In-Hand Manipulation

In-hand manipulation studies how multifingered robot hands move, rotate, and use objects while maintaining contact, enabling tasks such as reorientation, assembly, and tool use. Approaches include reinforcement learning with simulation-to-real transfer, imitation from human demonstrations, analytical contact and Jacobian-based control, and sensing with tactile, visual, acoustic, or proprioceptive signals. Researchers also design new hand mechanisms and benchmarks to measure how well different hands perform fine manipulation. Real-robot demonstrations are expanding, but reliable generalization across objects, grasps, hand designs, and changing contacts remains difficult.