Dynamic Event-Triggered Cluster Consensus of Multi-Agent Systems via PSO-GA Co-Design

摘要

This paper is concerned with cluster consensus of multi-agent systems (MASs) with homogeneous and heterogeneous dynamics. Firstly, a resilient dynamic event-triggered (RDET) mechanism is proposed by introducing two parameters in the dynamic triggering threshold to describe busy and free networks of each cluster. An auxiliary function is configurated to describe the network status of agents in different clusters. Agents allocated in multiple clusters are classified into two categories: some agents are in the busy network environment and the rest are in the free network environment. The resulting consensus error system is modeled as an augmented system involving two time-varying delays with different bounds. Then, some sufficient criteria are derived for the stability analysis of homogeneous and heterogeneous MASs, respectively. Based on the obtained stability criteria, an algorithm consisting of particle swarm optimization (PSO) and a genetic algorithm (GA) is designed for a co-design of control gains and event-triggered parameters. Finally, a numerical example of satellite formation flying is simulated to illustrate the merits of the developed theoretical results.Note to Practitioners—The objective of this article is to design addresses the challenge of achieving consensus in MASs with diverse dynamics, such as those found in autonomous vehicle fleets or sensor networks. We’ve introduced a RDET mechanism that can handle both busy and free network environments by adjusting its triggering threshold based on network status. This allows agents to communicate efficiently, reducing the need for constant communication and conserving resources. Agents in multiple clusters are categorized into those in high-traffic (busy) and low-traffic (free) networks. We’ve developed a model for the consensus error system that accounts for time-varying delays, which is crucial for stability analysis in both homogeneous and heterogeneous MASs. To optimize performance, we propose an algorithm combining PSO and GA. This powerful tool co-designs control gains and event-triggered parameters, ensuring efficient and stable communication strategies for all agents regardless of their specific dynamics. In practical terms, this work provides a robust framework to improve inter-agent cooperation in complex industrial scenarios, such as large-scale robot swarms, intelligent transportation systems, or even satellite constellation management. By implementing the RDET mechanism with our optimization algorithms, practitioners can expect to achieve tighter synchronization, reduced communication costs, and better overall system resilience, as demonstrated through a simulation of satellite formation flying.