Distributed implementation and design for state estimation

This paper presents an architecture for agent-based distributed state estimation of linear plants with distributed outputs. The estimation structure is based on an orthogonal decomposition of the local observable/unobservable subspaces associated to each set of locally accessible outputs. The design of the observers can be carried out in a distributed way, which might open the door to scalable designs when the number of agent grows. The proposed architecture is developed for a two-agents network, where we establish stability results for the error dynamics, but comments are given about the generalization to larger networks. Simulations are provided to illustrate the estimation scheme in such broader cases.