A global approach to fault detection in multi-agent systems with switching topologies subject to cyber-attacks

In this paper, a distributed methodology is proposed for attack detection and isolation in multi-agent systems (MAS) with high order integrator dynamics. The proposed methodology is capable of detecting an attack on a node/agent and on an edge/communication link anywhere in the fleet and identify its nature even in the presence of a switching communication topology. In order to effectively achieve this, a cascade of predefined-time switched observers is introduced for each agent to achieve a fast estimate of the global state of the MAS whereby the settling time is an a-priori user defined parameter, independently of the initial conditions. Then, through a conventional consensus algorithm, a set of residuals are generated by the agents that is capable of detecting and isolating attacks on nodes and edges in the network.