Self-Tuning Fusion Filter for Multi-Sensor Systems with Unknown Packet Receiving Rates

Abstract

This paper studies a self-tuning estimation problem for a multi-sensor networked linear stochastic discrete-time system with unknown packet receiving rates and unknown model parameters. Packet dropouts may occur when measurements of a sensor are transmitted to a local processor. Phenomena of packet dropouts are described by a group of Bernoulli distributed random variables. By model transformation, the identification problem of packet receiving rates is transformed into that of unknown model parameters for a new augmented system. The Recursive Extended Least Squares (RELS) algorithm is used to simultaneously identify packet receiving rates and model parameters of original systems. Using an optimal matrix-weighted fusion estimation algorithm in linear unbiased minimum variance sense, a fusion identifier of model parameters for original systems is presented. Further, a self-tuning state fusion filter is obtained by substituting identified fusion model parameters and packet receiving rates into the corresponding optimal estimation algorithms. A simulation example verifies the effectiveness of the proposed algorithms.