A Neuro-Fuzzy Solution for Inverse Kinematics of Redundant Manipulators

Abstract

Abstract: This paper presents a novel approach for solving the inverse kinematics (IK) problem of redundant robotic manipulators using an Adaptive Neuro-Fuzzy Inference System (ANFIS). Redundant manipulators, characterized by having more degrees of freedom (DOF) than necessary for a task, introduce complexity in their IK solutions due to their non-linear, time-varying, and transcendental nature. Traditional methods such as algebraic or geometric solutions are often computationally intensive and can suffer from issues such as singularities. This study leverages ANFIS, combining the learning capability of neural networks with the logic-based structure of fuzzy systems, to predict IK solutions for 5-DOF and 7-DOF manipulators. The Denavit-Hartenberg (D-H) notation is employed for the kinematic modeling of robot links. Comparative analysis of the predicted IK values using ANFIS demonstrates its high accuracy and efficiency, as validated by surface and residual plots. The results show that ANFIS can be effectively utilized for fast and reliable IK predictions, offering a robust alternative to traditional approaches. This method can be applied in robotic systems requiring high precision in constrained or dynamic environments.