Dynamic Leader-Follower Consensus with Adversaries: A Multi-Hop Relay Approach

2025-11-24

AI Summary

This work addresses the resilient consensus problem for multi-agent systems with adversarial neighbors, where normal followers must accurately track a time-varying leader’s state under misinformation attacks. Method: We propose a distributed resilient consensus protocol based on multi-hop relay communication, integrated with a Mean-Subsequence-Reduced (MSR) filtering mechanism, and applicable to both first- and second-order agent dynamics. Contribution/Results: We derive necessary and sufficient graph-theoretic conditions guaranteeing bounded tracking error; these conditions are less restrictive than those in prior works and yield tighter error bounds. Moreover, we characterize the minimal graph structure required to achieve resilient consensus. Numerical experiments demonstrate that the proposed method achieves superior robustness against diverse adversarial attack patterns—including Byzantine, spoofing, and miscommunication attacks—and delivers improved tracking accuracy compared to existing approaches.

Technology Category

• Multiagent Systems: Adversarial Agents
• Machine Learning: Adversarial Learning & Robustness
• Intelligent Robots: Multi-Robot Systems

Application Category

• Security and Privacy: Tracking, profiling, and countermeasures against them
• User Modeling, Personalization and Recommendation: Attacks and countermeasures in recommendation systems
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking

Abstract

This paper examines resilient dynamic leader-follower consensus within multi-agent systems, where agents share first-order or second-order dynamics. The aim is to develop distributed protocols enabling nonfaulty/normal followers to accurately track a dynamic/time-varying reference value of the leader while they may receive misinformation from adversarial neighbors. Our methodologies employ the mean subsequence reduced algorithm with agents engaging with neighbors using multi-hop communication. We accordingly derive a necessary and sufficient graph condition for our algorithms to succeed; also, our tracking error bounds are smaller than that of the existing method. Furthermore, it is emphasized that even when agents do not use relays, our condition is tighter than the sufficient conditions in the literature. With multi-hop relays, we can further obtain more relaxed graph requirements. Finally, we present numerical examples to verify the effectiveness of our algorithms.

Problem

• Resilient consensus tracking under adversarial misinformation attacks
• Multi-hop relay communication for improved error bounds
• Distributed protocols for dynamic leader-follower systems

Innovation

• Multi-hop relay communication for resilient consensus
• Mean subsequence reduced algorithm for adversarial environments
• Relaxed graph requirements with improved tracking error bounds