Surface Following using Deep Reinforcement Learning and a GelSightTactile Sensor
This addresses the need for robots to perform dexterous manipulation in unknown or inaccurately modeled environments, representing an incremental improvement in tactile-based control.
The paper tackled the problem of surface following for robots using tactile sensors by proposing an end-to-end deep reinforcement learning approach that maps raw GelSight tactile data to robot motions, achieving 80% proper actions with a trained model and successful autonomous tests.
Tactile sensors can provide detailed contact in-formation that can facilitate robots to perform dexterous, in-hand manipulation tasks. One of the primitive but important tasks is surface following that is a key feature for robots while exploring unknown environments or workspace of inaccurate modeling. In this paper, we propose a novel end-to-end learning strategy, by directly mapping the raw tactile data acquired from a GelSight tactile sensor to the motion of the robot end-effector.Experiments on a KUKA youBot platform equipped with theGelSight sensor show that 80% of the actions generated by a fully trained SFDQN model are proper surface following actions; the autonomous surface following test also indicates that the proposed solution works well on a test surface.