Event-Triggered Resilient Average Consensus With Adversary Detection in the Presence of Byzantine Agents

This paper addresses the problem of resilient average consensus in the presence of Byzantine agents in multi-agent networks. An event-triggered secure acceptance and broadcasting algorithm is proposed in which full knowledge of the network and high computational capabilities of each regular node are not required. The computational expense and communication times are also reduced for the event-triggered mechanism. We analyze the conditions for such a fully distributed algorithm to succeed in the f-local adversarial model. A new definition called an f-propagation graph, which is extended from r-robustness, turns out to be more accurate in describing the required topology conditions. Based on the proposed algorithm and topology conditions, we provide another algorithm to detect the adversarial nodes according to their abnormal behavior. When the network topology is an f-propagation graph, regular nodes that are equipped with the proposed algorithms update state values synchronously and eventually converge asymptotically to resilient average consensus. Simulation results are provided to verify the effectiveness of our proposed algorithms and the network topology conditions.