Physics-based Remote Touch

Humans rely on their sense of touch to achieve fine manipulation tasks and interact with their surroundings. However, feeling the tactile properties of a remote environment remains a challenge for two main reasons: first, current robotic systems are rarely equipped with tactile sensing and second, only forces are fed back to the users. In this project, the final goal is to create a bilateral tactile telemanipulation system, bridging the gap for touch over a distance. To cope with the large bandwidth required for this tactile interaction, I propose a new encoding of the tactile information, inspired by human touch. First, I will acquire a large database of robotic mechanical interactions in order to extract primitives of spatio-temporal deformation. Second, this dictionary will allow me to distill tactile features from any interaction and render tactile properties to a remote user.The robotic decoding part will be carried out under the supervision of Prof. Adelson in the Massachusetts Institute of Technology, using the GelSight tactile sensor. In the return phase, I will integrate humans in-the-loop by designing a novel haptic device and studying the effect of applied skin deformation on the perception of tactile properties. This integration will be done under the supervision of Dr. Wiertlewski from Delft University of Technology. This interdisciplinary project requires skills in biomechanics, perception, robotics and control as well as modeling and machine learning, which will be part of the applicant training.If granted, this project will for the first time encode the tactile information in a dictionary of primitives, essential for sensing, compression, and reproduction of rich tactile stimuli. In addition, the remote touch platform will have applications in a wide range of the fields where delicate touch is necessary, such as healthcare, or remote exploration and navigation in environments such as nuclear reactors or space.