Resilient Consensus of Higher-order Multi-agent Networks: An Attack-isolation-based Approach

Current resilient consensus algorithms deal with multi-agent networks with lower-order nodal dynamics where each agent excludes certain extreme values from its neighbors. This does not scale well in general networks. In this paper, a new type of resilient consensus algorithm based on distributed attack isolation (DAI-RC) is proposed for general higher-order networks. In the DAI-RC algorithm, the evolution of each normal agent can avoid influence from those neighbors which are isolated as victims of attack. To characterize a feasible communication topology to isolate attacked agents, a notion of graph isolability that captures attack isolability is proposed, based on which a sufficient condition to isolate the attacked agents is presented. Simulations are finally provided to illustrate the effectiveness of the proposed algorithm.