A new finger inverse kinematics method for an anthropomorphic hand

Abstract

In this paper, a new method for solving the inverse kinematics of the fingers of an anthropomorphic hand is proposed. Our approach combines a Modified Selectively Damped Least Squares (MSDLS) and Jacobian Transpose (JT) methods. The main advantages of this method with respect to the ordinary SDLS are: optimal Cartesian increment, shorter computation time and better response near singularity configurations. The original JT method exhibits a strong shattering with small magnitudes which occurs near the goal position or in the case of unreachable positions. Like in the SDLS, a damping factor was applied to each input singular vector to filter the undesirable behavior. A comparative study between the MSDLS applied to the inverse Jacobian and JT matrix is developed to investigate manipulator performance in critical end-point positions of the index finger of a commercial anthropomorphic robotic hand and also to evaluate the impact of the increment length on computation time.