Linear Optimal Filter for Descriptor Systems With Time-Correlated Measurement Noise

This paper is concerned with the linear optimal filtering problem for descriptor systems with correlated process and measurement noises. The process noise is a one-step auto-correlated noise and the measurement noise described by a first order Gauss-Markov process is time-correlated. Using a canonical form of descriptor and measurement differencing, the linear optimal filter in the linear minimum variance (LMV) sense is designed based on an innovation analysis method. It is computed by the solutions of some auto- and cross-covariance matrices. The simulation example verifies the effectiveness of the proposed algorithm.