A Complex Networks Approach for Reliability Evaluation of Swarm Systems under Malicious Attacks

Abstract

The reliability of swarm systems needs to be investigated because of their self-adaption and self-organized features. Previous studies mainly focused on the reliability of a single agent, whereas for swarm systems, the ability to maintain their overall connection and function under adverse conditions is more important. It remains challenging on how to evaluate the reliability of swarm systems at the system level. In this paper, we present a reliability evaluation method for swarm systems by characterizing the behavior of the whole system using the method of temporal network analysis. A novel comprehensive reliability metric, i.e. cooperation reliability (CR), is proposed, considering both system integrity and motion consensus. Meanwhile, by identifying critical individuals in swarm systems, we design different malicious attack strategies. It is found that the malicious attacks perform much more harmfully to the reliability of swarm systems than noise and random attack. Moreover, we find that the reliability of a swarm system is sensitive to the swarm density in our framework due to the dynamical interaction of the system. Our findings may shed light on understanding the complicated behaviors of swarm systems under attack and designing a more robust swarm system.