Simplified generation algorithm of regrasping motion - performance comparison online-searching approach with EP-based approach

This paper proposes a simplified generation method of regrasping motion. The regrasping motion is designed by evolutional programming (EP) in our conventional study and the obtained regrasping behavior should be adjusted according to some conditions; initial posture and dimensions of a grasping object. However, the designing process of the regrasping motion is off-line because it takes several minutes to find the optimal regrasping strategy by EP. In this paper, we therefore simplified the searching process for regrasping motion in order to apply it in the realtime process. Besides, this simplified generation method is enable to achieve a regrasping motion from arbitrary initial posture because there is no strategy that is previously designed. We show its effectiveness of the simplified generation method with numerical simulations, comparing the performance of the EP-based approach.