A Kalman gain modify algorithm based on BP neural network

摘要

In practical application of modified gain extended Kalman filter (MGEKF) algorithm, generally used erroneous measured values instead of the real values, so the modified results also contain errors. To solve this problem, this paper proposes an improved MGEKF based on back propagation neural network (BPNN), termed BPNN-MGEKF algorithm. At BPNN training time, it uses measured values as the input, and modified results by true values as the output. So the trained network includes correction of errors, even if input measurement values can be obtained more accurate results. This paper applies the BPNN-MGEKF to single moving station bearing-only position experiment, experimental results showed that: compare to other algorithms, BPNN-MGEKF has a faster convergence speed and higher accuracy.