Kinematics-Aware Multi-Policy Reinforcement Learning for Force-Capable Humanoid Loco-Manipulation

Fuente: arXiv
Guardado en:
Detalles Bibliográficos
Autores principales: Xiao, Kaiyan, Xu, Zihan, Zhe, Cheng, Liu, Chengju, Chen, Qijun
Formato: Preprint
Publicado: 2025
Materias:
Acceso en línea:
Etiquetas: Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
_version_ 1866914171491188736
author Xiao, Kaiyan
Xu, Zihan
Zhe, Cheng
Liu, Chengju
Chen, Qijun
author_facet Xiao, Kaiyan
Xu, Zihan
Zhe, Cheng
Liu, Chengju
Chen, Qijun
contents Humanoid robots, with their human-like morphology, hold great potential for industrial applications. However, existing loco-manipulation methods primarily focus on dexterous manipulation, falling short of the combined requirements for dexterity and proactive force interaction in high-load industrial scenarios. To bridge this gap, we propose a reinforcement learning-based framework with a decoupled three-stage training pipeline, consisting of an upper-body policy, a lower-body policy, and a delta-command policy. To accelerate upper-body training, a heuristic reward function is designed. By implicitly embedding forward kinematics priors, it enables the policy to converge faster and achieve superior performance. For the lower body, a force-based curriculum learning strategy is developed, enabling the robot to actively exert and regulate interaction forces with the environment.
format Preprint
id arxiv_https___arxiv_org_abs_2511_21169
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Kinematics-Aware Multi-Policy Reinforcement Learning for Force-Capable Humanoid Loco-Manipulation
Xiao, Kaiyan
Xu, Zihan
Zhe, Cheng
Liu, Chengju
Chen, Qijun
Robotics
Humanoid robots, with their human-like morphology, hold great potential for industrial applications. However, existing loco-manipulation methods primarily focus on dexterous manipulation, falling short of the combined requirements for dexterity and proactive force interaction in high-load industrial scenarios. To bridge this gap, we propose a reinforcement learning-based framework with a decoupled three-stage training pipeline, consisting of an upper-body policy, a lower-body policy, and a delta-command policy. To accelerate upper-body training, a heuristic reward function is designed. By implicitly embedding forward kinematics priors, it enables the policy to converge faster and achieve superior performance. For the lower body, a force-based curriculum learning strategy is developed, enabling the robot to actively exert and regulate interaction forces with the environment.
title Kinematics-Aware Multi-Policy Reinforcement Learning for Force-Capable Humanoid Loco-Manipulation
topic Robotics
url https://arxiv.org/abs/2511.21169