Compact Visuotactile World Models for Lifting: Prediction, Reward Alignment, and Force Constraints
A 652,157-parameter action-conditioned visuotactile world model with matched behavior cloning, policy learning in imagination, independent reactive implicit Q-learning, and model-assisted force feedback is studied, finding that accurate tactile forecasts need not improve force-constrained control.
Qing-Zhen Ma
· 0 citations