Adaptive neural controller for a class of plant with nonlinear uncertainties

Abstract

This paper presents a neural-based adaptive control (NBAC) for the torque control of a flexible beam with structural uncertainties. In the NBAC, a neural network (NN) is connected in parallel with a linearized plant model, so that the NN is expected to identify the uncertainties included in the plant. At the same time the NN works as an adaptive controller that can compensate for the unknown system dynamics. At first, stability of the NBAC system including the nonlinear NN is analysed and a sufficient condition of the local asymptotical stability is derived by the Lyapunov stability technique. Then, the NBAC is applied to the torque control of a flexible beam that includes linear and nonlinear uncertainties caused by a contact force and by not exactly known shape and material of the beam. Experimental results illustrate effectiveness and applicability of the NBAC.