This paper presents a manipulation planning approach for robot hands that enables the generation of finger trajectories. The planner is based on a hybrid approach that combines discrete-continuous kinematics using a fully discrete transition system. One of the main contributions of this work consists in the representation of the universe of different submodel combinations, as states in a discrete transition system. The manipulated object geometry is taken into account and the system composed by the object and the hand is modeled as a set of closed kinematical chains. The methodology enables the synthesis of complex manipulation trajectories, when one or more fingers change the contact condition with the object. Contact condition changes include rolling contact, sliding contact, contact loss and contact establishment. Tests were carried out employing a three finger manipulation task in computer simulations and with an experimental setup.
anthropomorphic hand; dexterous manipulation; trajectory planning
