Scholarlink

<p>To address the challenges of achieving consensus in dynamic and uncertain environments, this research presents an adaptive consensus control strategy for multi-agent systems (MASs) that utilizes reinforcement learning (RL). By integrating the actor-critic algorithm with a real-time feedback control mechanism, the method dynamically adjusts control inputs based on the agents’ states. Experimental results demonstrate rapid convergence and error reduction: in the star topology, state convergence is achieved, with the global error reducing to near zero in 2-3 seconds, while in the ring topology, convergence occurs within 1-2 seconds. Additionally, the method’s global asymptotic stability is rigorously proven using Lyapunov stability theory, confirming its robustness for complex MAS applications. A comparison with traditional control methods, including proportional-integral-derivative (PID) and Q-learning algorithms, highlights the superior performance of the proposed method in terms of convergence speed and error reduction, further validating its effectiveness in dynamic network environments.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Adaptive control for multi-agent systems in dynamic and uncertain environments

  • 期刊论文

  • 发表日期：

    2026年6月2日

查看全文
  • Xuqiong Luo,
  • Wenjing Ren,
  • Jing Li

摘要

To address the challenges of achieving consensus in dynamic and uncertain environments, this research presents an adaptive consensus control strategy for multi-agent systems (MASs) that utilizes reinforcement learning (RL). By integrating the actor-critic algorithm with a real-time feedback control mechanism, the method dynamically adjusts control inputs based on the agents’ states. Experimental results demonstrate rapid convergence and error reduction: in the star topology, state convergence is achieved, with the global error reducing to near zero in 2-3 seconds, while in the ring topology, convergence occurs within 1-2 seconds. Additionally, the method’s global asymptotic stability is rigorously proven using Lyapunov stability theory, confirming its robustness for complex MAS applications. A comparison with traditional control methods, including proportional-integral-derivative (PID) and Q-learning algorithms, highlights the superior performance of the proposed method in terms of convergence speed and error reduction, further validating its effectiveness in dynamic network environments.