NoContactNoWorries: Estimating Contact through Vision and Proprioception for In-Hand Dexterous Manipulation

This paper introduces NoContactNoWorries, a multimodal transformer framework that fuses RGB-D vision with robot proprioception to infer binary contact states as a pseudo-tactile signal for in-hand dexterous manipulation. The authors validate the approach by training a single model across multiple objects and demonstrating that the inferred contact signal supports downstream reinforcement learning agents for object reorientation, generalizing to novel objects. Experiments conducted in both simulation and on a real-world robot highlight the feasibility of inferring contact from vision and proprioception as a scalable alternative to dedicated tactile hardware.