Detection and mitigation of misbehaving effects in multiagent networks with limited control

We focus on multiagent networks composed of nodes exchanging local information over a time-varying, connected, and undirected graph. In the network, we consider driver, observer, misbehaving and floating nodes, where driver nodes receive a control input based on their own state information, observer nodes receive a control input using the state information of their neighbouring nodes, misbehaving nodes are subject to a disturbance and/or fault, floating nodes are subject to neither a control input nor a disturbance. We first propose a proportional-integral control to be executed by a subset of nodes (i.e. driver and observer nodes) to suppress the adverse effects of misbehaving nodes. Specifically, we characterise which nodes in the network behave as desired based on the selection of observer and misbehaving nodes. Following that we introduce a novel misbehaving detection mechanism for the first time in the literature, which continuously detects misbehaving nodes in real time by using observer nodes and assigning each observer node to a misbehaving node to suppress its undesired effect. Finally, we provide numerical examples to show the efficacy of the proposed approach given in this paper.