Remarks on Control of a Robot Manipulator using a Quaternion Recurrent Neural-Network-Based Compensator
Kazuhiko Takahashi, Lisa Watanabe, Hidemi Yamasaki, Satoka Hiraoka, Masafumi Hashimoto · 2020
This paper presents and investigates practical applications of a recurrent quaternion neural network in controlling a robot manipulator within a desired trajectory. To effectively track the robot manipulator's end-effector position against the desired trajectory, computed torque control is used to design a control system, and the recurrent quaternion neural network is used to synthesise an external input to compensate for the system's control input. Computational experiments for controlling a 3-link robot manipulator are conducted using the control system. Experimental results demonstrate the effectiveness of using the recurrent quaternion neural network for practical control applications in a control system.