Distributed adaptive sliding mode observers for a network of dynamical systems

This paper addresses the problem of monitoring a network of dynamical systems, by reconstructing the states of each node robustly, and in a distributed manner. The dynamics of each node of the network is assumed to have a known linear part and an unknown, but bounded uncertainty/nonlinearity element restricted to certain channels. For this class of systems a distributed observer, based on sliding mode principles is proposed. An adaptation law is part of the observer scheme and is used to ensure global state estimation properties are obtained. The distributed observer is structured such that each node relies on locally available information together with information collected from `neighbouring´ nodes in the network. The gains of the observer are synthesised in a decentralised manner based on node level models. The proposed methodology is verified by means of a simulation example of a network of Chua oscillators.