Two-level centralized fusion robust steady-state Kalman predictor over clustering sensor networks

This paper deals with the problem of designing the two-level centralized fusion robust Kalman predictors over clustering sensor networks with uncertain noise variances. The sensor network is partitioned as clusters by the nearest neighbor rule and every cluster consists of the sensor nodes and cluster head. According to the minimax robust estimation principle, based on the worst-case conservative system with conservative upper bounds of noise variances, applying the weighted least-squares optimal fusion rule, the two-level centralized fusion robust steady-state Kalman predictor is presented, which can significantly reduce the communication burden and save energy when the number of sensors is very large. Its robustness is proved by the Lyapunov equation approach. The robust accuracy relation of the two-level robust steady-state Kalman predictors is proved. A Monte-Carlo simulation example of a target tacking system shows the robustness and accuracy relations.