A New High-Precision Inertial Navigation Path Integration Algorithm Inspired by Grid Cells

In this paper, a path integration algorithm based on grid cell bio-intelligence is proposed to address the problem of navigation and positioning accuracy degradation due to the accumulation of inertial measurement unit errors over time. Based on the recurrent neural network, the bionic network is constructed by adding geometric constraints, and the path integration based on the bionic network is realized; the long- short-term memory(LSTM) network is added at the back-end of the model, which further improves the compensation effect. In this paper, two path integration methods are designed, namely the holistic method and the segmented method. When velocity is the input, the holistic path integral reduces the error by 77.99%, and the error can be reduced to 0.0062 m after further compensation by the LSTM network; the segmented path integral reduces the error to 0.10 m. The bionic network has a significant effect on the compensation of attitude angle, and the error is reduced by 63.81%. Using the compensated attitude angle, the speed is solved by the bionic network with acceleration as input, and the error is reduced by 99.07%; then the speed is input into the bionic model, and the holistic path integration reduces the error by 85.79%, and the error can be reduced to 0.0036 m after further compensation by the LSTM network; and segmented path integration reduces the error to 0.031 m. This path integration algorithm achieves the expected effect, and provides a good solution for the inertial navigation error. This path integration algorithm achieves the expected effect and provides a new idea for inertial navigation error compensation.