Task Space Robot Hand Manipulation and Optimal Distribution of Fingertip Force Functions

Abstract

Abstract : This research framework investigates the aspects of manipulation based on a four-digit, anthropomorphic robot hand (built for research purposes and designated as the Cybhand). The thesis has looked in detail at three aspects of task-space object manipulation, where position and force control act simultaneously on the object. First, a force feedback control system was constructed. Force sensing was achieved by sensing the tension in tendons. The second aspect, object manipulation and force distribution, investigates the techniques involved in generating torques to move the hand and grasp an object in an optimal way. Optimization of the contact forces was achieved by employing nonlinear inequality constraints with a quadratic cost function to reduce the grasping energy. This method has proven its advantage in reducing the computation time required to converge to an optimal force function. Use of optimization has been based on the computation of the dynamic model of the entire hand and compared with the four-fingered null-space solution or internal forces optimization The third aspect deals with the singularity problem. The research has investigated the problem of an ill-conditioned hand configuration, when the entire hand Jacobian becomes a singular matrix, resulting in infinite forces applied to the object. Robust singularity inversion techniques were used to invert the matrix of the multiple contact system. Using singular value decomposition techniques, it has been shown that: i) the manipulability measure of each finger can be used to invert the hand Jacobian. ii) the manipulability measure of a single finger is sufficient to be used if certain kinematic conditions are satisfied