Inverse Kinematics Based on Fuzzy Logic and Neural Networks for the WAM-Titan II Teleoperation System

The inverse kinematic problem is crucial for robotics. In this paper, a solution algorithm is presented using artificial intelligence to improve the pseudo-inverse Jacobian calculation for the 7-DOF Whole Arm Manipulator (WAM) and 6-DOF Titan II teleoperation system. An investigation of the inverse kinematics based on fuzzy logic and artificial neural networks for the teleoperation system was undertaken. Various methods such as Adaptive Neural-Fuzzy Inference System (ANFIS), Genetic Algorithms (GA), Multilayer Perceptrons (MLP) Feedforward Networks, Radial Basis Function Networks (RBF) and Generalized Regression Neural Networks (GRNN) were tested and simulated using MATLAB. Each method for identification of the pseudo-inverse problem was tested, and the best method was selected from the simulation results and the error analysis.

From the results, the Multilayer Perceptrons with Levenberg-Marquardt (MLP-LM) method had the smallest error and the fastest computation among the other methods. For the WAM-Titan II teleoperation system, the new inverse kinematics calculations for the Titan II were simulated and analyzed using MATLAB. Finally, extensive C code for the alternative algorithm was developed, and the inverse kinematics based on the artificial neural network with LM method is implemented in the real system. The maximum error of Cartesian position was 1.3 inches, and from several trajectories, 75 % of time implementation was achieved compared to the conventional method. Because fast performance of a real time system in the teleoperation is vital, these results show that the new inverse kinematics method based on the MLP-LM is very successful with the acceptable error.