Robust Consensus Tobit Kalman Filtering for Distributed State‐Saturated System Under Dynamic‐Disturbed Saturation Levels and Censored Measurements

ABSTRACT A robust consensus Tobit Kalman filtering (RCTKF) is derived in this paper for the distributed state‐saturated system suffering from dynamic‐disturbed saturation levels and censored measurements. A dynamic‐disturbance vector is selected to modify the classical state‐saturation model for the characterization of state saturations with dynamic‐disturbed saturation levels. Then, a distributed‐fusion cost function (DFCF) is founded on the maximum correntropy criterion (MCC) to enhance the accuracy of state estimates under non‐Gaussian system noise around the saturation and censoring regions through the fused adjusting factors. Next, the RCTKF is derived on the DFCF to obtain optimal global‐state estimates within limited consensus steps. A dynamic selection strategy is designed based on fused adjusting factors for the censoring probability and disturbance‐compensation probability to respectively enhance the one‐step prediction accuracy of state and measurement. In terms of the local state estimation, the upper bounds are deduced for the covariances of state one‐step prediction errors and filtering errors to obtain their analysis solutions, and then the filtering gains with two different expressions are derived on the DFCF to obtain the optimal local‐state estimates. The weighted average consensus‐based distributed fusion is considered for the double information pairs including state estimates and adjusting factors to obtain the optimal global‐state estimates within limited consensus steps. Finally, the numerical example and 3D‐target tracking example are chosen to verify the filtering performance of RCTKF.