Kaifeng Zhang

Robots can learn contact-rich dexterity directly from human skin: co-training on aligned human visual-tactile-action data more than doubles real-world manipulation success (70% vs. 29%) with zero performance saturation up to 1,000 demonstrations.

The proposed method accelerates training and enables stable policies where matched raw-action PPO remains near failure, with successful sim-to-real transfer in in-hand and arm-hand tasks.