Application Of Variational Bayesian-Based Adaptive Kalman Filter In Mobile In-Vehicle Integrated Navigation

With the rapid popularization of smartphones in China and the rapid growth of the number of motor vehicles,the demand for smartphone positioning in vehicle environments is increasing day by day.Due to the hardware cost and performance of smartphones,positioning methods based on the Global Navigation Satellite System(GNSS)are easily affected by the environment and cannot fully meet the continuous and accurate positioning requirements of smartphones in complex urban environments.By combining GNSS and Inertial Navigation System(INS),the advantages of the two systems can be complementary,effectively reducing the impact of complex urban environments on smartphone GNSS positioning.This thesis conducts research on the application of GNSS/INS integrated navigation algorithm in the smartphone car environment.Firstly,the limitations of Kalman filter in mobile integrated navigation systems with inaccurate initial measurement noise settings were explored.Then,the application effect of the variational Bayesian-based adaptive Kalman filter algorithm in the above situations was verified.Finally,the algorithm was improved to adapt to more complex observation environments.The specific research work and content are as follows:(1)Developed a mobile in-vehicle real-time GNSS/INS integrated navigation software.In order to address the issue of smartphone GNSS positioning accuracy being easily affected by the environment,and to verify the application effect of variational Bayesian-based adaptive Kalman filter in mobile integrated navigation systems,this thesis develops a mobile GNSS/INS real-time integrated navigation software based on pseudorange single point positioning,Doppler velocity measurement,INS mechanical arrangement,and GNSS/INS integrated navigation algorithm.The software realizes GNSS real-time positioning and speed measurement,GNSS/INS real-time integrated navigation,location information visualization,data storage and other functions.(2)Verified the accuracy and availability of mobile positioning system,mobile speed measurement system and the GNSS/INS integrated navigation system in the vehicle environment.The results of the vehicle dynamic experiment are as follows: the pseudorange single point positioning horizontal positioning accuracy of the smartphone in the urban in-vehicle environment is 4.801 m,and the vertical accuracy is 9.760 m;The horizontal accuracy of Doppler velocimetry is 0.264 m/s,and the vertical accuracy of velocity is 0.283 m/s;The horizontal position accuracy of the GNSS/INS integrated navigation is 3.130 m and the vertical position accuracy is 4.562 m;The horizontal accuracy of velocity is 0.286 m/s,and the accuracy of elevation direction velocity is0.155 m/s;The pitch angle accuracy is 0.576 °,and the heading angle accuracy is7.625 °.According to the experimental results,the positioning accuracy of GNSS/INS integrated navigation on mobile phones has been improved compared to GNSS positioning.However,due to inaccurate initial measurement noise settings,the speed accuracy of the integrated navigation has decreased compared to Doppler velocity measurement.(3)Verified the mobile phone based variational Bayesian adaptive integrated navigation system.Introduced the variational Bayesian-based adaptive Kalman filter algorithm,compared and analyzed the performance of variational Bayesian-based adaptive Kalman filter and Kalman filter through simulation data and vehicle measurement data.The data results indicate that when the initial measurement noise is not accurately set,the position,velocity,and attitude accuracy of the integrated navigation system based on Kalman filter will decrease;variational Bayesian-based adaptive Kalman filter can achieve better position,velocity,and attitude accuracy through online estimation of the covariance of measurement noise.Compared with Kalman filter,the variational Bayesian-based adaptive Kalman filter improves the horizontal accuracy of position by 4.0%,velocity by 33.74%,and heading angle by 5.8% when the initial measurement noise is set too high;When the initial measurement noise is set slightly,the horizontal accuracy of position is improved by 5.1%,the horizontal accuracy of velocity is improved by 12.1%,and the accuracy of the heading angle is improved by 9.1%.(4)Improved the variational Bayesian-based adaptive Kalman filter algorithm in complex mutation observation environments.To address the issue of measurement noise inconsistent with the true value caused by sudden changes in the observation environment,a mutation observation environment detection method was designed using smoothed average standardized innovation.By recalculating the forgetting factor,the efficiency of variational Bayesian-based adaptive Kalman filter in estimating the covariance of measurement noise was improved.The advantage of variable forgetting factor variational Bayesian-based adaptive Kalman filter based on innovation has been verified by simulating the observation environment of sudden changes based on measured data.The experiment shows that the improved variational Bayesian-based adaptive Kalman filter has similar horizontal position accuracy compared to the previous one,with an increase of 6.6% in horizontal velocity accuracy,6.01% in pitch angle accuracy,and 3.83% in heading angle accuracy.This thesis contains 39 figures,17 tables,and 96 references.