Grasp2Twist: Learning Bimanual Dexterous Jar Opening by Reinforcement Learning

2026-09-25

AI Summary

This study addresses the challenges of fragmented multi-stage policies, limited continuous twisting, and sim-to-real transfer in bimanual dexterous grasping and sustained cap unscrewing by proposing a unified reinforcement learning-based control framework. The core methodology introduces a geometry-based continuous enclosure metric coupled with a binary indicator to enable smooth transitions from grasping to twisting. Furthermore, a three-stage curriculum learning mechanism is constructed to guide finger contact reconfiguration for unlimited rotation. Experimental results demonstrate that the proposed policy achieves an 88% success rate across six categories of household containers and enables zero-shot sim-to-real transfer. These findings validate the effectiveness of the enclosure metric and curriculum learning in facilitating complex dexterous manipulation tasks.

Abstract

This paper presents Grasp2Twist, a bimanual dexterous manipulation system that learns to grasp and twist open jar lids using reinforcement learning. Learning this task raises three challenges: learning a unified policy for a multi-stage task, sustaining lid twisting, and sim-to-real transfer. To address the first challenge, we introduce a continuous enclosure measure to guide grasp formation and a binary enclosure indicator to guide the grasp-to-twist transition for unified policy learning. We derive both from the geometric relationship between the object center and the convex hull formed by the hand's palm and fingertips. Kinematic constraints limit how far the hand can rotate the lid with fixed contacts, so sustained twisting requires finger contact reconfiguration. We use a three-stage curriculum to facilitate exploration of these contact changes and also improve robustness for sim-to-real transfer. With our approach, the learned policy demonstrates finger gaiting, reconfiguring hand-object contacts to sustain lid rotation. It transfers zero-shot to the physical system and achieves an 88% task success rate across six household containers, including peanut-butter, vitamin, and instant-coffee jars. Ablations further validate the roles of the geometric enclosure in grasp formation and the curriculum in contact-reconfiguration exploration.

Problem

• bimanual dexterous manipulation
• reinforcement learning
• sim-to-real transfer
• contact reconfiguration
• multi-stage task

Innovation

• Bimanual Dexterous Manipulation
• Reinforcement Learning
• Sim-to-Real Transfer
• Curriculum Learning
• Finger Gaiting