Scholarlink

This study proposes the design and testing of a quadruped robot with crawling posture, namely Ambot-N1. The mechatronics design is firstly introduced. A reinforcement learning motion control framework is proposed and applied to the robot. The influence of the initial joint angle on the convergence characteristics of the policy is tested. The optimal control policy trained using optimized joint initial angles demonstrates excellent generalization ability, enabling the robot to achieve adaptive motion in three terrains: grassland, gravel and solid ground.

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

Design and Test of a Quadruped Robot Using Crawling Posture and Reinforcement Learning Algorithm for Motion Control

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Xuan Wu,
  • Hongmin Zhou,
  • Yifan Zeng,
  • Jianqiang Yu

摘要

This study proposes the design and testing of a quadruped robot with crawling posture, namely Ambot-N1. The mechatronics design is firstly introduced. A reinforcement learning motion control framework is proposed and applied to the robot. The influence of the initial joint angle on the convergence characteristics of the policy is tested. The optimal control policy trained using optimized joint initial angles demonstrates excellent generalization ability, enabling the robot to achieve adaptive motion in three terrains: grassland, gravel and solid ground.