Scholarlink

<p>Energy efficiency remains a critical challenge in quadruped robot locomotion, particularly at low to moderate walking speeds where rigid foot–ground interactions lead to excessive energy dissipation. This study proposes a porous metastructure-based robot foot and the corresponding morphology-aware reinforcement learning (RL) framework to improve energy efficiency and stability of quadruped locomotion. To this end, triply periodic minimal surface (TPMS) metastructures were integrated into the robot’s foot to enable passive elastic energy storage and release, which was explicitly accounted for during policy learning. TPMS hemispherical foot modules were designed, additively manufactured, and experimentally characterized under compression. The diamond-type TPMS exhibited the most favorable balance between compliance, energy absorption, and reduced hysteresis-induced energy loss, with a relative density of 60% identified as optimal. These properties were incorporated into a TPMS dynamics emulation model for deep RL, along with an energy-phase reward that aligns actuator power with the variation in elastic energy at the foot–ground interface. The proposed framework was validated through locomotion experiments on a quadruped robot. Compared with a conventional solid hemispherical foot as a baseline, the TPMS-foot module combined with the compliant dynamics-aware controller reduced the battery power consumption by 1.4%–6.2% across walking speeds from 0.4 to 1.0&#xa0;m/s. These results demonstrate that TPMS metastructures, when properly modeled and exploited through learning-based control, can serve as effective energy-shaping components for energy-efficient quadruped locomotion.</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

Energy-efficient Quadruped Robot Locomotion Via TPMS-Based Foot Design and Deep Reinforcement Learning

  • 期刊论文

  • 发表日期：

    2026年5月5日

查看全文
  • Kang-Hui Lee,
  • Delgermurun Bartogtokh,
  • Mahdi Bodaghi,
  • Jung-Yup Kim,
  • Keun Park

摘要

Energy efficiency remains a critical challenge in quadruped robot locomotion, particularly at low to moderate walking speeds where rigid foot–ground interactions lead to excessive energy dissipation. This study proposes a porous metastructure-based robot foot and the corresponding morphology-aware reinforcement learning (RL) framework to improve energy efficiency and stability of quadruped locomotion. To this end, triply periodic minimal surface (TPMS) metastructures were integrated into the robot’s foot to enable passive elastic energy storage and release, which was explicitly accounted for during policy learning. TPMS hemispherical foot modules were designed, additively manufactured, and experimentally characterized under compression. The diamond-type TPMS exhibited the most favorable balance between compliance, energy absorption, and reduced hysteresis-induced energy loss, with a relative density of 60% identified as optimal. These properties were incorporated into a TPMS dynamics emulation model for deep RL, along with an energy-phase reward that aligns actuator power with the variation in elastic energy at the foot–ground interface. The proposed framework was validated through locomotion experiments on a quadruped robot. Compared with a conventional solid hemispherical foot as a baseline, the TPMS-foot module combined with the compliant dynamics-aware controller reduced the battery power consumption by 1.4%–6.2% across walking speeds from 0.4 to 1.0 m/s. These results demonstrate that TPMS metastructures, when properly modeled and exploited through learning-based control, can serve as effective energy-shaping components for energy-efficient quadruped locomotion.