Its really only a matter of it being an adaptive control system design problem, I say only because the major stumbling block with adaptive control is to develop a tractable mathmatical model of the system that you are trying to control, with neural networks the adaptive algorim is effective in that it works but the details are hidden so from a point of view of an effective system a neural network would be best but if you want to understand HOW it works then you need to go the classical adaptive route.Unfortuantly adaptive control systems involve systems of simultanious non-linear integrals and the maths becomes wildly complex very quickly.
mind you the maths behind neural networks beomes wildly complex very quickly as well but at a much more deeply mysterious life the universe and everything level ;-)