Trajectory Generation for Juggling Considering Dynamics of Multifingered Hand-arm

Abstract

To realize human-like dynamic manipulation, robotic juggling has been studied so far, in which simple end effectors were used. The purpose of this research is to achieve juggling with a multi-fingered hand-arm like a human. In the case that a multi-fingered hand arm is used, it becomes a serious problem that the weight of the multi-fingered hand is heavy, and it is necessary to optimize the movement of the arm in order to compensate for the hand's heavy weight. In this paper, we propose a method to optimize motion trajectory using a smoothing spline. Parameters of the smoothing spline are optimized offline using a genetic algorithm. The robot hand-arm performs an optimum throw-up operation online according to the falling position of a ball using the optimized parameters. We show the effectiveness of the proposed method by experiments, in which the hand-arm realized throw-up and catch motions in seven consecutive times. Moreover, we show that the proposed method reduces the magnitude of the joint torques and reduces the tracking error on the target trajectory, as compared with conventional methods.