J-GLOBAL ID:200904098860509643  Research Project code:0150001179 Update date:Mar. 14, 2006

Research on manipulation with a multifingered hand which control a continuous contact with an object

Study period:1997 - 2000
Research overview:
An object model is needed to manipulate the object with a multifingered hand. There are a wide variety of objects in our environment, making it difficult to construct a model for each and every object in advance. To construct all these models will require many man-hours at a substantial cost. To manipulate a variety of objects, we developed a multifingered hand system which acquires an object model while manipulating the object. The multifingered hand has 6-axis force-torque sensor at each fingertip. A contact point between the fingertip and a grasped object can be calculated by using the force-torque data at the fingertip. The contact surface of an object can be calculated, due to the fact that the fingertip and the objects surface have the same normal vector at contact point. For a polyhedral object, the contact surface calculated from the force-torque data coincide with one facet of the polyhedral object. Acquisition of an object model is achieved by connecting the contact facets. During manipulation, the system checks whether the contact facet change or not. If the contact facet changes, the new facet is added to the object model. The acquired object model is used to calculate a position and orientation of the object for the next motion. The grasping force pattern to achieve a stable grasp, and a withdrawing finger for manipulation, are selected by using contact point information. Human beings manipulate a variety of objects by using their hands. In order to achieve similar manipulation tasks by a robot, it is necessary to obtain adequate interpretation of grasping and manipulation tasks that humans perform. A lot of research has been conducted on the dynamics of grasping, but there was no method available to interpret and describe grasping and manipulations systematically in order to derive optimal grasping mechanisms for the target tasks. For this purpose, we have developed a method to interpret and describe manipulation tasks, in simple expressions that is useful for the consideration of the structure of those tasks. In this method, grasping and manipulation are described from fundamental functional units. We can obtain abstract descriptions which are independent from specific hardware and a number of hands. It can also be used to derive novel mechanisms that can achieve the same grasping states in a completely different manner. We have also proposed a new method for reducing the weight of a robot hand by exchanging hand units of simple functions. We invented a simple mechanical joint system, which requires no special actuators for exchanging hands, and we have also developed basic hand pieces equipped with the system.
Keywords (7):
multifingered hand ,  force-torque sensor ,  exchangeable hand ,  desorption mechanism ,  object model ,  contact point ,  motion capture
Research program: Special Research and Development
Ministry with control over the research :
Ministry of Economy, Trade and Industry
Research budget: 2000: \13,000,000

Return to Previous Page