A resilient distributed optimization algorithm based on consensus of multi-agent system against two attack scenarios

In this paper, a resilient distributed consensus algorithm is proposed based on multi-agent system to defend against the Byzantine attack. Two attack scenarios including node attack and edge attack are considered and the differences between them are described, in which the edge attack scenario is seldom considered before. The assumption for a network under g-local attacks is relaxed from (2g + 1)-robust directed graph to (g + 1)-robust one based on a novel filtering method. The convergence and optimality are guaranteed by theoretical analysis, which ensures all the agents without attack can make an agreement and converge to a certain range. The simulation results show that for the two attack scenarios, the proposed algorithm can converge to the optimal solution of the compensation problem and the original problem, respectively. The proposed algorithm is more effective and has faster convergence speed compared with the contrast one.& COPY; 2022 The Franklin Institute. Published by Elsevier Ltd. All rights reserved.