DexWild: Dexterous Human Interactions for In-the-Wild Robot Policies

TL;DR

DexWild introduces a scalable, low-cost method for collecting diverse human-robot interaction data using a novel device, enhancing robot policy generalization across environments and embodiments.

Contribution

The paper presents DexWild-System and a co-training framework that leverage human demonstrations to improve dexterous robot manipulation in varied settings.

Findings

Achieved 68.5% success rate in unseen environments, nearly four times higher than robot-only training.

Demonstrated 5.8x better cross-embodiment generalization.

Significant performance improvements over existing methods.

Abstract

Large-scale, diverse robot datasets have emerged as a promising path toward enabling dexterous manipulation policies to generalize to novel environments, but acquiring such datasets presents many challenges. While teleoperation provides high-fidelity datasets, its high cost limits its scalability. Instead, what if people could use their own hands, just as they do in everyday life, to collect data? In DexWild, a diverse team of data collectors uses their hands to collect hours of interactions across a multitude of environments and objects. To record this data, we create DexWild-System, a low-cost, mobile, and easy-to-use device. The DexWild learning framework co-trains on both human and robot demonstrations, leading to improved performance compared to training on each dataset individually. This combination results in robust robot policies capable of generalizing to novel environments, tasks, and embodiments with minimal additional robot-specific data. Experimental results demonstrate that DexWild significantly improves performance, achieving a 68.5% success rate in unseen environments-nearly four times higher than policies trained with robot data only-and offering 5.8x better cross-embodiment generalization. Video results, codebases, and instructions at https://dexwild.github.io