Abstract: An adaptive neuro-fuzzy inference system (ANFIS) based inverse kinematics solution of a robotic manipulator is presented. The research object is SCARA type four degrees of freedom manipulator. The corresponding relationship between robot end-effector Cartesian coordinates and manipulator joints space is studied. We select the starting point, intermediate point and end point according to the target trajectory, and use the traditional method to obtain the corresponding joint variable values. Then we get the joints space curve by using cubic splines. Sample points are randomly selected from the curve to train the ANFIS. The results show that ANFIS in calculating precision and computing time of inverse kinematics is superior to the BP neural network.

Key words: adaptive neuro-fuzzy inference system, inverse kinematics solution, robotics;cubic spline, neural network