ROBBRIEF
A 1 to 1 joint mapping between a motorized hand rig and a seven joint robot hand unifies in the wild data collection with force feedback, a trade off prior setups could not resolve.
A researcher slides an exoskeleton over their hand, reaches for a mug, and feels the contact just as the robot hand across the room feels it. DITTO, described in a new arXiv preprint as a Dexterous Interface for Transparent TeleOperation, is a research platform for teaching dexterous robot hands by human demonstration.
The system pairs a motorized hand exoskeleton with a seven-joint dexterous robot hand in a 1-to-1 mapping. Each joint on the human-worn rig drives exactly one joint on the robot, which the authors say is what allows the same hardware to support both in-the-wild data collection and bilateral teleoperation with joint-level force feedback. Prior setups forced researchers to choose: teleoperation matched the robot's body at deployment but offered no touch, while handheld capture gave natural force transparency but introduced a visual embodiment gap when the learned policy was transferred.
DITTO attempts to dissolve that trade-off. The exoskeleton spans the operator's natural index-to-thumb workspace, and the authors report learned policies on contact-rich tasks as evidence that the dual use works. The paper is a single preprint, not a deployment claim or a benchmark sweep; what remains open is whether the 1-to-1 mapping generalizes beyond the demonstrated tasks, and whether a platform like this can scale to the data volumes dexterous manipulation still needs.