Robust Fault Estimation Using Relative Information in Linear Multi-Agent Networks

Abstract

Abstract-In this paper a robust fault estimation method, based on sliding mode observers, is proposed for a collection of agents undertaking a shared task and exchanging only relative information over a communication network. Since the 'system of systems' formed by the agents is not observable with respect to relative sensing information, by appropriate transformations and scalings of the inputs and outputs of the actual system, a meaningful observable subsystem is created. For this new subsystem, after modal decomposition based on the associated Laplacian, decoupled sliding mode observers, depending only on the individual node level dynamics of the network, can be created exploiting an existing design philosophy. These collectively form a centralized fault estimation scheme for the original system.

A Survey of Fault Diagnosis and Fault-Tolerant Techniques—Part I: Fault Diagnosis With Model-Based and Signal-Based Approaches

“…A general idea of the distributed fault diagnosis is to design local estimators or fault detection filters by intelligent agents according to the local sensing and computing resources, and a consensus strategy is utilized to ensure the whole detection or estimation performance of all the agents in the network. Recent developments on distributed fault detection [143] and distributed fault estimation [144] are developed respectively by using adaptive thresholds and sliding modes techniques to improve the robustness against noises and modelling errors. Moreover, applications on unmanned airships [145] and power networks [146] are reported as well.…”

Stochastic and deterministic fault detection for randomized gossip algorithms

“…In the context of fault detection in distributed systems, [8] addresses the problem by looking at the whole system and constructing a batch of observers for each subsystem. By looking at the outputs of these observers it is possible to detect and isolate faults affecting one of the sub-systems.…”

Distributed Fault Detection Using Relative Information in Linear Multi-Agent Networks ★ ★The work from Daniel Silvestre was supported by the project FCT [UID/EEA/50009/2013] and with grant SFRH//BD/71206/2010, from Fundação para a Ciência e a Tecnologia. J. Hespanha was supported by the U.S. Army Research Laboratory and the U.S. Army Research Office under grants No. W911NF-09-1-0553 and W911NF-09-D-0001. C. Silvestre was supported by project MYRG117(Y1-L3)-FST12-MKM of the University of Macau.

“…The importance of this problem is reported in [1] and later in [2], where the detection is crucial as a single malfunctioning node can severely impact the overall network performance. Applications span the areas of mobile robots, cooperating unmanned vehicles tasks such as surveillance and reconnaissance, distributed state estimation, among others (see [3] and the references therein).…”

“…In [3], one of the main results is showing that the overall system of the dynamical system is unobservable when only considering relative information of the states of the individual systems. A transformation is introduced that allows to perform fault detection and isolation by considering the observable subspace of the overall system.…”