Stability analysis of robot manipulators subject to feedforward neural network controllers

In this paper a new approach, based on Lyapunov's direct method, for the design of multilayer feedforward neural network (NN) controllers for uncertain robot manipulators is presented. Furthermore, two different feedforward NN controllers are proposed to stabilize uncertain robot manipulators. The first NN controller is shown to render the closed-loop system globally practically stable while the second NN controller guarantees global uniform asymptotic stability of the system. The new approach ensures stability of the control system without using any learning or adaptive algorithm. Moreover, using nonlinear control theory, the proposed approach to neural networks provides sufficient conditions for determining the number of hidden layers, the dimension of neurons, the architecture of the neural network and the weights among the layers in order to guarantee stability of the system. The theoretical results are illustrated by application to a two-link manipulator.<>