📡 Embodied AI Radar具身智能雷达

Topics主题 › Locomotion & Humanoids运动控制与人形机器人

Paper论文 2022 · MIT CoRL 2022

Walk These Ways: Tuning Robot Control for Generalization with Multiplicity of Behavior

Trains one quadruped policy that exposes gait parameters (frequency, foot swing, posture) as commands, so a human can tune behavior at deploy time for new tasks; open sim-to-real stack for the Unitree Go1.训练一个把步态参数(频率、抬脚高度、姿态等)作为指令开放的四足策略,部署时可由人调节行为以适应新任务;附带面向 Unitree Go1 的开源仿真到真实工具链。

BibTeX
@misc{margolis2022walk,
  title = {{Walk These Ways: Tuning Robot Control for Generalization with Multiplicity of Behavior}},
  author = {Gabriel B Margolis and Pulkit Agrawal},
  year = {2022},
  eprint = {2212.03238},
  archivePrefix = {arXiv},
  primaryClass = {cs.RO},
  url = {https://arxiv.org/abs/2212.03238},
}
Locomotion & Humanoids运动控制与人形机器人Quadruped / legged四足 / 足式realquadrupedsim2realrlunitree-go1

Authors (from arXiv): 作者(来自 arXiv):Gabriel B Margolis, Pulkit Agrawal

More from: 同机构更多条目:MIT麻省理工学院

More in Locomotion & Humanoids更多运动控制与人形机器人条目

All Locomotion & Humanoids entries →查看全部运动控制与人形机器人条目 →