Geometrical approach for grasp synthesis on discretized 3d objects

Grasp synthesis on real 3D objects is a critical problem in grasp and manipulation planning. This paper presents a geometrical approach to compute force closure (FC) grasps, with or without friction and with any number of fingers. The object's surface is discretized in a cloud of points, so the algorithm is applicable to objects of any arbitrary shape. One or more FC grasps are obtained with a geometrical approach, which embeds the FC test in the algorithm to simplify achieving the force-closure property. This initial FC grasp may be improved with a complementary optimization algorithm. The grasp quality is measured considering the largest perturbation wrench that the grasp can resist with independence of the perturbation direction. The efficiency of both algorithms is illustrated through numerical examples.