Omnibody Hand

Reward AI's OM-1 pushes each robot's dynamics into a simulation-trained controller

The model reads hand trajectories, tactile and proximity data from a seven-degree-of-freedom wearable, then emits motion direction, speed, force and grasp timing. Per-machine execution goes to a second controller trained in simulation.