Scholarlink

<p>This study addresses the leader-follower formation control problem for wheeled mobile robots by developing a trajectory tracking control strategy based on actor-critic reinforcement learning. The proposed method enables the follower robot to maintain the desired formation relative to the leader under varying conditions. Both the actor and critic components are implemented using diagonal recurrent neural networks (DRNNs), which process trajectory and formation tracking errors as inputs. The actor network outputs are employed to adaptively update the trajectory tracking control gains, while the critic network evaluates the performance of the actor. To ensure stability of the closed-loop system, the DRNN weight update laws are derived using Lyapunov stability theory. Comparative simulation and experimental results demonstrate that both trajectory and formation tracking errors converge asymptotically to zero. Furthermore, the proposed method achieves significant improvements over existing approaches, with a 65.42% reduction in mean absolute error and a 19.88% reduction in root mean square error compared to previously published high-performance controllers. These results indicate that the proposed leader-follower formation control strategy is not only effective but also exhibits superior robustness to environmental uncertainties, including external disturbances and obstacles, when compared with state-of-the-art methods.</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

Stable online adaptive formation control for leader-follower mobile robots based on reinforcement learning

  • 期刊论文

  • 发表日期：

    2026年4月13日

查看全文
  • Ahmad M. El-Nagar,
  • Raouf Fareh,
  • Sofiane Khadraoui,
  • Maamar Bettayeb,
  • Tamer Rabie,
  • Ibrahim Kamel,
  • A. Aziz Khater

摘要

This study addresses the leader-follower formation control problem for wheeled mobile robots by developing a trajectory tracking control strategy based on actor-critic reinforcement learning. The proposed method enables the follower robot to maintain the desired formation relative to the leader under varying conditions. Both the actor and critic components are implemented using diagonal recurrent neural networks (DRNNs), which process trajectory and formation tracking errors as inputs. The actor network outputs are employed to adaptively update the trajectory tracking control gains, while the critic network evaluates the performance of the actor. To ensure stability of the closed-loop system, the DRNN weight update laws are derived using Lyapunov stability theory. Comparative simulation and experimental results demonstrate that both trajectory and formation tracking errors converge asymptotically to zero. Furthermore, the proposed method achieves significant improvements over existing approaches, with a 65.42% reduction in mean absolute error and a 19.88% reduction in root mean square error compared to previously published high-performance controllers. These results indicate that the proposed leader-follower formation control strategy is not only effective but also exhibits superior robustness to environmental uncertainties, including external disturbances and obstacles, when compared with state-of-the-art methods.