Improved decentralized GNSS/SINS/odometer fusion system for land vehicle navigation applications

Due to low cost and complementary performance advantages, global navigation satellite system (GNSS)/ strapdown inertial navigation system (SINS) integrated systems have established themselves in certain areas of land vehicle navigation. However, this integrated system cannot maintain reliable positioning solutions in challenging environments due to the inherent fragility of GNSS signals and time accumulated errors of a stand-alone SINS. To address this challenge, a multi-source information fusion system based on the decentralized system architecture and sequential Kalman filter for a land vehicle is proposed, which can fuse information from an odometer and motion aided constraints selectively and adaptively in different driving environments. Moreover, a comprehensive calibration and compensation strategy is designed to enhance the information fusion. On the one hand, a real-time calibration algorithm is designed to estimate the time-varying odometer scale factor and the misalignment between the inertial measurement unit (IMU) and vehicle body frame when GNSS signals are available. On the other hand, the forward velocity error caused by the lever arm, and the non-zero lateral velocity generated by the turning maneuver are compensated by the introduced velocity compensation method. A real car experiment in urban areas is carried out to illustrate the effectiveness of the proposed system. It shows that the proposed decentralized GNSS/SINS/odometer fusion system can maintain an average horizontal positioning root mean square error (RMSE) of 1-meter level when GNSS signals are cut off about 1-2 min. In addition, compared with the traditional centralized fusion structure, the proposed decentralized fusion structure can mitigate the horizontal positioning RMSE of the whole trajectory from 2.95 m to 0.59 m, which verifies that it can obtain better performance for the application of low-cost sensors in complex GNSS environments.