Scholarlink

To solve the problem of motion planning of the legged robot due to the complexity of system structure and decision making, a transferable control framework for hexapod robot based on deep reinforcement learning is proposed. The framework constructs a simulation environment to obtain training data, conducts data-driven training on the network through deep reinforcement learning to obtain control strategies, and migrates to the physical robot to evaluate and verify the gait and control performance of the hexapod robot. The results show that the motion of the trained policy has good performance in reward value, speed and stability, which verifies the effectiveness of the proposed method.

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

Reinforcement Learning for Motion Control of Legged Robots

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Xiang Fu,
  • Jiake Wu,
  • Xingshang Wang,
  • Kewei Chen

摘要

To solve the problem of motion planning of the legged robot due to the complexity of system structure and decision making, a transferable control framework for hexapod robot based on deep reinforcement learning is proposed. The framework constructs a simulation environment to obtain training data, conducts data-driven training on the network through deep reinforcement learning to obtain control strategies, and migrates to the physical robot to evaluate and verify the gait and control performance of the hexapod robot. The results show that the motion of the trained policy has good performance in reward value, speed and stability, which verifies the effectiveness of the proposed method.