Shared Grasping: a Combination of Telepresence and Grasp Planning

Telepresence systems can be used for repair and maintenance in environments that are hard to access or dangerous for the human. Current systems can effectively the gross motor skills of the human operator into actions in the remote environment. In comparison, fine manipulation, like the secure grasping and handling of complex objects, is error-prone and strenuous for the operator. To execute such tasks in a more precise and robust way, this thesis increases the level of autonomy of a complex telepresence system in order to help the operator during the interaction with the remote environment. The focus of the work is to assist teleoperated grasp and manipulation tasks, which the operator should be able to execute with the lowest possible workload. Therefore, immersion is crucial, and hence, the developed approach does not decrease the high immersion of the human in a telepresence system, but retains important properties like haptic force feedback and 3D stereo view on the remote environment. The operator stays in the loop during prehension and participates actively by commanding continuously the position of the end effector and hence, the hand position from where the object is grasped. To realize this seamless integration of semi-autonomy in a telerobotic system, first, a new grasp planner is developed. It computes physically feasible and robust grasps fast enough to cope with the movement speed of the operator. The main feature of the approach is, next to the low computation time, the consideration of contact forces that the robotic end effector is able to apply on the surface of the object. Hence, the assumption of normalized contact forces, common in state of the art approaches, is overcome and the grasp can be evaluated realistically with respect to external disturbance forces. Second, an assistance for positioning the end effector relative to the object is developed, which combines a dense database of grasps with the concept of adaptive virtual fixtures. This enables the operator to determine the approach direction towards the object