A Multi Feature Fusion Aided Positioning for INS/Beidou with Combination of BP Neural Network and Differential Evolution Kalman Filter

Inertial navigation system (INS) is a pure autonomous navigation system, which can provide continuous real-time position, speed, and attitude information. The INS has the characteristics of short-term high accuracy and strong anti-interference ability, but the position error will accumulate with the extension of time. In order to further improve the accuracy of the traceless Kalman filter (KF) in Beidou/inertial conduction,.a Back Propagation (BP) neural network aided approach is developed in combination with improved differential evolution algorithm (DE). Specifically, we designed a new BP network that fuse multi features of signals. the INS and Beidou fusion data are collected as samples to train the BP neural network. Then, the improved DE algorithm is employed to optimize KF, attaining superior fusion efficiency. Simulation results demonstrate that the stability and accuracy of position have been significantly improved compared with the original combinatorial positioning model.