A Robust Vehicle Positioning Method Based on Low-Cost Sensors Under Short-Term Failure of Global Positioning System 2021-01-7010

Combining the advantages of GPS and INS, GPS/INS has been widely used in vehicle localization. However, there are GPS failures in actual scenes, and even a short-term failure will cause the positioning accuracy of the low-cost MEMS-SINS to decline rapidly. Therefore, aiming at the problem of vehicle positioning under short-term GPS failure, the following three aspects of work are mainly carried out. Firstly, the unscented quaternion estimator is introduced, and the velocity-combined GPS/SINS based on this algorithm is realized. Secondly, for the positioning problem under short-term GPS failure, a USQUE-LSTM positioning method is proposed by introducing the long short-term memory neural network, which is the main contribution of this paper. Finally, considering the defects of the neural network, in order to improve the adaptability of the positioning strategy to high-speed obstacle avoidance and other scenes, the USQUE-VDM positioning method is designed by introducing the two-degree-of-freedom dynamic model of vehicle; then based on the federated Kalman filter, the USQUE-VDM and USQUE-LSTM are further integrated, and the USQUE-LSTM-VDM positioning method under GPS short-term failure is proposed. To verify the proposed positioning methods, real vehicle tests and CarSim + MATLAB/Simulink co-simulation are carried out. The results show that the proposed USQUE-LSTM method can effectively improve the vehicle positioning performance under short-term GPS failure, and the positioning root-mean-squared error is less than 5m within 60s; and the proposed USQUE-LSTM-VDM strategy has higher positioning accuracy and better scene adaptability, which provides an effective solution to the positioning problem under short-term GPS failure.