DexEXO: A Wearability-First Dexterous Exoskeleton for Operator-Agnostic Demonstration and Learning
🤖 AI Summary
This work addresses the challenges of scaling high-quality demonstration data collection across users in dexterous robot learning and the limitations of existing wearable devices in comfort, universality, and kinematic alignment. To this end, the authors propose DexEXO, a wearability-first hand exoskeleton featuring a pose-tolerant thumb mechanism and a slider-based finger interface that accommodates hand lengths from 140 to 217 mm without requiring user-specific calibration. Its passive structure aligns naturally with the wearer’s anatomy, eliminating the need for post-hoc visual alignment or processing. Integrated with wrist-mounted RGB observations and diffusion-based policy training, DexEXO enables efficient end-to-end learning of high-performance policies using only raw visual inputs. User studies demonstrate that DexEXO significantly outperforms existing systems in comfort and usability while substantially streamlining the end-to-end imitation learning pipeline.
Application Category
📝 Abstract
Scaling dexterous robot learning is constrained by the difficulty of collecting high-quality demonstrations across diverse operators. Existing wearable interfaces often trade comfort and cross-user adaptability for kinematic fidelity, while embodiment mismatch between demonstration and deployment requires visual post-processing before policy training. We present DexEXO, a wearability-first hand exoskeleton that aligns visual appearance, contact geometry, and kinematics at the hardware level. DexEXO features a pose-tolerant thumb mechanism and a slider-based finger interface analytically modeled to support hand lengths from 140~mm to 217~mm, reducing operator-specific fitting and enabling scalable cross-operator data collection. A passive hand visually matches the deployed robot, allowing direct policy training from raw wrist-mounted RGB observations. User studies demonstrate improved comfort and usability compared to prior wearable systems. Using visually aligned observations alone, we train diffusion policies that achieve competitive performance while substantially simplifying the end-to-end pipeline. These results show that prioritizing wearability and hardware-level embodiment alignment reduces both human and algorithmic bottlenecks without sacrificing task performance. Project Page: https://dexexo-research.github.io/
Problem
Research questions and friction points this paper is trying to address.
dexterous robot learning
wearable exoskeleton
cross-operator demonstration
embodiment mismatch
scalable data collection
Innovation
Methods, ideas, or system contributions that make the work stand out.
wearable exoskeleton
embodiment alignment
cross-operator adaptability
diffusion policy
dexterous teleoperation
🔎 Similar Papers
No similar papers found.
