MIT News November 19, 2024
Fast and accurate physics simulation is an essential component of robot learning, where robots can explore failure scenarios that are difficult to produce in the real world and learn from unlimited on-policy data. Yet, it remains challenging to incorporate RGB-color perception into the sim-to-real pipeline that matches the real world in its richness and realism. Researchers at MIT trained a robot dog in simulation for visual parkour. They proposed a way to use generative models to synthesize diverse and physically accurate image sequences of the scene from the robot’s ego-centric perspective. They presented demonstrations of zero-shot transfer to the RGB-only observations of the real world on a robot equipped with a low-cost, off-the-shelf color camera. website … read more. Open Access TECHNICAL ARTICLE