Research On GNSS/SINS Integrated Positioning Error Suppression Technology Based On Deep Learning

This article takes cadastral survey as the application background,and relies on the topic of "Research on the Technology of Rapid Altitude Measurement and Accurate Positioning by Geographic Information Satellites for Villages and Towns Spatial Planning".It aims at the special measurement phase when the filtered Global Navigation Satellite System(GNSS)signal is abnormal or even fails during the measurement process.The low dynamic characteristics of the cadastral survey operation,the "stop and follow the measurement" and the measurement characteristics of the data can be processed after the fact,and the error suppression technology of the combined positioning of GNSS and strapdown inertial navigation system(SINS)has been studied.The main research work of this paper includes the following aspects:(1)Research on GNSS/SINS data fusion algorithm for cadastral survey.Aiming at the problem that the GNSS/SINS combined positioning system using conventional Kalman filtering has insufficient prior knowledge or inaccurate GNSS signal observations,resulting in a decrease in system positioning accuracy,an improved adaptive Kalman filtering algorithm(IAKF)is proposed.Observed outliers are detected by chi-square test,and improved Sage-Husa adaptive Kalman filter algorithm(ISHAKF)is used when no abnormalities are detected,and the noise parameters of the system are estimated in real time through innovation to solve the problem caused by insufficient prior knowledge The problem of increased filtering error;when anomalies are observed,the prior state mean square error matrix is adaptively adjusted according to the innovation to solve the problem that the innovation covariance estimation seriously deviates from the actual situation after the GNSS observation data abruptly,thereby improving the positioning accuracy of the system.Finally,a comparative simulation experiment was made between IAKF and conventional Kalman filtering.The simulation results show that the algorithm has good adaptability and stability,and can significantly improve navigation accuracy.(2)Research on error suppression technology for cadastral survey GNSS in the failure phase.In view of the observation failure area caused by interference or occlusion of the GNSS signal,the GNSS/SINS combined positioning system enters the pure SINS solution mode,and the positioning error gradually accumulates and cannot meet the positioning accuracy requirements.A long short-term memory(LSTM)neural network is proposed.Networkassisted combined positioning error suppression algorithm.According to the characteristics of LSTM neural network that can be effectively applied to long-distance time series,in the effective area of GNSS,the Kalman filter algorithm is used to perform data fusion on GNSS/SINS to obtain precise positioning information.At the same time,it uses inertial measurement unit(IMU),GNSS and SINS to output The information trains the neural network;in the GNSS signal failure area,the trained neural network is used to predict the GNSS position information,so that the system can continue to use the Kalman filter to filter.Through simulation experiments,it is proved that the algorithm can effectively suppress the system error divergence and improve the positioning accuracy when the GNSS signal fails,and it can still meet the positioning accuracy requirements and strong robustness in different motion states.(3)Experiment to verify the error suppression technology of the GNSS/SINS combined positioning system.Experiments are used to verify the improved adaptive Kalman filter algorithm proposed in this paper and the neural network-assisted combined positioning error suppression algorithm in the GNSS failure phase.The experimental results show that the combined GNSS/SINS positioning error suppression proposed in this paper can effectively reduce Accumulation of errors improves the accuracy and stability of cadastral surveying.