Low-Level Finger Coordination for Compliant Anthropomorphic Robot Grippers

The execution of a grasp process with an anthropomorphic multi-finger robot gripper has to base on an appropriate movement planning and an accurate low-level control strategy. Stability problems may arise, if there is a deviation, e.g. through position, orientation and/or size perturbation of the object, between planned and real executable movement patterns. One opportunity to react to such perturbations is the use of visual information. But if the there is no visual system available or even the object is concealed, e.g. through the robot gripper, the robot arm or another object, then the use of a visual system is insufficient. For this reason a low-level finger coordination depending on the actual state of the grasp (grasp phase) is presented. With this concept it is possible to coordinate the grasp on the level of the low-level controller and therefore to improve the grasp performance.