LAUSR

Abstract Distributed consensus-based estimation is one of the main applications of sensor networks. Most approaches are highly dependent on exact model knowledge. This limitation motivated the development of robust distributed filters that deal with model uncertainties. Many of these works, however, are not fully distributed filters or demand high communication and computational efforts. In this paper, we propose a robust distributed consensus-based filter for uncertain discrete-time linear systems. We assume norm-bounded parametric uncertainties in all matrices of both the target system and sensing models. The approach consists of adopting a purely deterministic interpretation of the robust distributed estimation problem, formulated by combining the penalty function method and the robust regularized least-squares estimation problem. The filter is presented in a fully distributed Kalman-like structure that is suitable for online applications, requiring acceptable computational and communication efforts. We evaluate the effectiveness of the proposed filter by comparing its performance with an existing robust distributed filter, as well as with a centralized strategy.