Pseudo-measurement Optimal Information Fusion Kalman Filter Based on Least Squares Method

For the time-variant multisensor systems with correlated measurement noises and different measurement matrices, on the basis of recursive least squares (RLS) method, least squares(LS) method and Kalman filtering theory, two information fusion Kalman filters are put forward. The theory is that firstly Cholesky factorization is used to convert the former multisensor systems into noise irrelative and equivalent multisensor pseudo-measurement, and then in the next measurement cycle, pseudo-measurement model is considered to be AR model, and LS and RLS methods are respectively used to estimate the state, their global optimalities are proved, and that weighted measurement fusion is a special example based on LS method is proved. The proposed RLS algorithm strength is to transform inverse operation of gain matrix into scalar division, so operation burden is reduced and is convenient for real-time use. Simulation result confirms the effectiveness of the algorithm.