Haptic-Guided Shared Control Grasping for Collision-Free Manipulation

We propose a haptic-guided shared control system that provides an operator with force cues during reach-to-grasp phase of tele-manipulation. The force cues inform the operator of grasping configuration which allows collision-free autonomous post-grasp movements. Previous studies showed haptic guided shared control significantly reduces the complexities of the teleoperation. We propose two architectures of shared control in which the operator is informed about (1) the local gradient of the collision cost, and (2) the grasping configuration suitable for collision-free movements of an aimed pick-and-place task. We demonstrate the efficiency of our proposed shared control systems by a series of experiments with Franka Emika robot. Our experimental results illustrate our shared control systems successfully inform the operator of predicted collisions between the robot and an obstacle in the robot's workspace. We learned that informing the operator of the global information about the grasping configuration associated with minimum collision cost of post-grasp movements results in a reach-to-grasp time much shorter than the case in which the operator is informed about the local-gradient information of the collision cost.