Multi-Modal World Model for Physical Robot Interactions: Simultaneous Visual and Tactile Predictions for Enhanced Accuracy

arXiv:2304.11193v2 Announce Type: replace-cross Abstract: Predicting the outcomes of robotic actions, often referred to as learning a world model, in complex environments remains a fundamental challenge in robotics. Existing approaches primarily rely on visual observations and action inputs to...