Scholarlink

<p>Quadruped robots are challenged during the traversal of soft and yielding terrain because of extensive non-rigidity in terrain substrata. Existing approaches to motion control often neglect specific ground deformation effects in addressing foot–terrain interaction. This study introduces a novel strategy for enhancing quadruped robot locomotion stability on deformable terrains. Our strategy relies on robot foot–ground contact perception and terrain classification. A self-designed contact dual-sensor module perceives robot foot contact with the ground and classifies deformable terrains into different categories based on stiffness and deformation. A deformation force model estimates the ground reaction forces (GRFs) variation during the feet stance phase. This model integrates into robot GRF optimization, enhancing terrain-adaptive locomotion. Experiments show that the robot dynamically adjusts GRFs, substantially improving stability and adaptability in deformable terrain. Compared with standard model predictive control, the proposed strategy reduces the root mean square errors in robot roll, pitch, and yaw angles by 58.03%, 48.90%, and 57.28%, respectively, in unstructured deformable terrains. These results demonstrate the effectiveness of the strategy in soft terrains, enhancing performance while ensuring robustness and stability.</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

Enhancing quadruped robot locomotion stability on deformable terrains through contact perception and terrain classification

  • 期刊论文

  • 发表日期：

    2026年6月30日

查看全文
  • Yangyang Han,
  • Huaizhi Zong,
  • Zhenyu Lu,
  • GuoPing Liu,
  • Xiaohui Yang,
  • Junhui Zhang

摘要

Quadruped robots are challenged during the traversal of soft and yielding terrain because of extensive non-rigidity in terrain substrata. Existing approaches to motion control often neglect specific ground deformation effects in addressing foot–terrain interaction. This study introduces a novel strategy for enhancing quadruped robot locomotion stability on deformable terrains. Our strategy relies on robot foot–ground contact perception and terrain classification. A self-designed contact dual-sensor module perceives robot foot contact with the ground and classifies deformable terrains into different categories based on stiffness and deformation. A deformation force model estimates the ground reaction forces (GRFs) variation during the feet stance phase. This model integrates into robot GRF optimization, enhancing terrain-adaptive locomotion. Experiments show that the robot dynamically adjusts GRFs, substantially improving stability and adaptability in deformable terrain. Compared with standard model predictive control, the proposed strategy reduces the root mean square errors in robot roll, pitch, and yaw angles by 58.03%, 48.90%, and 57.28%, respectively, in unstructured deformable terrains. These results demonstrate the effectiveness of the strategy in soft terrains, enhancing performance while ensuring robustness and stability.