting functions, including state-of-the-art drone equipment, safe and insured flights by experienced drone pilots, efficient data processing that pinpoints both blade damage and severity, and damage reports and analytics available through a secure online portal.
Dry Lake Wind Power Project, Arizona (Photo: U.S. DOE)
Measure’s new wind farm inspection solution expands the company’s services to the renewable energy sector, which also include a robust suite of drone inspection solutions for solar plants that was announced in July.
The suite includes solar-panel inspections, drone-based site overview and maintenance, site shading and terrain analysis, thermal inverter scans, tracker misalignment detection and vegetation management analysis.
Benefits of Measure’s drone-based blade and tower inspections include:
75% faster inspections than other methods, averaging 30 minutes or less per turbine compared to as much as two hours for manned inspections. This reduces excessive time commitments and allows large wind farms to be inspected more frequently. It also reduces labor costs for inspection and frees employees for other tasks.
Decreased injury risk in the field, with no threat of falls to inspectors climbing turbine structures or blades.
Better defect and damage detection because drones get closer to turbine blades than ground cameras, capturing clearer images. Undetected defects on the blades can result in continuous efficiency losses as high as 6% and associated revenue loss of up to $10,000 annually per turbine.
Maximized turbine availability and revenue generation through early problem detection that helps prevent critical failures and associated downtime for repairs.
Actionable data, including classified damage reports and historical portfolio analysis documenting turbine defects, failure rates and efficiency losses over time. Damage reports can be customized to display only the information needed by blade repair technicians with a few clicks.
“Many wind farms don’t inspect their turbines on a preventive maintenance basis, and those that do use ground crews with conventional cameras and zoom lenses. Under both conditions, there is a risk of failing to detect turbine damage or structural defects on blades that can worsen over time and lead to a catastrophic failure,” said Harjeet Johal, Measure vice president of energy infrastructure and a 10-year veteran of the renewable energy industry with a Ph.D. in electrical engineering. “Our drone-based inspections provide multiple advantages that can help wind farm operators operate at peak capacity.”
“Our global wind portfolio is currently 1,033 MW with 877 MW in the U.S. alone. Knowing the health of our wind assets is essential for us to provide reliable power to our customers,” said Adam Brown, U.S. Drone Program Lead at The AES Corporation, a Fortune 200 global power company. “Using drones to inspect the blades and towers makes it safer for our people as they can stay firmly on the ground while still being able to inspect, at scale, hundreds of wind turbines to ensure they have the highest availability.”
Post-Irma hurricane damage is captured in aerial imagery by EagleView.
EagleView Technologies has captured post event aerial imagery of two million properties in the state of Florida following Hurricane Irma.
EagleView is a provider of aerial imagery and property data analytics for government agencies, insurance carriers and other private-sector organizations,
With an image library dating back to 2002 in the state of Florida, EagleView is able to provide emergency services, public safety agencies, property assessors and county GIS departments with ample imagery from before and after Hurricane Irma occurred. Combining high-resolution imagery and advanced machine learning capabilities, EagleView can identify the severity of property damage following a hurricane or other extreme weather event.
“Hurricane Irma inflicted severe damage on properties all over Florida and affected millions of people throughout the state,” said EagleView President Rishi Daga. “With a view of more than two million properties in Florida, we are assisting the agencies that use our imagery with their efforts, so they can continue to help all of those who have been affected.”
The two million properties have been photographed via specialized camera rigs in fixed-wing aircraft. The images are taken from an orthogonal (top-down) perspective as well as at oblique angles from all four cardinal directions. Oblique aerial imagery enables insurance claims adjusters to view all sides of a home’s exterior and gives emergency response crews greater insight into the storm’s effects in their communities.
“Our goal was to begin capturing and processing imagery as soon as possible to assist in recovery efforts, and we have done so at record speed,” said Jay Martin, Senior Vice President of Operations at EagleView. “Our next phase is to put boots on the ground and complete property inspections up close using drones as part of our EagleView OnSite solution.”
Post-hurricane image capture and processing will continue to take place throughout the upcoming weeks.
EagleView is completing the phase of image capture via fixed-wing aircraft and will soon move in to completing property inspections with the use of unmanned aerial systems (UAS), bringing post-event data directly to insurance claims adjusters.
As of Sept. 18, thousands of drone inspections have been scheduled through Friday, Sept. 22.
Antenova Ltd. has launched the Robusta GNSS antenna for tracking applications and smart cities. The Robusta is a very low-profile antenna in a new patent design for metal surfaces.
Antenova, manufacturer of antennas and RF antenna modules for M2M and the Internet of Things, launched the Robusta (part no. SR4G031) at the GSMA Mobile World Congress Americas show, held Sept. 12-14 in San Francisco.
The antenna operates in the 1559-1609 MHz bands and is designed for tracking metal objects and smart city applications.
The Reflector family is designed to answer to the challenge of operating on a metal surface or housing, where it is extremely difficult for an antenna to operate. The Reflector antennas use a patented new technology with two layers. The first layer is electrically isolated from the second layer to provide RF shielding to the second layer. This allows the antenna to radiate effectively in the direction pointing away from the base material.
The Robusta antenna has two key features for discreet tracking. It is extremely low profile so it can be mounted onto a metal object such as a bicycle frame or concealed under a label. Being able to operate directly on a metal surface, it can be used on bicycles, motorcycles, vehicles, containers or other property that needs to be tracked and located accurately.
The Robusta antenna is also a good choice for smart lighting and smart meters in smart city applications, where it can be fixed to metal fittings.
The antenna is manufactured from rigid FR4 material and measures 23 x 16 x 1.7 millimeters high, and comes with 100-millimeter or 150-millimeter cable and IPEX MHF connector and an adhesive pad for easy integration into a device.
Antenova provides resources on its website to help with integration.
The NTP (Network Time Protocol) Reflector is a fast, accurate NTP server. It features denial of service resilience, monitoring and notification functions.
Characteristics include 100 percent hardware NTP time-stamping for accuracy and high performance; NTP packet monitoring for DoS detection; bandwidth limiting and packet filtering for CPU protection; and alarming if NTP loading is above expected levels.
To help users better understand the advantages of Microsemi’s NTP Reflector and packet limiting/monitoring technology, the company explains the underlying technology and its security benefits in the new application note, available for download.
Key Characteristics
100 percent hardware NTP timestamping for accuracy and high performance
NTP Packet monitoring for DoS detection
Bandwidth limiting and packet filtering for CPU protection
Alarming if NTP loading is above expected levels.
The NTP Reflector is one of the many differentiating features of Microsemi’s new SyncServer S600 series network time servers.
The reflector is a real-time, hardware-based NTP packet identification and time-stamping engine uniquely designed to protect the SyncServer CPU from excessive network traffic denial of service attacks and notify the operator if NTP traffic is above expected levels.
The innovative technology enables extremely high-bandwidth, high-accuracy, high-reliability and security-hardened NTP operations.
Tersus GNSS Inc. has released a new AutoSteer autopilot for agricultural machinery.
The AG960 AutoSteer System is designed to accelerate the application of autopilot for precision agricultural machinery and enhance and optimize operational accuracy and productivity for modern farmers.
By integrating high-precision real-time kinematic (RTK) receiver and software, the AG960 enables agricultural machines to operate in accordance with a pre-set planning path. Using precise GNSS guidance, the hydraulic system of the agricultural machinery is steered by the vehicle controller.
Agricultural machines can operate aligned with the set route automatically, while graphical detailsare displayed on the vehicle display panel. The system is easy to use and applicable for each working cycle of agriculture, such as soil tillage, plowing, building of ditches and ridges, seeding, spraying and harvesting.
Tersus plans to launch a series of solutions that meet the requirements of different farming machines. The AG960 was first commercially deployed in China, and will be rolled out in other regions around the world.
Remote sensing of atmospheric gas concentrations is important in monitoring global greenhouse gas levels and industry monitoring. Monitoring is usually carried out via satellite sensing or laborious ground-based measurements.
With aerial measurement, a wider area can be measured efficiently, and repeat measurements taken of days, weeks and months gathering time-series data.
Custom drone with gas sensor (circled).
This spring, a study by QuestUAV and the British Geological Survey (BGS) used a custom QuestUAV Q200 airframe equipped with two sensors, one tuned for methane (CH4) and one for CO2. The sensors use an open-path gas mass spectrometer — a fiber-guided laser beam passed laterally across open atmosphere on top of the drone to a reflector and then back to the sensor itself.
Signals from the sensors were fed into a multi-core processing unit on board the drone. All readings were stamped with time and location provided by the standard GPS and flight units in the Q200.
The completed drone was commissioned in March. Over several months, trial flights were run over gas releases initiated manually on the ground over the test site. The recorded sensor data was processed immediately on return to base, and the data passed to BGS for analysis and appraisal.
The team plans to fine-tune the operational workflow and maintenance tasks for regular missions.
Public Services and Procurement Canada has awarded a contract to Ottawa-based Kongsberg Geospatial for an emergency operations airspace UAV tracking system.
Kongsberg Geospatial, an Ottawa-based developer of geospatial software technology, was awarded the contract to produce an Emergency Operations Airspace Management System (EOAMS) for evaluation by Canadian government agencies for safely managing drones at emergency and disaster scenes.
The contract was awarded via a competitive request for proposals under the Canadian Safety and Security Program in a project for Defense R&D Canada’s Centre for Security Science.
A small UAV is shown surveying the movement of a forest fire. The EOAMS would allow first responders to deploy drones at disaster scenes without endangering other emergency response aircraft or commercial flights. (Photo illustration: Kongsberg Geospatial)
The EOAMS is a portable display that interfaces with a variety of local sensors, including radar and Automatic Dependence Surveillance — Broadcast (ADS-B) receivers to give a clear picture of the airspace around disaster areas.
The system is intended to allow first responders to safely use unmanned aerial vehicles (UAVs) to survey the area, without risking collision with other emergency aircraft, including water bombers or rescue and police helicopters.
The system would also provide a warning to first responders if unapproved UAVs approach the area – providing a degree of protection against what is becoming an increasing problem with the proliferation of small consumer camera drones at fires and accident scenes.
The Government of Canada is expected to begin flight operations testing with the new Emergency Operations Airspace Management System in the summer of 2018.
“Securing and managing the airspace around disaster scenes or at big public events is becoming a real concern for all levels of government,” said Paige Cutland, IRIS program director for Kongsberg Geospatial. “Even if a drone operator isn’t acting with malicious intent, they have the potential to cause considerable harm if, for example, they fly into the path of an air ambulance. We need effective tools to help prevent this while also allowing legitimate UAV operations to be safely integrated into the emergency airspace.”
The new EOAMS will be based on Kongsberg Geospatial’s IRIS UAS airspace visualization system. The IRIS spatial awareness system evolved from technology originally developed for air traffic management display systems, and for supporting flight operations for military UAV systems like the U.S. Navy Triton Global Hawk.
The system has been developed for safely operating UAVs beyond visual line-of-sight (BVLOS), and has been adopted by the FAA ASSURE group for use in research toward developing regulations for commercial BVLOS operations in the United States.
“Kongsberg Geospatial has been pioneering innovation in airspace management for unmanned aircraft for over a decade,” said Ranald McGillis, president of Kongsberg Geospatial. “With the EOAMS project, we have the opportunity to introduce some really exciting capabilities in a portable system that will help first responders use UAVs in new and effective ways to support emergency response efforts.”
A GPS-lidar fusion technique implements a novel method for efficiently modeling lidar-based position error covariance based on features in the point cloud. The fusion uses a three-dimensional (3D) city model to detect and eliminate non-line-of-sight (NLOS) GPS satellites to improve global positioning.
The technique has potential application in UAV missions such as 3D modeling, filming, surveying, search and rescue, and package delivery.
By Akshay Shetty and Grace Xingxin Gao, University of Illinois
Unmanned aerial vehicles (UAVs) commonly rely on GPS for continuous and accurate outdoor position estimates. However, in certain urban scenarios, additional onboard sensors such as light detection and ranging (lidar) are desirable due to errors in GPS measurements. To fuse these measurements it is important, yet challenging, to accurately characterize their error covariance. We propose a GPS-lidar fusion technique with a novel method for efficiently modeling the position error covariance based on surface and edge features in point clouds. We use the lidar point clouds in two ways: to estimate incremental motion by matching consecutive point clouds; and, to estimate global pose (position and orientation) by matching with a 3D city model. For GPS measurements, we use the 3D city model to eliminate NLOS satellites and model the measurement covariance based on the received signal-to-noise-ratio (SNR) values. Finally, all the above measurements and error covariance matrices are input to an unscented Kalman Filter (UKF), which estimates the globally referenced pose of the UAV. To validate our algorithm, we conducted UAV experiments in GPS-challenged urban environments on the University of Illinois at Urbana-Champaign campus.These experiments demonstrate a clear improvement in the UAV’s global pose estimates using the proposed sensor fusion technique.
SITUATION
Emerging applications in UAVs such as 3D modeling, filming, surveying, search and rescue, and package delivery all involve flying in urban environments. In these scenarios, autonomously navigating a UAV has certain advantages such as optimizing flight paths and sensing and avoiding collisions. However, to enable such autonomous control, we need a continuous and reliable source for UAV positioning. In most cases, GPS is primarily relied on for outdoor positioning. However, in an urban environment, GPS signals from the satellites are often blocked or reflected by surrounding structures, causing large errors in the position output.
In cases when GPS is unreliable, additional onboard sensors such as lidar can provide the navigation solution. An onboard lidar provides a real-time point cloud of the surroundings of the UAV. In a dense urban environment, lidar can detect a large number of features from surrounding structures such as buildings.
Positioning based on lidar point clouds has been demonstrated primarily by applying different simultaneous localization and mapping (SLAM) algorithms. In many cases, algorithms implement variants of iterative closest point (ICP) to register new point clouds.
APPROACH
The main contribution of this article is a GPS-lidar fusion technique with a novel method for efficiently modeling the error covariance in position measurements derived from lidar point clouds. Figure 1 shows the different components involved in the sensor fusion.
Figure 1. Overview of sensor fusion architecture.
We use the lidar point clouds in two ways: to estimate incremental motion by matching consecutive point clouds; and, to estimate global pose by matching with a 3D city model. We use ICP for matching the point clouds in both cases.
For the lidar-based position estimates, we proceed to build the error covariance model depending on the surrounding point cloud. First, we extract surface and edge feature points from the point cloud. We then model the position error covariance based on these individual feature points. Finally, we combine all the individual covariance matrices to model the overall position error covariance ellipsoid.
For the GPS measurement model, we use the pseudorange measurements from a stationary reference receiver and an onboard GPS receiver to obtain a vector of double-difference measurements. Using the double-difference measurements eliminates clock bias and atmospheric error terms, hence reducing the number of unknown variables. We use the global position estimate from the lidar-3D city matching to construct LOS vectors to all the detected satellites. We then use the 3D city model to detect NLOS satellites, and consequently refine the double-difference measurement vector. We create a covariance matrix for the GPS double-difference measurement vector based on SNR of the individual pseudorange measurements.
We implement a UKF to integrate all lidar and GPS measurements. Additionally, we incorporate orientation, orientation rate and acceleration measurements from an onboard inertial measurement unit (IMU). Finally, we test the filter on an urban dataset to show an improvement in the navigation solution.
LIDAR-BASED ODOMETRY
ICP is commonly used for registering three-dimensional point clouds. It takes a reference point cloud q, an input point cloud p, and estimates the rotation matrix R and the translation vector T between the two point clouds. Different variants of the algorithm generally consist of three primary steps.
Matching. This involves matching each point pi in the input point cloud, to a point qi in the reference point cloud. The most common method is to find the nearest neighbors of each point in the input point cloud. For our application, a kDtree performs best since the two point clouds are relatively close to each other.
Defining Error Metric. This defines the error metric for the point pairs. We choose the point-to-point metric, which is generally more robust to difficult geometry than other metrics such as point-to-plane. The total error between the two point clouds is defined as follows:
(1)
where N is the number of points in the input point cloud p.
Minimization. The last step of the algorithm is the minimization of the error metric with regard to the rotation matrix R and the translation vector T between the two point clouds.
We use ICP to estimate the incremental motion of the lidar between consecutive point clouds. Figure 2 shows our implementation of ICP to estimate the lidar odometry.
Figure 2. The input to ICP is a reference point cloud q and an input point cloud p as shown in (a). The algorithm calculates the rotation matrix R and the translation vector T such that the error metric E is minimized. (b) shows the reference point cloud q and the transformed input point cloud R • p + T.
MATCHING LIDAR, 3D MODEL
We generate our 3D city model using data from two sources: Illinois Geospatial Data Clearinghouse and OSM. The Illinois Geospatial Data were collected by a fixed-wing aircraft flying at an altitude of 1700 meters, equipped with a lidar system including a differential GPS unit and an inertial measurement system to provide superior global accuracy. Since the data were collected from a relatively high altitude, it primarily contains adequate details for the ground surface and the building rooftops. In order to complete the 3D city model, we need additional information for the sides of buildings. We use OSM to obtain this information. OSM is a freely available, crowd-sourced map of the world, which allows users to obtain information such as building footprints and heights. Figure 3 shows a section of the 3D city model for Champaign County.
Figure 3. Section of the point cloud for Champaign County dataset. (Left) shows the 3D city model using only the Illinois Geospatial Data. (Right) fhows the model after incorporating building information from OpenStreetMap.
To estimate the global pose of the lidar, we match the onboard lidar point cloud with the 3D city model using ICP, in these steps:
Use the position output from onboard GPS receiver as an initial guess. If position output is unavailable, use the position estimate from the previous iteration as an initial guess. For orientation, use the estimate from the previous iteration. Thus, we obtain an initial pose guess .
Project the onboard lidar point cloud pL to the same space as the 3D city model qcity using .
Implement ICP, to obtain the rotation RL and translation TL between the two point clouds. Use this output to obtain an estimate for the global pose .
Figure 4 shows the results of implementation of the above method. While navigating in urban areas, the GPS receiver position output used for the initial pose guess might contain large errors in certain directions. This might cause ICP to converge to a local minimum, depending on features in the point cloud pL generated by the onboard lidar.
Figure 4. Global pose estimation with the aid of 3D city model. (Left) shows the intial position guess (red dot, with term in red outlined box) and the onboard lidar point cloud pL projected on the same space as the 3D city model qcity. (b) shows the updated global position (green dot, with term in green outlined box) after the ICP step. We see an improvement in the global position, as the point cloud matches with the 3D city model.
To evaluate how our lidar-3D city model matching algorithm performs in such challenging cases, we test it in two different urban areas as shown in Figure 5. We begin by selecting a grid of initial position guesses up to 20 meters away from the true position. With an adequate distribution of features, ICP is able to correctly match the two point clouds and provide an accurate position estimate after matching. In contrast, when there’s an urban scenario with a relatively poor distribution of features, ICP is unable to estimate the position accurately.
Figure 5. Lidar-3D city model matching in two different urban areas. We begin with a grid on initial position guesses (red) around the true position (black). In (a) and (b), there are adequate features. The position estimates after matching (blue) converge to the true position. In (c) and (d) the feature distribution is relatively poor. The position estimates after matching (blue) are parallel to the building surface.
MODELING ERROR COVARIANCE
We model the lidar position error covariance as a function of the surrounding features. In urban environments, we typically observe structured objects such as buildings, hence we focus primarily on surface and edge features in the point cloud. We extract these feature points based on the curvature at each point. Points with curvature values above a threshold are marked as edge points, whereas points with curvature values below a threshold are marked as surface points. (For detailed discussion of the algorithms involved, see GPS-LiDAR_AkshayShetty-algorithms.
For each surface feature point, we first compute the normal by using 9 of the neighboring points to fit a plane. We model the error covariance ellipsoid with the hypothesis that each surface feature point contributes in reducing position error in the direction of the corresponding surface normal. Additionally, we assume that surface points closer to the lidar are more reliable than those further away, because of the density of points.
For each edge feature point, we first find the direction of the edge using the closest edge points in the scans above and below. We model the error covariance ellipsoid with the hypothesis that each edge feature point helps in reducing position error in the directions perpendicular to the edge vector. A vertical edge, for example, would help in reducing horizontal position error. Additionally, we assume that edge points closer to the lidar are more reliable than those further away, again because of the density of points. Figure 6 shows sample error covariance ellipsoids for a surface point and an edge point.
Figure 6. Position error covariance ellipsoid for surface and edge feature points. The exact sizes of the ellipsoids are tuned during implementation.
To obtain the overall position error covariance, we combine the error covariance matrices for all the individual surface and edge feature points. Figure 7 shows the combined covariance ellipsoid for two different scenarios. We observe that while passing through a corridor, the covariance ellipsoid is larger in the direction parallel to the building sides due to a poor distribution of features.
Figure 7. Overall position error covariance ellipsoids (black) for two point clouds (green). We combine the error ellipsoids from individual surface (red) and edge (blue) feature points.
GPS MEASUREMENT MODEL
We use pseudorange measurements from the GPS receiver to create the measurement model. To eliminate certain error terms, we use double-difference pseudorange measurements, which are calculated by differencing the pseudorange measurements between two satellites and between two receivers. Before proceeding to use the pseudorange measurements, we check if any of the satellites detected by the receiver are NLOS signals. We use the 3D city model mentioned earlier to detect the NLOS satellites. We use the position output generated by the lidar-3D city model matching to locate the receiver on the 3D city model.
Next, we draw LOS vectors from the receiver to every satellite detected by the receiver and eliminate satellites whose corresponding LOS vectors intersect the 3D city model. Figure 8 shows the above implementation in an urban scenario.
Figure 8. Elimination of NLOS satellite signals. LOS vectors are drawn to all detected satellites: SV3, SV14, SV16, SV22, SV23, SV26, SV31. The LOS vectors to satellites SV23 and SV31 intersect (red) the 3D city model and are eliminated from further calculations.
After eliminating the NLOS satellites, we select satellites that are visible to both the user and the reference receivers to create the GPS double-difference measurement vector and its covariance. We assume that the individual pseudorange measurements are independent, and that the variance for each measurement is a function of the corresponding SNR. We propagate the covariance matrix for the individual pseudorange measurements, to obtain the covariance matrix for the double-difference measurements.
GPS-Lidar Integration
In addition to using a lidar and a GPS receiver, we use an IMU on board the UAV. Figure 9 shows the experimental setup: the UAV designed and built by our research group. For the double-difference GPS measurements, we use a reference receiver within a kilometer of our data collection sites. We implement a UKF to fuse measurements from the sensors and estimate the global pose of the UAV.
Figure 9. Experimental setup for data collection. Our custom-made iBQR UAV mounted with a lidar, a GPS receiver and antennas, an IMU, and an onboard computer.
Position and orientation estimates from lidar and GPS are incorporated via the correction step of the filter, whereas the IMU measurements are included in the prediction step. For position corrections from lidar, we use our point cloud feature based model for the error covariance. For GPS double-difference measurements, we use the covariance based on the individual pseudorange measurement SNR.
We implement our algorithm on an urban dataset collected on our campus of University of Illinois at Urbana-Champaign. As shown in Figure 10, the GPS measurements and the GPS position output contain large errors, due to the presence of nearby urban structures. Here we stack all the double-difference measurements and compute the unweighted least square estimate of the baseline between the UAV and the reference receiver.
Figure 10. Position estimates from GPS measurements. The position output from the GPS receiver (blue) and the unweighted least-squares position estimate (red) contain large errors.
For the lidar measurements, we check the output from our incremental ICP odometry method and the lidar-3D city model matching algorithm. Furthermore, we implement an ICP mapping algorithm to check the performance of existing ICP-based methods on the dataset. In Figure 11, the ICP odometry method and the ICP mapping algorithm accumulate drift over the course of the trajectory. The lidar-3D city model matching algorithm does not drift over time; however, the position still contains errors in situations where the lidar does not detect enough number of points or the matching algorithm converges to a local minimum.
Figure 11. Position estimates from lidar point clouds. The incremental ICP odometry (green) and the ICP mapping (blue) estimates accumulate drift over time. The lidar-3D city model matching (yellow) does not drift over time, but contains errors where the ICP algorithm might converge to a local minimum.
Figure 12 shows the output of the filter for the same trajectory. The filter output estimates the actual path much more accurately than the individual measurement sources by themselves.
Figure 12. Position estimates from UKF, integrating GPS and lidar measurements. The filter position output (blue) resembles the actual trajectory, more accurately than any individual source of GPS or lidar measurements.
CONCLUSION
In summary, we proposed a GPS-lidar integration approach for estimating the navigation solution of UAVs in urban environments. We used the onboard lidar point clouds in two ways: to estimate the odometry by matching consecutive point clouds, and to estimate the global pose by matching with an external 3D city model. We built a model for the error covariance of the lidar-based position estimates as a function of surface and edge feature points in the point cloud. For GPS measurements, we eliminated NLOS satellites using the 3D city model and used the remaining double-difference measurements between an onboard receiver and a reference receiver. To construct the covariance matrix for the double-difference measurements, we used the SNR values for individual pseudorange measurements.
Finally, we applied an UKF to integrate the measurements from lidar, GPS and an IMU. We experimentally demonstrated the improved positioning accuracy of our filter.
ACKNOWLEDGMENTS
The authors would like to sincerely thank Kalmanje Krishnakumar and his group at NASA Ames Research Center for supporting this work under the grant NNX17AC13G.
The material in this article was first presented at the ION GNSS+ 2017 conference in Portland, September 2017.
MANUFACTURERS
The lidar used aboard the UAV in these tests is a Velodyne VLP-16 Puck Lite. The GPS receiver is a u-blox LEA-6T with a Maxtena M1227HCT-A2-SMA antenna. The IMU is an Xsens Mti-30, and the onboard computer an AscTec Mastermind 3a.
The iBQR UAV was designed and assembled by the authors.
AKSHAY SHETTY received an M.S. degree in aerospace engineering from University of Illinois at Urbana-Champaign. He is also pursuing a Ph.D. at the same university.
GRACE XINGXIN GAO received a Ph.D. degree in electrical engineering from Stanford University. She is an assistant professor in the Aerospace Engineering Department at the University of Illinois at Urbana-Champaign.
In recent years there has been a proliferation of software-defined radio (SDR) data-collection systems and processing platforms designed for GNSS receiver applications or those that support GNSS bands. For post-processing, correctly interpreting the GNSS SDR sampled datasets produced or consumed by these systems has historically been a cumbersome and error-prone process.
This is because these systems necessarily produce datasets of various formats, the subtleties of which are often lost in translation when communicating between the producer and consumer of these datasets.
This specification standardizes the metadata associated with GNSS SDR sampled data files and the layout of the binary sample files.
The GNSS SDR Metadata Standard defines parameters and schema to express the contents of SDR sample data files. The standard is designed to promote the interoperability of GNSS SDR data collection systems and processors.
The metadata files are human readable and in XML format. A compliant open source C++ API for reading metadata and binary samples is also officially supported to promote ease of integration into existing SDR systems.
Review the formal standards document and click on Submit a Comment to provide feedback. Comments will be accepted through December 31, 2017.
Surrounding sounds may not be a common way of determining location. But on the battlefield, warfighters need to know the direction of gunshots to enable a proper response.
Weighing 12 ounces, the Boomerang Warrior-X by Raytheon BBN Technologies provides immediate hostile fire location awareness to individual soldiers and gives unit leaders shooter grid coordinates, according to the company. These situational awareness enhancements improve coordinated team responses to hostile fire.
Incoming shot announcements are transmitted to a built-in speaker or an earpiece while a lightweight display provides range and azimuth of the shooter position. As the soldier moves, the system compensates for the soldier’s motion and continually updates the threat’s location on a wrist display.
The Boomerang Warrior X system.
This summer, an undisclosed Gulf nation has awarded a direct commercial sales contract to Raytheon BBN Technologies valued at more than $10 million for the delivery of 2,000 Boomerang Warrior-X systems during the next 12 months.
“This technology is a proven life saver on the battlefield,” said Ed Campbell, president of Raytheon BBN Technologies. “Boomerang delivers the best performance of any available shooter detection system today at the lowest cost.”
Raytheon BBN Technologies is a wholly owned subsidiary of Raytheon Company.
I had my special ISO-certified glasses ready. Living in Oregon, I wasn’t about to miss the once-in-a-lifetime chance to see a total eclipse of the sun.
On Aug. 21, my family drove a few miles north to get into the path of totality, which for us lasted about a minute. It was definitely worth the field trip.
Besides regular folk like me, experts in numerous fields turned their eyes — and their instruments — to the eclipse.
The National Center for Atmospheric Research took to the air with a Gulfstream V fitted out with sensors and equipment for atmospheric research. The flight gathered data about the sun that can’t be collected from the ground.
With better instruments than ever before, for the first time scientists had the chance to observe the corona in the infrared spectrum, which may provide insight into the sun’s magnetic fields.
Back on terra firma, atmospheric scientists closely monitored changes in temperature and other weather effects. The temperature dropped as much as 7 degrees in Crossville, Tennessee, reports the National Weather Service.
Scientists at zoos and aquariums across the country closely watched animal behavior during totality. Species exhibiting unusual behavior included elephants, hippos, crocodiles and penguins.
The scientists found a “decrease in the number of free electrons in the part of the Earth’s ionosphere along the eclipse path where sunlight was temporarily blocked by the moon…
“TEC [total electron content] time series from two continuously operating GPS monitoring stations near the path of totality…show a small dip of about 2 TECU [TEC units] or so around 18:00 UTC on Aug. 21, coincident with the timing of the eclipse.”
The eclipse also affected WAAS real-time correction data from geostationary satellites.
While study of the data continues, it’s clear that GPS easily withstood the eclipse. Learn more here.
.
.
The Institute of Navigation’s 
Surrounding sounds may not be a common way of determining location. But on the battlefield, warfighters need to know the direction of gunshots to enable a proper response.
