Performance Analysis of Robust Tightly Coupled GNSS/INS Integration Positioning Based on M Estimation in Challenging Environments

Abstract

AbstractThe phenomena of limited satellite visibility and multipath is serious in challenging environments, which causes to decrease of the positioning accuracy dramatically. An effective solution is tightly coupled integration of Global Navigation Satellite Systems (GNSS) and Inertial Navigation System (INS) based on robust M estimation. However, the robust performance is directly influenced by both the selection of equivalent weight functions and the INS accuracy difference. In this paper, theoretical analysis and experiment validation of positioning accuracy and reliability for the robust integration system are implemented under the condition of various equivalent weight functions and different INS systems. Firstly, for constructing equivalent least square form of EKF, we combine classical EKF observation equation with redundant observation equation established by predicted state and corresponding covariance in propagation step. Then, robust extended Kalman filter (REKF) algorithm which is based on M estimation is introduced for purpose of weakening the influence of faults on location. Moreover, the features of three typical equivalent weight functions (Huber, Tukey, IGG-III) are analyzed respectively. Finally, a simulation experiment platform is built to analyze the influence of the choice of equivalent weight functions and the IMU grades on positioning accuracy and reliability, which is realized by integrating GNSS and various grades IMU to locate with three REKF solutions against traditional EKF. Simulation experimental results show significant enhancements for REKF based on M estimation in terms of precision and reliability of tightly coupled system. Furthermore, in the setting failure scenario, the solution scheme based on IGG-III equivalent weight function can obtain the optimal robust effect in comparison with REKF(Huber) and REKF(Tukey), whose horizontal positioning accuracy is kept within −0.5 m–0.5 m and vertical positioning accuracy is no more than 1 m. Meanwhile, compared with consumer grade IMU, using tactical grade IMU can further improve positioning accuracy and reliability.KeywordsExtended Kalman FilterEquivalent least squareM estimationEquivalent weight functionINS