Tactile Experience-based Robotic Grasping

摘要

We propose an experience-based approach to the problem of blind grasping, stable robotic grasping using tactile sensing and hand kinematic feedback. We first collect a set of stable grasps to build a tactile experience database which contains tactile contacts for each stable grasp. Using the tactile experience database, we propose an algorithm to synthesize local hand adjustment that controls the hand pose and improves the grasp based on tactile sensor readings. Simulation experiments show that local adjustment of the hand improves the grasping performance.