Artificial neural networks for automatic control.

摘要

Artificial neural networks are massively parallel networks of simple computational elements which collectively have the ability to learn and to generalize. This paper presents an application of artificial neural networks to aircraft flight control. A full authority neural network is trained using supervised learning via the back-propagation algorithm to provide stabilization of the airframe and to shape the desired dynamic behavior (handling qualities) of a linear aircraft model. The final system performance is demonstrated through simulations and linear analysis of the resulting aircraft and neural controller. These results demonstrate the neural controller's excellent closed-loop performance.