Tactile Data

Tactile data research focuses on effectively utilizing information from touch sensors to enhance robotic perception and manipulation. Current efforts concentrate on overcoming challenges posed by the high dimensionality of 3D tactile data, often employing techniques like canonical representations, self-supervised pretraining, and generative models (e.g., diffusion models) to learn robust features and improve data fusion with visual information. This research is significant for advancing robotics capabilities in tasks requiring fine motor skills and physical interaction, with applications ranging from dexterous manipulation and object recognition to surface exploration and safe navigation in cluttered environments.