Intermittent event-triggered circle formation control for multi-agent systems based on reinforcement learning

Abstract

This paper investigates circle formation control for the first-order multi-agent systems with privacy protection using a reinforcement-learning intermittent event-triggered method. Firstly, when the communication network suffers risks of information disclosure and denial-of-service (DoS) attack, an adaptive state observer is proposed to estimate the state of the multi-agent system. Moreover, based on the state estimation, a novel hybrid circle formation algorithm including the optimal controller is proposed as well as it can minimize the cost function. In light of stochastic analysis and Lyapunov stability theory, sufficient conditions for circle formation are derived. Furthermore, owing to the eavesdropping threat between the observer and the controller, an improved Paillier homomorphic encryption algorithm is employed to ensure data privacy, which requires less computation compared to some existing methods. Finally, simulation examples prove the accuracy of the theoretical results.