ANFIS application for calculating inverse kinematics of programmable universal machine for assembly (PUMA) robot

Abstract

This paper focused on the robot arm's kinematics problem or the connection between angle in each joint and the end-effector's position.f Forward kinematics problem will be deduced using D-H (Denavit- Hartenberg) parameter method. The inverse kinematics problem will be solved using ANFIS (Adaptive Neuro-Fuzzy Inference System) instead of calculating the solution. ANFIS is a feature in MATLAB using ANFIS toolbox. PUMA 560 robot arm virtual model is used in this paper. Three ANFIS training conditions are made to test the influence of training conditions with the result's performance. The difference between end effector's position that using ANFIS and from calculation in forward kinematics will be calculated to test the end effector's position error. By making ANFIS solutions with three different MFs (Membership Functions), influence of MF number are known. With more MF will decrease the position's error. The most MF in this paper is 10 MFs resulting position error by 27.974mm.