Distributed recursive linear fusion predictor for multi-model multi-sensor systems with linear correlated noises

This paper studied the distributed recursive linear fusion prediction problem for multi-model multi-sensor systems with linear correlated noises. Firstly, local predictors and prediction error cross-covariance matrices for the common state components are proposed in the linear unbiased minimum variance (LUMV) sense via the innovative analysis approach. Then, a distributed recursive fusion predictor (DRLFP) is presented based on local predictors (LPs), cross-covariance matrices between local predictors, and cross-covariance matrices between the prior fusion predictor and local predictors. Finally, the stability and the steady-state property are analyzed. Its estimation accuracy is superior to that of LPs and distributed fusion predictor by only matrix-weighting local predictors (DFPMW), and is inferior to centralized fusion predictor (CFP). A simulation example shows the effectiveness of the proposed algorithms.