Fully Distributed Dynamic State Estimation With Uncertain Process Models

Abstract

This paper studies the problem of distributed state estimation in the scenario where the process model of interest is not certain to any local agent. By extending the previously proposed distributed hybrid information fusion (DHIF) algorithm to this considered scenario, two algorithms are proposed by following the frameworks of two well-known multiple-model (MM) approaches, namely, the first-order generalized pseudo-Bayesian and interacting MM approaches. The extended algorithms inherit the advantages of the original DHIF algorithm and are hence fully distributed, robust against agents not directly sensing the target, and only require a single communication iteration among agents during each sampling interval. It is also shown in the case when the unknown underlying model/mode is fixed; all local agents are able to asymptotically identify the true underlying model/mode and estimate the state of interest simultaneously. Sufficient conditions are formulated. Simulations are shown to illustrate the analytical results as well as the performances of the proposed algorithms.