Yawei Zhai, Shizhuang Wang, Yuanwen Fu, Xingqun Zhan, Shanghai Jiao Tong University, China

View Abstract Sign in for premium content


Future autonomous Urban Air Mobility (UAM) transportation systems are expected to significantly mitigate the heavy traffic load on the ground. As the key enabler of these systems, the next generation Unmanned Aerial Vehicles (UAV) will provide transportation services for both cargo and passengers, and will operate in low altitude of urban area with high traffic density. To ensure operation safety, their corresponding navigation systems must continuously provide position solutions with high integrity. Therefore, this paper develops, analyzes and tests an efficient sequential integrity monitoring scheme against multiple sensor faults in an integrated navigation system, using three most popular sensors in UAV applications: multi-constellation Global Navigation Satellite Systems (GNSS), Inertial Measurement Unit (IMU) and Visual Odometry (VO). The baseline sensor integration architecture is established through a Kalman Filter (KF), and Solution Separation (SS) based test statistics are employed. Unlike many prior work whose focus is only on GNSS integrity, the main contribution of this work is providing a detailed, step-by-step derivation of the integrity risk upper bound that rigorously capture all the potential faults from each sensor. Integrity is a key metric in safety-critical applications: it measures the trust that can be placed in the correctness of the information supplied by the navigation system. Historically, the main drive for developing integrity monitoring techniques was to improve civil aviation safety, because GNSS measurement faults can potentially lead to major integrity threats to aircraft. To achieve this, Receiver Autonomous Integrity Monitoring (RAIM) has been known since mid-1990s, and it has been evolved into Advanced RAIM (ARAIM) now [1, 2]. RAIM/ARAIM is widely acknowledged as the most mature and efficient method to provide integrity monitoring capability at the user receiver. It is typically achieved by performing real-time consistent check and integrity risk evaluation in a ‘snapshot’ way. With the boom of autonomous systems over the past decade, ensuring navigation safety has become a solid demand in a various of applications. As a result, more researchers outside of aviation field have been attracted by the integrity concept. This is particularly the case for Highly Automated Vehicles (HAV) and future passenger-carrying UAMs, for which their navigation systems usually consist of multiple navigation sensors including GNSS, IMU, VO, LiDAR, etc. To maximize each sensor’s natural advantage and to improve the system’s robustness, the position solutions are obtained by integrating the raw measurements into a filter. Therefore, to accommodate the intended application of this work (i.e., UAM/UAV), the integrity monitoring scheme design must be filter based with multiple navigation sensors. Due to its prevalence, VO is selected in addition to GNSS/IMU. But the theoretical contributions developed in this paper is also applicable for other sensors. Comparing to ground vehicles, UAM will be subject to less GNSS signal blockage and high multipath. We will assume the users always have access to sufficient satellites from multiple GNSS constellations. There has been prior work on KF-based sequential integrity monitoring using GNSS/IMU coupling [3 ,4]. The principles are employing IMU as an unfaulted redundant measurement to detect GNSS faults or spoofing, where innovation-based detectors are widely selected [5]. However, employing a snapshot innovation will lead to poor detection capability against small ramp fault, and using an innovation sequence monitor will eventually lose bounding as the filter time increases. In addition, the innovation-based integrity risk evaluation requires computing the worst-case fault, which is built upon a series of complicated and computational expensive derivations [6]. Recently, there has been exploration on SS-based sequential integrity monitoring, but most of the existing work are heuristic that lacks comprehensive build-up [7, 8]. Also, those work only consider GNSS/IMU with limited satellite number while assuming IMU measurement is always fault free. So, they are not directly applicable for our scenario. In this paper, we firstly established an Extended KF architecture by tightly coupling GNSS/IMU, and by using the velocity output from VO as one of the measurements. Then, fault vectors are introduced in the original process model and measurement equation to reveal how they can impact the state update, the measurement update, and eventually the position error. The mean of the position error is clearly expressed in terms of prior-state fault, IMU fault and measurement fault (including GNSS and VO). Meanwhile, Multi-Hypothesis (MH) SS (MHSS) test statistics are derived, where their means capture the presence of faults from each sensor. Finally, an overall integrity risk bound is established following a similar process as snapshot ARAIM. Because SS-based sequential integrity monitoring requires running a bank of KFs, computational load has become the main concern for this type of approach. This issue becomes serious (a) when many more satellites are used, and (b) when exclusion function is enabled, because the algorithm has to continuously perform a second layer sub-filter so that the exclusion statistics are available. This could lead to a squared increase of the KF number as compared to detection-only scenario. In response, this paper leverages the fault grouping methodology to mitigate (a), and develop a novel exclusion scheme to resolve (b). Unlike conventional methods whose exclusion determination mechanism is based on second layer detection tests, the new approach exploits the relative magnitudes of the normalized detection test statistics to make the final exclusion decision. As a result, if the maximum statistic is greater than the second largest statistic for a certain amount, the satellite(s) corresponding to this fault mode will be chosen to be excluded. The safety is ensured by developing an overbounding integrity risk associated with the proposed scheme, which is pre-evaluated before the detection and exclusion steps. The new SS-based sequential integrity monitor approach is validated and tested through simulation and experimental flight test using our self-built UAV platform. Similar parameters as baseline ARAIM simulation condition are employed to describe the GNSS constellation, expect we use a higher mask angle to address the potential surroundings around future UAM. Raw observation data from two constellations (GPS and BDS), IMU and a stereo camera is collected over the flight and post processed thereafter. The performance analyses are carried out by comparing the protection levels of dual-constellation ARAIM, GNSS/IMU and GNSS/IMU/VO integration. Preliminary results suggest employing IMU and VO can significantly improve integrity monitoring capability of the system. Moreover, the proposed exclusion scheme is tested by manually injecting faults with varying magnitudes, and high efficiency and effectiveness can be simultaneously achieved. [1] Joerger, M., Chan, F.C., Pervan, B., “Solution Separation Versus Residual-Based RAIM,” NAVIGATION: Journal of the ION, 61(4), 2014, pp. 273-291. [2] Blanch, J., Walter, T., Enge, Per., et al., “Baseline Advanced RAIM User Algorithm and Possible Improvements,” IEEE Transactions on Aerospace and Electronic Systems, 51(1), 2015, pp. 713-732. [3] Tanil, C., Khanafseh, S., Pervan, B., “Detecting Global Navigation Satellite System Spoofing Using Inertial Sensing of Aircraft Disturbance,” Journal of Guidance, Control, and Dynamics, 40(8), 2017, pp. 2006-2016. [4] Lee, J., Kim, M., Lee, J., Pullen, S., “Integrity assurance of Kalman-filter based GNSS/IMU integrated systems against IMU faults for UAV applications,” Proceedings of the 31st International Technical Meeting of the Satellite Division of The Institute of Navigation (ION GNSS+ 2018), Miami, Florida, September 2018, pp. 2484-2500. [5] Joerger, M., Pervan, B., “Kalman Filter-Based Integrity Monitoring Against Sensor Faults,” Journal of Guidance, Control, and Dynamics, 36(2), 2013, pp. 349-361. [6] Tanil, C., Khanafseh, S., Joerger, M., Pervan, B., “Sequential Integrity Monitoring for Kalman Filter Innovations-based Detectors,” Proceedings of the 31st International Technical Meeting of the Satellite Division of The Institute of Navigation (ION GNSS+ 2018), Miami, Florida, September 2018, pp. 2440-2455. [7] Tanil, C., Khanafseh, S., Joerger, M., et al., “Optimal INS/GNSS Coupling for Autonomous Car Positioning Integrity,” Proceedings of the 32nd International Technical Meeting of the Satellite Division of The Institute of Navigation (ION GNSS+ 2019), Miami, Florida, September 2019, pp. 3123-3140. [8] Gunning, K., Blanch, J., Walter, T., et al., “Integrity for Tightly Coupled PPP and IMU,” Proceedings of the 32nd International Technical Meeting of the Satellite Division of The Institute of Navigation (ION GNSS+ 2019), Miami, Florida, September 2019, pp. 3066-3078.