Grasp planning based on dynamics shaping

The present paper proposes a novel grasp planning for a robotic hand to achieve a manipulation task with high accuracy against unknown disturbances. In precision tasks such as a peg-in-hole insertion or a manipulation with a tool, the manipulation accuracy of the task frame on the grasped target is important. We verify that the manipulation task accuracy for the robot hand system against unknown disturbances is determined by the direction of the singular vector and condition number of the output controllability matrix for the manipulation system. The grasp planning based on dynamics shaping is proposed, which finds a grasp configuration so that the error of the task coordinate position toward the direction which requires the high positional precision can be suppressed. The validity of the grasp planning is shown by several simulations concerning a peg-in-hole insertion task.