Photo: eli_asenova/E+/Getty Images
Researchers presented hundreds of papers at the 2021 Institute of Navigation (ION) GNSS+ conference, which took place virtually and in person Sept. 20–24 in St. Louis, Missouri. The following five presentations focused on the challenges of urban navigation. The papers are available at www.ion.org/publications/browse.cfm.
Integrating Autonomous Air Vehicles
The emergence and development of advanced technologies and vehicle types has created a growing demand for the introduction of new forms of flight operations. These new and increasingly complex operational paradigms, such as Advanced and Urban Air Mobility (AAM/UAM) present regulatory authorities and the aviation community with several design and implementation challenges — particularly for highly autonomous vehicles.
An overarching and daunting task is finding methods to integrate these emerging operations without compromising safety or disrupting traditional airspace operations. Predictive risk mitigation is critical to meeting this challenge. The authors of this study focus on the development and testing of a prognostic service aimed at estimating the quality of GNSS performance for an autonomous aircraft in complex environments. Flight operations would be able to factor into pre-flight and in-flight route planning an estimate of GNSS quality, thereby predicting poor or unacceptable navigation system performance. The authors provide methodologies for producing quality estimates, and provide results for selected simulation and flight-test cases.
Citation. Dill, Evan, Gutierrez, Julian, Young, Steven, Moore, Andrew, Scholz, Arthur, Bates, Emily, Schmitt, Ken, Doughty, Jonathan, “A Predictive GNSS Performance Monitor for Autonomous Air Vehicles in Urban Environments,” Proceedings of the 34th International Technical Meeting of the Satellite Division of The Institute of Navigation (ION GNSS+ 2021), St. Louis, Missouri, September 2021, pp. 125–137. https://doi.org/10.33012/2021.18138
Processing Scheme for Integrity Monitoring
Integrity monitoring is of great importance for GNSS applications. Unlike classical approaches based on probabilistic assumptions, the alternative interval-based integrity approach depends on deterministic interval bounds as inputs. Different from a quadratic variance propagation, the interval approach has intrinsically a linear uncertainty propagation adequate to describe remaining systematic uncertainty.
To properly characterize all ranging error sources and determine the improved observation interval bounds, the authors propose a processing scheme. The team validated how the sensitivity analysis is a feasible way to determine uncertainty intervals for residual ionospheric errors and residual tropospheric errors, taking advantage of long-term statistics against reference data. Transforming the navigation problem into a convex optimization problem, the interval bounds are propagated from the range domain to the position domain. The authors implemented this strategy for multi-GNSS positioning in an experiment with static data from International GNSS Service (IGS) station Potsdam (POTS) and an experiment with kinematic data from a measurement campaign conducted in the urban area of Hannover, Germany, on Aug. 26, 2020.
Citation. Su, Jingyao, Schön, Steffen, “Improved Observation Interval Bounding for Multi-GNSS Integrity Monitoring in Urban Navigation,” Proceedings of the 34th International Technical Meeting of the Satellite Division of The Institute of Navigation (ION GNSS+ 2021), St. Louis, Missouri, September 2021, pp. 4141–4156. https://doi.org/10.33012/2021.18078
Removing Multipath Errors
In the urban environment, multipath and non-line-of-sight cause measurement errors and signal power loss. In urban canyons, while multi-GNSS provides the required number of satellites to obtain a position, the signals may be affected by gross multipath errors, leading to a potentially unsafe position. In this paper, the authors use machine-learning techniques to model multipath error distributions. The features assessed are commonly used parameters such as elevation, S/N and user speed.
The authors drove a sensor-equipped vehicle in Toulouse, France, collecting hours of experimental data for evaluation of their model’s validity. The multipath error component was extracted from data processed from a single-frequency GNSS receiver using measurement differential, clock bias estimation and other techniques. The quantile of multipath error was then modeled using a neural-network-based regression technique. Results using the proposed method are validated by an integrity assessment of the experimental data.
Citation. No, Heekwon, Milner, Carl, “Machine Learning Based Overbound Modeling of Multipath Error for Safety Critical Urban Environment,” Proceedings of the 34th International Technical Meeting of the Satellite Division of The Institute of Navigation (ION GNSS+ 2021), St. Louis, Missouri, September 2021, pp. 180–194. https://doi.org/10.33012/2021.17874
GNSS/INS/Radar Sensor Fusion
Autonomous driving has gathered much interest in recent years with significant research directed at solving the localization problem. To enable a fully autonomous platform, the navigation system must provide accurate solutions at high rates, be reliable, and be available in all types of environments. These requirements necessitate the use of multiple sensors while remaining cost-effective to enable widespread adoption.
To maintain accurate positioning in GNSS-challenged areas, perception sensors such as cameras, lidar or radar provide another source of absolute positioning information. This paper presents a multi-radar integrated version of AUTO, a real-time integrated navigation system that provides an accurate, reliable, high rate and continuous (always available) navigation solution for autonomous platforms by integrating INS, GNSS-RTK, odometer and multiple radars sensors with high-definition maps. AUTO uses a tight nonlinear integration scheme to fuse information from multiple imaging radars with the INS/GNSS/odometer solution. The HD maps may come from a map provider or be crowdsourced from radar data.
The results in this paper compare multi-radar configurations of one to five imaging radars for a vehicle and demonstrate the accurate solution achieved through the tightly integrated system. Key performance indices are presented for a multi-radar configuration of AUTO for vehicle and robot. The results show how radar data contributes significantly with other sensors to provide a high-rate, accurate, reliable and robust navigation solution in GNSS-degraded environments and adverse weather conditions.
Citation. Krupity, Dylan, Ali, Abdelrahman, Chan, Billy, Omr, Medhat, Salib, Abanob, Al-Hamad, Amr, Wang, Qingli, Georgy, Jacques, Goodall, Christopher, “AUTO: Multiple Imaging Radars Integration with INS/GNSS for Reliable and Accurate Positioning for Autonomous Vehicles and Robots,” Proceedings of the 34th International Technical Meeting of the Satellite Division of The Institute of Navigation (ION GNSS+ 2021), St. Louis, Missouri, September 2021, pp. 77–92. https://doi.org/10.33012/2021.17903
Feature Matching for Visual Nav
Typical feature matching on aerial imagery results in a majority of features being placed on trees and other seasonally variable features. The researchers tested the effectiveness of using semantic segmentation to create and force robust features onto desired areas of an image for the purpose of visual navigation. The process involves testing several segmentation algorithms to achieve state-of-the-art segmentation results and evaluating the effectiveness of feature matching on segmented imagery. The aim is to develop a near state-of-the-art semantic segmentation model for aerial imagery that can extract desired buildings from an image.
The research will then focus on feature-selection and feature-matching algorithms to compare the segmented aerial key features with a database of features from satellite imagery. So far, results show that feature selection algorithms such as SIFT fail to overcome the nuances among multisource aerial imagery. Improving the feature selection algorithm ideally will allow for an increased quantity and quality of matches, ultimately resulting in a camera pose estimation sufficient to be a reliable alternative to GPS.
Citation. Hussey, Tyler, Leishman, Robert C., Woodburn, David, “Towards More Robust Vision-based Map Matching Through Machine Learning & Improved Feature Matching,” Proceedings of the 34th International Technical Meeting of the Satellite Division of The Institute of Navigation (ION GNSS+ 2021), St. Louis, Missouri, September 2021, pp. 1647–1653. https://doi.org/10.33012/2021.17911