Real Time Implementation of NARMA-L2 Control of a Single Link Manipulator

Abstract

Nonlinearities and parametric uncertainties are unavoidable problems faced in controlling robot manipulator. A single link manipulator driven by a permanent magnet brushed dc motor is a nonlinear dynamics due to effects of gravitational force, mass of the payload, posture of the manipulator and viscous friction coefficient. Furthermore, uncertainties arise because of changes of the rotor resistance with temperature and random variation of friction while operating. Due to this fact, classical PID controller can not be used effectively since it is developed based on linear system theory. In order to overcome this problem, in this research, a neural network control scheme, NARMA-L2 Control is adopted and implemented in real time for controlling a DC motor driven single link manipulator with unknown dynamics. However, the real time experimentation showed that the proposed system results in chattering of the control signal. Hence, the system also chatters within the desired trajectory. As a solution, real time Smoothed NARMA-L2 Control scheme is implemented. Physical results showed that the improved control scheme has not only reduced the chattering but has successfully controlled the single link manipulator for both point-to-point and continuous path motion control.