Approximating a Robot Inverse Kinematics Solution Using Fuzzy Logic Tuned by Genetic Algorithms

Abstract

A new scheme based on recursive fuzzy logic is presented in this paper for solving the point-to-point inverse kinematics problem of serial robots. To improve the convergence problem in the whole workspace, the membership functions of the fuzzy logic are searched for, tuned, and optimised using a simple genetic algorithm. A dominant joint, which brings the end-effect closer to the desired target, has to be selected before the implementation of the fuzzy logic in order to reduce the number of fuzzy logic iterations. The inverse kinematics solution of robots is usually obtained by direct inversion of the kimenatics equations, but this technique often leads to a singular Jacobian matrix during the calculations. The work presented in this paper provides a direct approach to the calculation of the kinematics inverse problem which bypasses the kinematic singularities. Computer simulations of the proposed scheme confirm the findings of the theoretical developments.