DexCtrl: Towards Sim-to-Real Dexterity with Adaptive Controller Learning

Fuente: arXiv
Guardado en:
Detalles Bibliográficos
Autores principales: Zhao, Shuqi, Yang, Ke, Chen, Yuxin, Li, Chenran, Xie, Yichen, Zhang, Xiang, Wang, Changhao, Tomizuka, Masayoshi
Formato: Preprint
Publicado: 2025
Materias:
Acceso en línea:
Etiquetas: Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
_version_ 1866916716913623040
author Zhao, Shuqi
Yang, Ke
Chen, Yuxin
Li, Chenran
Xie, Yichen
Zhang, Xiang
Wang, Changhao
Tomizuka, Masayoshi
author_facet Zhao, Shuqi
Yang, Ke
Chen, Yuxin
Li, Chenran
Xie, Yichen
Zhang, Xiang
Wang, Changhao
Tomizuka, Masayoshi
contents Dexterous manipulation has seen remarkable progress in recent years, with policies capable of executing many complex and contact-rich tasks in simulation. However, transferring these policies from simulation to real world remains a significant challenge. One important issue is the mismatch in low-level controller dynamics, where identical trajectories can lead to vastly different contact forces and behaviors when control parameters vary. Existing approaches often rely on manual tuning or controller randomization, which can be labor-intensive, task-specific, and introduce significant training difficulty. In this work, we propose a framework that jointly learns actions and controller parameters based on the historical information of both trajectory and controller. This adaptive controller adjustment mechanism allows the policy to automatically tune control parameters during execution, thereby mitigating the sim-to-real gap without extensive manual tuning or excessive randomization. Moreover, by explicitly providing controller parameters as part of the observation, our approach facilitates better reasoning over force interactions and improves robustness in real-world scenarios. Experimental results demonstrate that our method achieves improved transfer performance across a variety of dexterous tasks involving variable force conditions.
format Preprint
id arxiv_https___arxiv_org_abs_2505_00991
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle DexCtrl: Towards Sim-to-Real Dexterity with Adaptive Controller Learning
Zhao, Shuqi
Yang, Ke
Chen, Yuxin
Li, Chenran
Xie, Yichen
Zhang, Xiang
Wang, Changhao
Tomizuka, Masayoshi
Robotics
Systems and Control
Dexterous manipulation has seen remarkable progress in recent years, with policies capable of executing many complex and contact-rich tasks in simulation. However, transferring these policies from simulation to real world remains a significant challenge. One important issue is the mismatch in low-level controller dynamics, where identical trajectories can lead to vastly different contact forces and behaviors when control parameters vary. Existing approaches often rely on manual tuning or controller randomization, which can be labor-intensive, task-specific, and introduce significant training difficulty. In this work, we propose a framework that jointly learns actions and controller parameters based on the historical information of both trajectory and controller. This adaptive controller adjustment mechanism allows the policy to automatically tune control parameters during execution, thereby mitigating the sim-to-real gap without extensive manual tuning or excessive randomization. Moreover, by explicitly providing controller parameters as part of the observation, our approach facilitates better reasoning over force interactions and improves robustness in real-world scenarios. Experimental results demonstrate that our method achieves improved transfer performance across a variety of dexterous tasks involving variable force conditions.
title DexCtrl: Towards Sim-to-Real Dexterity with Adaptive Controller Learning
topic Robotics
Systems and Control
url https://arxiv.org/abs/2505.00991