Adaptive neural controller based on convex parametrization

Abstract

The problem of control of a class of nonlinear plants has been addressed by using neural networks together with sliding mode control to lead to global boundedness. We revisit this problem in this thesis and suggest a specific class of neural networks that employ convex activation functions. By using the algorithms that have been proposed previously for adaptive control in the presence of convex/concave parameterization for adjusting the weights of the neural network, it is shown that global boundedness of all signals can be achieved together with a better tracking error than non-adaptive controllers. It is also shown through simulation studies of an aircraft landing problem that the proposed adaptive controller can lead to better learning of the underlying nonlinearity.