Remarks on a Commutative Quaternion Neural Network–based Controller and Its Application in Controlling a Robot Manipulator

Abstract

This study examines the possibility of using a commutative quaternion neural network in control system applications. A multi-layer commutative quaternion neural network and its training algorithm are derived and the network is applied to develop a feedforward-feedback controller, with the network input consisting of a reference output and some tapped-delay input-output sets of the controlled plant while the network output is employed to synthesise the control input. Training of the commutative quaternion neural network in the control system is conducted in real-time by integrating feedback error learning. To evaluate the effectiveness of a commutative quaternion neural network-based controller, computational experiments on trajectory tracking control of a three-link robot manipulator are conducted. Simulation results show the suitability of the commutative quaternion neural network for controlling the robot manipulator and the characteristics of the commutative quaternion neural network-based controller are clarified when compared with those of a quaternion neural network-based controller.