Research on Navigation and Positioning of Mobile Robot in Non-stationary Environment Based on Process Neural Network

China's independently developed Beidou 3 system has been fully operational and has achieved global positioning. In order to further improve the satellite navigation and positioning function of the ground mobile robot terminal, the influence of the high-frequency oscillating random disturbance signal received by the mobile robot data and the high-order nonlinear dynamics of the system on the navigation and positioning accuracy was analyzed, and the time-varying characteristics of the dynamic adaptive RTK-GPS positioning algorithm were used. A process neural network (PNN) based on empirical pattern decomposition (EMD) is proposed. Firstly, the existing input signal of the satellite positioning terminal is decomposed into several intrinsic mode functions (IMFs) using the EMD method. Then, for each IMF, the neural network model is constructed, and the dynamic error data is used as the sample for the neural network model correction training. For the satellite signal interference or lock loss process, the trained neural network is used to predict the output divergence to suppress the position and speed errors, so as to improve the accuracy of positioning and navigation. Experimental results show that this method is still suitable to improve the positioning accuracy in non-stationary environment, enhances the acquisition and tracking characteristics of the system, especially when the observation satellite is maneuvering, and the error of positioning results can be significantly reduced.