Distributed particle filter for nonlinear hybrid systems based on average consensus filter

This paper presents a new distributed monitoring approach for nonlinear, non-Gaussian hybrid systems incorporating multiple sensors in an embedded network configuration. The estimation engine of the proposed approach is particle filter (PF) which estimates locally the mode and continuous state of hybrid system at each sensor location or node. Decision on the mode of the system is established locally in a distributed network in which only nearby nodes exchange information. This objective is obtained by using an average consensus filter that allows the nodes of the system to reach agreement on values acquired by the nodes of the network. The effectiveness of the proposed method is demonstrated through simulation studies. © 2011 IFAC