Online Motion Sensors Error Modelling for Robust Navigation Using Fast Orthogonal Search

Abstract

Global Navigation Satellite System (GNSS) and Dead Reckoning (DR) techniques are typically integrated to provide a robust and continuous navigation solution. However, frequent GNSS outages due to signal deterioration and blockage can severely impact the performance of the integrated navigation, which will be deprived of accurate GNSS updates and have to rely solely on the DR solution. The shortcomings of DR navigation solutions are due to the presence of several sensor errors such as biases, scale factor errors, thermal drifts, misalignment errors, and stochastic errors. Despite sensor calibration procedures, the impact of sensor errors may persist, propagating through the DR algorithm and leading to significant drifts, especially with Micro-Electro-Mechanical Systems (MEMS) sensors. In this paper, the objective is to improve the standalone navigation performance of Vehicle Sensors Dead Reckoning (VSDR) during GNSS out-ages. To be specific, the Fast Orthogonal Search (FOS) system identification technique is utilized to model Inertial Measurement Unit (IMU) sensor errors utilizing the availability of the accurate integrated navigation solution. The sensor error models are to be utilized when the integrated solution is compromised (i.e. GNSS outage) to estimate improved sensor measurements, thus reducing drifting navigation errors and achieving robust stan-dalone VSDR operations over extended durations. The proposed method is verified using real data from vehicle motion sensors on real road test experiments performed on a land vehicle in downtown Kingston, Ontario, Canada. Our results demonstrate significant improvements when utilizing the sensor error models for rectifying the raw sensor measurements achieving position accuracy enhancements of 56% on average across different outage durations.