Fully Distributed Event-Triggered Formation Control With Collision-Free for Nonlinear Multiagent Systems Under Directed Graphs
提出一种最优参考轨迹生成器，结合强化学习和事件触发机制，解决非线性多智能体系统的动态编队控制问题，同时通过碰撞避免函数确保安全，减少通信和计算负担。
By proposing an optimal reference trajectory generator (RTG), the dynamic formation control problem for nonlinear multiagent systems (MASs) under an event-triggered mechanism (ETM) is addressed via reinforcement learning (RL). The introduction of the RTG enables the removal of several commonly used but stringent constraints while still ensuring that the desired control objectives are achieved. This broadens the applicability of the proposed control strategy. Leveraging actor-critic-identifier networks, each agent is able to optimally track the reference trajectory under the proposed ETM framework. The developed hybrid ETM reduces both interagent communication frequency and the computational burden on agents. An RL-based controller is designed using not only actor-critic-identifier networks but also a novel collision-avoidance function to ensure safe agent behavior. Using a Lyapunov-like function analysis, we prove the effectiveness of the proposed control strategy and the convergence of the actor-critic-identifier network weights. Finally, the effectiveness and superiority of the proposed approach are further validated through a numerical simulation case and a comparative study.