Dextrous hand grasping strategies using preshapes and digit trajectories

Inspired by human grasping behavior, use of hand pre-shapes has for some time been recognized as a useful way of reducing the complexity of planning grasps for a dextrous hand. We define four types of task-specific preshape and two modes of digit closure, and give criteria for the choice and positioning of the pre-shapes, based on intersecting digit trajectories with the graspable surfaces of the object. Grasps are planned in simulation using real range data from a laser striper and an anthropomorphic hand model. Our results show that a range of polyhedral and curved objects can be grasped using relatively simple, fast algorithms when the hand movements are constrained in this way