ROBOTERA opened orders for its XHAND 1 Pro-U on November 16, with an initial release of 40 units. The five-finger hand was used in NVIDIA’s GR00T 2 demonstration at the Conference on Robot Learning. The launch release gives no price.

The hand has 16 degrees of freedom and joints aligned with NVIDIA’s DexUMI wearable exoskeleton. A person handles real objects while DexUMI records movement and fingertip forces. The recordings train a model that then runs on a two-arm robot fitted with the hands.

Teaching without occupying the robot

ROBOTERA says training used DexUMI demonstrations “without additional task-specific teleoperation or extra robot play data.” Its point is about collecting training data: the demonstrator need not remotely drive the target robot. It does not mean the robot learned without human examples.

The company reports task success of 60% for banana yogurt preparation, 80% for granola scooping and 75% for a match-and-lamp sequence. It provides no trial counts in the release.

For a research lab, the attraction is a way to capture hand skills directly. Before treating those percentages as a buying guide, ask how many attempts were run, which failures counted and whether the results hold with different objects.

Source: ROBOTERA’s November 16 announcement, distributed by PR Newswire.