Integrated flight/propulsion control system design via neural network

Neural network design methodology is utilized to design an integrated airframe/engine control system for a modern, statically unstable, fighter aircraft in landing. The resulting neural controller structure consists of a feedback compensator and a feedforward filter formulated in the form of a three-layer feedforward network whose parameters are trained by a static backpropagation method. The number of parameters is chosen by an ad hoc procedure. The feedback compensator satisfies closed-loop stability, while the feedforward filter provides command shaping. Once training has been completed, and the parameters are fixed, overall system performance evaluation results are then presented.<
>