GPS/INS integrated navigation method based on self-learning cubature Kalman filter

Abstract

The invention discloses a GPS/INS integrated navigation method based on self-learning cubature Kalman filter. Running stages of a GPS/INS integrated navigation system are divided into a training stagand an error compensation stage. At the training stage being a GPS signal effective stage, a self-learning Kalman filter device formed by two cyclic filtering subsystems carries out optimal estimationon a speed error and a position error of an INS by using the speed difference and position difference of the INS and the GPS as observed quantities, thereby realizing the self-learning function. At the error compensation stage being a GPS signal out-of-lock stage, the Kalman filter device has the function of predicting the observation quantities through self-learning, the prediction result of theLSTM network can be trusted fully, seamless navigation under the out-of-lock condition of the GPS signal is realized, and the optimal estimation error value of the Kalman filter device is compensated, thereby improving the positioning accuracy of intelligent vehicle navigation in complex environments. The GPS/INS integrated navigation method can be applied occasions like vehicle navigation positioning in a complex urban environment; and the autonomy of navigation positioning is improved effectively and thus the intelligence degree of the vehicle is increased.

Description

technical field [0001] The invention relates to the technical field of navigation positioning and deep learning, in particular to a GPS / INS combination method based on self-learning volumetric Kalman filtering. Background technique [0002] The inertial navigation system (Inertial Navigation System, INS) has the advantages of full autonomy, complete navigation parameters, and high short-term accuracy, but its errors will accumulate over time, and the long-term flight time will lead to a serious decrease in accuracy. Global Positioning System (Global Positioning System, GPS) can provide full-time, continuous real-time information, but the dynamic performance of GPS signals is poor, and it is easily interfered by external information. In areas without GPS signals, such as indoors, urban areas with dense high-rise buildings, Areas with poor observation environment such as underground tunnels or inside buildings will greatly reduce the accuracy of navigation and positioning, whi...

AI Technical Summary

Problems solved by technology

[0003] Purpose of the invention: The present invention proposes a GPS / INS integrated navigation method based on self-learning volumetric Kalman filter, which overcomes the problem of decreased accuracy of integrated navigation in the GPS out-of-lock environment, and the method can be used in long-term voyage out-of-lock environments. Provides persistent high-precision navigation

Images