Scholarlink

Throughout the navigation of unmanned surface vessel (USV), they frequently encounter interference, such as waves, resulting in deviations from their intended trajectories. To tackle this challenge, we present a disturbance-resilient approach grounded in Stackelberg game theory. Initially, a dynamic system is established to capture velocity error, and a non-cooperative game model encompassing control inputs for unmanned vessels and external disturbances is devised. This framework is designed to mitigate the load on the optimal controller of USV. Next, to address the challenges posed by unmodeled dynamics of unmanned vessel systems and external disturbances, this paper integrates the effective approximation capability of integral reinforcement learning evaluation-action optimal control architecture with neural networks. It directly deduces an approximate optimal solution for the Stackelberg game, effectively mitigating the instability in the decision-making process of unmanned vessel anti-saturation control and enhancing the robustness and adaptability of USV. Ultimately, simulation results demonstrate the effectiveness of the proposed interference-resistant control strategy.

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

Integral Reinforcement Learning Based Robust Anti-disturbance Control of Unmanned Surface Vehicle Based on Stackelberg Game

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Dong Zhao,
  • Wenjing Ren,
  • Yizhen Meng

摘要

Throughout the navigation of unmanned surface vessel (USV), they frequently encounter interference, such as waves, resulting in deviations from their intended trajectories. To tackle this challenge, we present a disturbance-resilient approach grounded in Stackelberg game theory. Initially, a dynamic system is established to capture velocity error, and a non-cooperative game model encompassing control inputs for unmanned vessels and external disturbances is devised. This framework is designed to mitigate the load on the optimal controller of USV. Next, to address the challenges posed by unmodeled dynamics of unmanned vessel systems and external disturbances, this paper integrates the effective approximation capability of integral reinforcement learning evaluation-action optimal control architecture with neural networks. It directly deduces an approximate optimal solution for the Stackelberg game, effectively mitigating the instability in the decision-making process of unmanned vessel anti-saturation control and enhancing the robustness and adaptability of USV. Ultimately, simulation results demonstrate the effectiveness of the proposed interference-resistant control strategy.