EaDex: A Cross-Embodiment Dexterous Manipulation Framework from Low-Cost Demonstrations

2026-06-02

AI Summary

Learning dexterous manipulation is hindered by the high cost of acquiring high-quality demonstration data and the low sample efficiency of reinforcement learning exploration. This work proposes the EaDex framework, which leverages a monocular RGB-D camera to capture human hand motions and combines MANO modeling, data normalization, and motion retargeting to construct structured, low-cost demonstrations. To bridge the gap between imitation and autonomous policy improvement, the method introduces a contact-aware reward-based dynamic demonstration annealing mechanism that enables a smooth transition from demonstration-guided learning to self-optimization. Evaluated on three object-opening tasks across nine cross-embodiment settings, the approach achieves a 55.3% higher success rate compared to a non-annealing baseline, demonstrating its effectiveness and strong generalization capability across diverse robotic hands.

Abstract

Dexterous manipulation learning has long been hindered by the high costs of data and training, as pure reinforcement learning typically requires large-scale interactive exploration and imitation learning depends on high-quality demonstrations that are expensive to collect. To address this problem, we propose EaDex, a multi-embodiment dexterous manipulation learning framework under low-cost demonstration conditions, which enables rapid generation of demonstration data and consequently reduces training time for efficient dexterous manipulation. At the data level, EaDex captures human hand motions using only a single RGB-D camera and constructs structured demonstration data through MANO-based hand modeling, data normalization, and motion retargeting. At the learning level, we introduce a contact-reward-based dynamic demonstration annealing mechanism, which guides early-stage exploration under demonstration and gradually transitions to autonomous optimization with accumulating contact rewards. Using our custom dataset, we evaluate EaDex on three dexterous hands and three articulated object-opening tasks, covering nine cross-embodiment manipulation settings, achieving a 55.3% relative improvement over the baseline without demonstration annealing. These results validate the effectiveness of the proposed low-cost demonstration pipeline and the dynamic demonstration annealing strategy for dexterous manipulation learning.

Problem

• dexterous manipulation
• imitation learning
• high-cost demonstrations
• reinforcement learning
• data efficiency

Innovation

• dexterous manipulation
• low-cost demonstration
• dynamic demonstration annealing
• cross-embodiment learning
• contact reward