Neuro-Adaptive Cooperative Tracking Control for Nonlinear Multiagent Systems Against DoS Attacks

This paper studies the issue of adaptive cooperative tracking control for MASs in the presence of DoS attacks. The agents' communication structure is characterized by directed links, ensuring that at least one leader maintains a directed route to every follower despite the presence of attacks. Initially, a distributed observer is designed to enable each follower to precisely estimate the leader's information. Furthermore, the study utilizes the average dwell time method in conjunction with adaptive backstepping strategies to create a secure consensus controller, ensuring that the tracking error has asymptotically approached zero. Additionally, the detrimental effects of DoS attacks have been fully addressed and neutralized. Ultimately, simulation outcomes have validated the effectiveness of the proposed approach for achieving adaptive asymptotic tracking control.

I. Introduction

Cooperative control of multi-agent systems (MASs) has extensive applications across diverse research areas, including power systems, transportation control, and robot formation control, among others [1]–[3]. Consensus is a basic problem of cooperative control, which can be roughly classified into leaderless cooperative tracking control and consensus with a leader. Notably, MASs heavily rely on sensors and communication networks, where each agent operates based on its unique control algorithm and interacts with adjacent agents through a network, fostering coordination towards a collective goal. However, a small cyber-attack on the communication network can degrade agents' control performance and even destroy the whole MASs [4]. Therefore, ensuring optimal consensus performance in nonlinear MASs subject to attacks is of paramount importance.