Distributed Fault Detection Using Relative Information in Linear Multi-Agent Networks★

This paper addresses the problem of fault detection in the context of a collection of agents performing a shared task and exchanging relative information over a communication network. A techniques in the literature is used to construct a meaningful observable system equivalent to the original unobservable system of systems. A solution involving Set-Valued Observers (SVOs) and a left-coprime factorization of the system is proposed to estimate the state in a distributed fashion and a proof of convergence of the estimates is given under mild assumptions. The performance of the detection algorithm against deterministic and stochastic faults; and when the system is subject to unmodeled disturbances is assessed through simulations.