Improvement Of GPS Vector Tracking Algorithm Based On Cubature Kalman Particle Filter And Recurrent Neural Network

The Global Navigation Satellite System,with its advantages such as highprecision,all-weather,and easy-operated,has been widely applied to the long-term stability analysis and real-time monitoring of geohazard.However,the quality of satellite signals will be degraded due to blockage,multipath,and electromagnetic interference.As a result,the situation of tracking loss will often occur in traditional receiver based on scalar tracking algorithm,and the precision of positioning cannot be guaranteed.In this thesis,we propose to improve the GPS vector tracking algorithm and verify it through the data generated by the signal simulator.The main findings of this research are as follows:(1)A software and hardware experimental platform for generating,collecting and processing GPS signals is built according to the algorithm flow of the vector tracking loop.Besides,the system test results show that the experimental system platform can realize the whole process of signal simulation,acquisition,tracking and positioning.(2)The cubature Kalman particle filter is researched and used as the navigation filter of the vector tracking loop instead of the traditional Kalman filter to improve the stability and precision of positioning in harsh environments.Experiments show that the scalar receiver without integrity monitoring failed to position in weak signal conditions which the carrier to noise ratio of a satellite signal drops to 31 d B-Hz since the scalar tracking loop was unlocked.In this case,however the vector tracking loop based on the cubature Kalman particle filter still maintains a stable tracking and positioning.In high dynamic conditions,the proposed algorithm has a smaller position root mean square error in X and Y direction in WGS-84 coordinates,with a reduction of 3.6% and 14.6%,respectively,when compared with the vector tracking based on extended Kalman filter algorithm.(3)The innovative sequence values of the extended Kalman filter is predicted by a Long Short Term Memory-Recurrent Neural Network which is optimised by the sparrow search algorithm.The neural network is trained by the innovative values with normal signals.When the signal is interrupted,the trained neural network will be used to predict the innovative sequence values to improve the precision of vector tracking.Based on the results of the experiment,it is shown that the vector tracking aided by neural network reduced the position root mean square error in the X,Y,Z direction in WGS-84 coordinates by 65.4%,5.3%,and15.4% during the short-time or temporary signal blockage.