We present a generic neural network architecture capable of con(cid:173) trolling non-linear plants. The network is composed of dynamic. Using a recur(cid:173) rent form of the back-propagation algorithm, control is achieved by optimizing the control gains and task-adapted switch parame(cid:173) ters. A mean quadratic cost function computed across a nominal plant trajectory is minimized along with performance constraint penalties. The approach is demonstrated for a control task con(cid:173) sisting of landing a commercial aircraft in difficult wind conditions.

A recurrent back-propagation neural network architecture was then designed to numerically estimate the parameters of an optimal nonlinear control law for landing the aircraft. The performance of the network was then evaluated.

A recurrent back-propagation neural network architecture was then designed to numerically estimate the parameters of an optimal nonlinear control law for landing the aircraft. The performance of the network was then evaluated.

A recurrent back-propagation neural network architecture was then designed to numerically estimate the parameters of an optimal nonlinear control law for landing theaircraft. The performance of the network was then evaluated.