Resilient Prescribed-Time Fully Distributed Consensus of Multiagent Systems Under DoS Attacks via Adaptive Time Regulation

Abstract

This article presents a prescribed-time (PT) fully distributed control scheme for multiagent systems to address the leader-following consensus challenges under denial-of-service (DoS) attacks. Initially, two novel practical PT stability criteria incorporating attack characteristics are developed. These criteria accommodate controlled Lyapunov evolution under nonvanishing uncertainties and facilitate dynamic adjustment of the actual convergence time to cope with unpredictable attack conditions. Following this, a fully distributed PT observer is formulated for each follower to reconstruct the leader’s unknown state based solely on the leader’s output and neighboring estimated outputs, without requiring global topology information. Furthermore, we propose a PT neural backstepping controller to address the “explosion of complexity” issue. Notably, by recursively aggregating unknown nonlinearities at each layer into the terminal control step for unified neural compensation, this design avoids the explicit computation of successive derivatives of virtual control laws, as well as the layer-by-layer deployment of neural networks and auxiliary filters. Comparative experiments on a robot formation system corroborate the theoretical superiority and practical applicability of the proposed scheme. Note to Practitioners—The motivation for this work arises from the leader-following consensus problem in multiagent systems (MASs) under denial-of-service (DoS) attacks, with particular emphasis on convergence performance and implementation efficiency. Existing prescribed-time (PT) consensus schemes are frequently characterized by stringent settling time assumptions, global topology reliance, and substantial computational overhead under DoS attacks. In response, we develop a novel PT fully distributed tracking framework relying solely on local outputs to address limited information availability and system uncertainties. The proposed strategy is particularly suitable for uncertain MASs subject to DoS attacks, such as spacecraft formations, intelligent robotic systems, and autonomous vehicles. Its key merits lie in the ability to adaptively regulate the actual convergence time according to attack conditions, while preserving a lightweight computational structure. Future research will extend these results to more complex adversarial settings, including communication delays, random packet losses, and actuator constraints.