Distributed Multiple Model Estimation

Abstract

A distributed estimation algorithm for discrete time systems with multiple models described by Markovian parameters is presented. Several approaches have been proposed for the multiple model problem, however, they assume a centralized processing architecture in which all the measurements are sent to and processed at a central node. In this paper, the problem of constructing the global conditional pdf of states and model parameters by fusing the local pdfs from multiple nodes is considered. The results can be applied to multi-target tracking with maneuvers in distributed sensor networks.