A Novel Robust Kalman Filter for Unmanned Ground Vehicles Positioning under GNSS Abnormal Measurements

Abstract

For unmanned ground vehicles (UGV), reliable and precise navigation solution is a main challenge in complex environment, especially when measurements of global navigation satellite system (GNSS) are abnormal. In order to address this challenge, we propose an algorithmic solution strategy and present a novel robust Kalman filter for UGV positioning via fusing information from GNSS and inertial navigation system (INS). Firstly, we review the positioning requirements of UGVs by analyzing the technical needs of continuously determining a vehicle’s location on road and precise navigation of lane level. Secondly, a new robust algorithm of Kalman filter is designed to reduce the positioning errors of GNSS/INS integrated navigation system when GNSS signals are abnormal. Thirdly, the application of the proposed algorithm to UGV positioning is illustrated. Simulation results with real data sets gathered from road tests show that the new robust filter can help us to evaluate the information quality of measurement, and can further autonomously adjust the Kalman gain and error covariance estimation matrices online. As a result, the accuracy and robustness of integrated navigation with the new filter can be improved in GNSS-challenged environments.