Multifingered hand system which can acquire the object model by manipulation with a multifingered hand

This paper describes a multifingered hand system which can acquire the object model while manipulating the object with multifingered hands. The multifingered hand has 4 fingers and a 6-axis force-torque sensor are mounted on each fingertip. The multifingered hand system classify the finger into grasping fingers (gf) and manipulating fingers (mf). The system has 4 grasp modes which is classified by the number of grasping fingers, and has 12 grasping force patterns. The contact faces of the grasped object can be calculated by using the contact point information, when the object is a polyhedron. Acquisition of an object model is achieved by connecting the contact faces calculated by using the contact point information. When the multifingered hand grasps an unknown object, the system select the grasping force pattern so as to achieve stable grasp. At the same time, an initial model of the grasped object is acquired. Next, the hand starts manipulating the object. During manipulation, the system checks whether the contact face at the fingertip changes or not. If the contact face of the fingertip changes, new contact face is added to the object model. The object model is constructed incrementally with the progress of manipulation. Finally, experimental results are shown to demonstrate the effectiveness of the proposed scheme.