Grasping Torque Optimization for a Dexterous Robotic Hand Using the Linearization of Constraints

Abstract

A new algorithm is proposed for the optimization of grasping torques. Previous work revealed that the optimization of the contact force can be simplified as a linear programming problem by replacing the nonlinear friction cone with a polyhedral cone. On this basis, further simplification and linearization is completed for various grasp contact constraints of the grasping system, and an optimization model is established with the minimum sum-of-squares of joint torques as an objective function. A timely and effective algorithm based on key constraint sets is then established. This methodology optimizes the joint torque directly and facilitates the control of the dexterous hand. A numerical example and experimental results show that the optimization algorithm has high precision and good real-time performance. The joint torque output obtained using the proposed algorithm was 27.6% lower than the result of the previous algorithm. The optimization algorithm can thus be used for optimal grasping control of a dexterous hand.