Design of adaptive NNs-robust-PID controller for a robot control

This paper investigates the trajectory control of a robot using a new type of recurrent neural network. A three-layered recurrent neural network is used to estimate the forward dynamics model of the robot manipulator. The standard backpropagation (BP) algorithm is used as a learning algorithm for this network to minimise the difference between the robot manipulator actual response and that predicted by the neural network. This algorithm is employed to update the connection weights of a recurrent neural network controller with three layers using a stochastic gradient function. The control architecture consists of a neural feed-forward model which is a recurrent network used for identification of the robot dynamics, a conventional PID controller, a robust controller and a neural controller. Simulations illustrate that the proposed neural control approach which is applied to some nonlinear processes can gain satisfactory performance results. The results of the simulations are presented to show the promising performance of the neural controller.