Distributed state estimation with compressed and synchronized auxiliary particle filters using graph theory

In this paper, we propose a novel compressed distributed auxiliary particle filter that uses graph theory (CDAPF-GT) to reduce the communication cost and improve the estimation accuracy of a nonlinear state space model. Our method compresses the global loglikelihood function into a set of parameters that are updated by an average consensus algorithm over a network of nodes. Unlike the existing methods, our method synchronizes the particle sets among all the nodes and uses the latest measurements to construct a better proposal distribution. Our method is suitable for applications that require fast and reliable distributed state estimation. We show through various simulation scenarios that our method outperforms the common counterpart method in terms of estimation accuracy with the same level of communication.