GNSS researchers presented hundreds of papers at the 2022 Institute of Navigation (ION) GNSS+ conference, which took place Sept. 19–23, 2022 in Denver, Colorado, and virtually. The following four papers focused on autonomous applications in transportation. The papers are available here.
Addressing integrity monitoring of autonomous navigation
There are critical issues for the integrity monitoring of autonomous navigation applications, which include an adequate uncertainty budget in the observation domain, redundancy for the determination of the navigational states, and the capability of fault detection and exclusion.
Several aspects are addressed in the paper, including how to: determine interval bounds to handle GNSS multipath effects in urban environments, realize fault detection and exclusion based on constraint satisfaction and set membership, and improve the detector using weighting models.
The authors of the paper aim to contribute to the alternative integrity approach based on interval and set representations for bounding and propagating system uncertainty. Simulated and real-world experiments are carried out to demonstrate the feasibility of the authors’ proposed methods.
The authors note that statistical evaluation of integrity will not always suffice due to the presence of remaining systematic uncertainty, but state the alternative integrity approach will contribute to future autonomous navigation applications.
Su, Jingyao; Schön, Steffen; “Advances in Deterministic Approaches for Bounding Uncertainty and Integrity Monitoring of Autonomous Navigation.”
Estimation and reference systems in automation
For a high level of automation, estimation is crucial, and to achieve a full and reliable navigation evaluation, a trustable reference system needs to be developed.
Although the presence of a reference system and of an inertial measurement unit with GNSS through the multi-sensor fusion scheme was integrated, in GNSS-denied or challenging environment the navigation solution could not be accurately estimated and still needs to be fixed.
The authors of the paper propose new strategies to better estimate the lidar-based position uncertainty and to update the reference system.
The first strategy proposed involves determining the appropriate position error covariance matrix, based on the Hessian matrix and the scale of covariance obtained from a normal distribution transform (NDT) scan matching technique and the geometric dilution of precision computed from the distribution of point cloud segments in each scan.
In the second strategy proposed in the paper, the updated reference system was post-processed according to the loosely coupled INS/GNSS/NDT integration scheme with a forward and backward smoothing process.
The results of the proposed strategies indicated that the updated reference system provides more reliable navigation estimation compared to an existing reference system from commercial software and can be used for accurate evaluation of positioning, navigation and timing with automated vehicle applications.
Srinara, Surachet; Chiu, Yu-Ting; “Adaptive Covariance Estimation of Lidar-Based Positioning Error for Multi-Sensor Fusion Scheme with Autonomous Vehicular Navigation System.”
Evaluating TerraStar-X
GNSS performance using typical, low-cost GNSS devices in vehicles is not enough to achieve the positioning and availability needed for lane-level accuracy on autonomous vehicles. The antenna and receiver hardware available in standard vehicles limits the position accuracy and convergence performance. These limitations make the positioning more susceptible to error sources such as receiver multipath, noise, carrier tracking and stability.
GNSS correction services with additional design considerations and sophisticated algorithms are needed to work within the constraints of automotive-grade GNSS devices to achieve the performance required for lane-level positioning.
TerraStar X technology from NovAtel enables these applications. It includes an orbit and clock determination system (OCDS), which produces a set of corrections, precise satellite orbits and clocks, and satellite-specific biases for individual signals augmented by the computation of additional regional corrections.
The authors of the paper outline the design and performance of the combined OCDS and regional correction system. They demonstrate the performance of the TerraStar X technology across a variety of applications.
The addition of regional corrections enables automotive and mass-market applications to achieve in-lane positioning in seconds, using any dual-frequency, dual-constellation GNSS hardware. The result is software that provides a continuous stream of multi-constellation, multi-frequency GNSS corrections — enabling a correction service that makes the affordable GNSS device ecosystem possible.
Regional corrections also improve the performance of survey-grade GNSS receivers.
Mervart, Leos; Lukes, Zdenek; Alves, Paul; “TerraStar X Technology: Design of GNSS Corrections for Instantaneous Lane-Level Accuracy on Large Scale Connected Vehicles and Devices.”
Solving the localization problem in autonomous driving
The localization problem in autonomous driving imposes two criteria on the navigation solution: accuracy and reliability or integrity. According to the authors of this paper, solving the localization problem is a key requirement to enabling the development of autonomous platforms.
This paper presents AUTO, a real-time integrated navigation system that tightly integrates INS, GNSS-RTK, odometer, and multiple radars sensors with high-definition maps to achieve a high-rate, accurate, continuous, and reliable navigation solution. It also shows how AUTO leverages a tight integration of imaging radars with other traditional sensors to provide a robust navigation solution with corresponding estimates of the uncertainty.
The AUTO solution was tested in a variety of environments and locations, including a range of conditions such as winter weather, to assure the robustness and reliability required by autonomous applications.
The results demonstrate the lane level accuracy of the solution in a variety of challenging urban and downtown environments. Additionally, the tight integration enables the determination of protection levels to describe upper bounds on the uncertainty.
The results in the paper are illustrated using a Stanford Diagram, along with a user-defined alert limit to describe the solution integrity and availability. The proposed algorithm uses a map matching technique between the imaging radar data and a globally referenced high-definition map to better estimate the solution uncertainty and protection levels.
AUTO’s tightly integrated approach to integrity monitoring means uncertainties and protection levels can be determined even in areas where the system may experience extended periods of GNSS unavailability.
Krupity, Dylan; Chan, Billy; Ali, Abdelrahman; Salib, Abanob; Georgy, Jacques; Goodall, Christopher; “Integrity Monitoring and Uncertainty Estimation with AUTO’s Non-linear Integration of Multiple Imaging Radars and INS/GNSS for Autonomous Vehicles and Robots.”
