Design of improved adaptive Kalman filter based on data fusion

Abstract

This paper designs a new adaptive Kalman filter (KF) to solve the filtering problem of with unknown noise statistics. The estimation error of the process noise covariance matrix will destroy the state estimation, and the state estimation error will reduce the estimation of the process noise covariance matrix, thus affecting the stability and optimality of the filter. In addition, most adaptive KF algorithms adopt the measurement method of single sensor, which is easy to cause the shortage of measurement data and affect the accuracy of the algorithm. Therefore, we propose a measurement sequence fusion adaptive Kalman filter (MSFAKF) algorithm to solve this series of problems. Based on MSFAKF algorithm, the estimation of process noise covariance and measurement noise covariance are obtained by measurement fusion sequence and innovation sequence respectively. Under certain conditions, the stability of MSFAKF algorithm is proved, and the filtering result converges to that of ideal KF with known second moment of noise. Simulation experiment shows that MSFAKF algorithm has the characteristics of fast convergence speed, high accuracy and strong adaptability.