Modeling and control of one-link robotic manipulator using neural network based PID controller

The dynamics of one-link robotic manipulator is complex and non linear and hence, cannot be easily controlled by conventional PID controller. The severity of the problem further increases when the plant's mathematical model is unknown or partially known which makes the use of PID control more difficult because it requires the dynamics of the system for tuning its parameters. Even if the dynamics are known, the parameters of PID controller are required to be retuned when external disturbance signals and/or parameter variations occurs in the system. In this paper, the PID controller is implemented using a multilayer feed forward neural network (MLFFNN) for the desired trajectory tracking control of one-link robotic manipulator (plant). To make the controller adaptive, the dynamics of plant is assumed to be unknown and hence, a separate multilayer feed forward neural network identification model is used which will approximate the plant's dynamics and operate simultaneously with the controller. The other benefits of using an identification model is that it can adjust its own parameters to reflect the effects of the disturbance signal and parameter variations on the system and provides this information to the controller which then makes necessary adjustment to its output to compensate these effects. Simulation results shows that MLFFNN based PID controller is able to control the plant and provides the desired trajectory in the presence of parameter variations and disturbance signal.