Regrasping behavior and object transition generated by EP

Abstract

We have studied the generation of regrasping motion for a four-fingered robot using evolutionary programming (EP). EP has the advantage of finding the set of optimal numerical values but it requires many iterations to find it. The regrasping strategy obtained cannot be applied to other object shapes than those used in the search process. We consider that the regrasping strategy generated by EP should be reused for other sizes of object. We propose an expansion of the method for other sizes of grasped object. We show the effectiveness of the method with experimental results.