Real-Time Grasping-Force Optimization for Multifingered Manipulation: Theory and Experiments

Abstract

Real-time grasping-force optimization is a vital problem in dextrous manipulation with multifingered robotic hands. In this paper, we review two prominent approaches that have recently been proposed for this problem, and identify their common need for an appropriate initial condition to start the recursive algorithms. Then, we propose an efficient algorithm that combines a linear-matrix-inequality method with a gradient method for automatic generation of initial conditions. By incorporating the initial conditions, optimal grasping forces can be generated efficiently in real time. Finally, we implement and evaluate all the algorithms for grasping-force control in the Hong Kong University of Science and Technology three-fingered hand platform. Experimental results demonstrate the efficiency of the proposed approach.