This Motorized Glove Lets You Feel What a Robot Hand Touches

Robot hands are clumsy compared with yours, and a big reason is training data. To teach a robot to pick up a grape or plug in a cable, someone has to show it — usually by remote-controlling it like a puppeteer, or by wearing a motion-tracking glove while doing the task. Both approaches have a flaw. Remote control matches how the robot actually moves, but the operator can't feel anything. A handheld glove feels natural, but the robot's hand and eyes don't experience the world the way the human did.

Researchers at Columbia University built a fix. DITTO is a pair of matched devices: a robotic hand with seven independent joints (roughly the range of movement in a human hand), and a motorized exoskeleton you wear on your hand. Every motor in the wearable glove lines up one-to-one with a motor in the robot hand. Move your fingers and the robot copies you instantly. When the robot presses against something, the glove pushes back on your fingers, so you feel the surface.

That combination — natural movement plus real feeling — matters because touch is how humans handle fragile things without crushing them. The team showed the system covers the natural range of motion between thumb and index finger, and used it to teach robots tasks that require contact: pressing, gripping, steadying objects. Those are exactly the jobs that trip up today's robots.

The catch: this is a lab prototype, not a product. It covers one hand, focuses on thumb-and-index pinching, and comes with no price or ship date. Robots also still need enormous amounts of demonstration data before they can work unsupervised. But if collecting that data gets cheaper and easier, the helper robots pitched for warehouses, hospitals and homes get closer to being genuinely useful.