Session B2b: Autonomous Navigation and Safety-Critical Applications

Robust Sequential Integrity Monitoring for Integer Ambiguity Resolution-enabled RTK/INS Tightly Coupled Positioning
Jiachang Jiang, Lin Zhao, Harbin Engineering University; Jiaxiang Li, Marine Design and Research Institute of China; Chun Jia, Xin Xu, Harbin Engineering University
Location: Beacon B

High accurate and high trustworthy positioning is essential for autonomous vehicles. It will play a significant role in the future transport system. Recently, advancements in computing and communication technologies have accelerated the development and commercialization of autonomous vehicles. With the increasingly stringent requirements of the safety-critical autonomous vehicle applications, robust capability is required for GNSS-based navigation, which is the ability to provide continuous and reliable high-precision position information, regardless of the environment. The integration of inertial navigation systems (INS) with real-time kinematic (RTK) has been widely applied in autonomous vehicle fields for their superior complementary characteristics to enhance positioning performance and ambiguity resolution. However, the positioning accuracy and continuity dramatically deteriorate under challenging conditions caused by dynamics or harsh signal environments, where the satellite tracking geometry is unfavorable or the level of noise and multipath are severe, i.e., in urban or dense urban environments. Reliable and rapid carrier phase ambiguity resolution (AR) is still challenging for INS/RTK integrated positioning under those harsh scenarios. Generally, an instantaneous or fix-and-hold ambiguity filter scheme is utilized to process ambiguity state for INS/RTK integrated positioning. The instantaneous ambiguity resolution scheme ignores the integer constant constraint from epoch to epoch, weakening the model strength. However, it renders the navigation filter entirely immune to cycle slips, which may be a coarse but practical for ambiguity resolution under urban navigation. On the contrary, the fix-and-hold methodology, greatly increases the integer model strength for subsequent epochs, allowing the filter to hold on to fixes by the tight epoch-to-epoch state-space constraint provided by the inertial sensor and the vehicle dynamics pseudo-measurements. However, it is perilous because false fixes eventually contaminate the filter state with incorrect but highly confident priors. Worse still, this causes subsequent GNSS measurement updates to accept a similarly incorrect fix with high probability, repeatedly conditioning the filter state on incorrect ambiguities. To mitigate this issue, we introduce a sequential monitoring scheme against ambiguity resolution within an Extend Kalman filter (EKF) that ensures a high level of safety for RTK/INS integrated navigation applications. After exploring the sequential false ambiguity fixes impact on filter state errors analytically, a cumulative set-based monitoring scheme using a varying sliding window length-based innovation vector is developed to mitigate decreasing sensitivity to faults over time, which can safely protect against sequential GNSS incorrect ambiguity resolution. To quantify the fault detection capability, a real-time protection level (PL) against incorrect-fix events is formulated and developed. Meanwhile, to isolate the impact of false ambiguity resolution, a parallel backup float-only filter is applied to provide a seed for the ambiguity states in urban environments. Additionally, we utilize the best integer equivariant estimation (BIE) to assure the accuracy of float ambiguity. It is optimal in the minimum mean squared error (MSE) sense to rapidly recover from (re-) initialization. Land vehicle road tests have been conducted with various grades of IMU/GNSS datasets collected in open-sky and urban environments. Tightly coupled positioning with instantaneous and fix-and-hold ambiguity resolution modes are compared in the simulation. Specifically, the proposed PL for the RTK/INS tightly coupled positioning can bound the position error against false ambiguity resolution. The results demonstrate the effectiveness of the proposed algorithm, in which the ambiguity resolution success rate and positioning accuracy of the proposed method can be improved in restricted satellite visibility environments. The sequential innovation-based ambiguity resolution monitoring method proposed in this paper exploits consecutive windows of innovation vector with no requirement for a bank of filter, which efficiently improves the reliability of RTK/INS tightly coupled positioning under urban environments.