| This research has been concerned about manipulation by using robot hands and has three major aspects. They are (i) reconstruction of the contact information, (ii) calculation of distributed forces, and (iii) the control strategies.; We began with the test of the six-axis force sensor. The mapping matrix between the output signal and the resultant wrench was built by analyzing the test data. The ambiguity problem for soft point contact has been shown to be nonexistent if the surface of the fingertip is convex. By integrating the ambiguity analysis and the equations derived by the other researchers, the contact information: contact position, force, and moment can be correctly reconstructed.; The contact forces applied by the fingers were solved for by optimizing the contact friction angles. Three steps: projection, residual wrench equilibrium, and optimization were used to solve the problem. All of the pseudoinverse method, the velocity field method, and the force field method can be used to solve the distributed forces which equilibrate the residual wrench. The relationship between these three methods was discussed. The first and the second steps can provide a good initial point for the optimization process. Therefore, a solution close to the best friction angles can be found efficiently.; In order to manipulate the grasped object, the computed torque method has been combined successfully with a force allocation rule which optimizes the friction angles. All of the force feedback, regrasping, and rolling motion were considered. The method proposed was based on three dimensional manipulation and can be applied to any tasks as long as the desired trajectory is defined. The simulation examples showed that the object was moved along the desired trajectory very well and the friction angles were well optimized. The time needed to find the commanded torque is only 5.2 ms when running on an 80486 (50 MHz) personal computer for the four finger case. It is probably quick enough for real-time control.; The work developed in this research was based on the model of the DIGITS system. However, the proposed strategies and techniques can be applied to other multifingered systems or even to a walking machine with minor modification. |
