NEURO-CONTROLLERS FOR ADAPTIVE HELICOPTER HOVER TRAINING

This paper presents an application of artificial neural networks in ad aptive helicopter hover training of novice student pilots. The design of the adaptive trainer utilizes the hypothesis that novices can be tr ained to fly a helicopter system automatically (with no human interact ion) if the helicopter system adapts to the learning curve of the stud ent. Two different techniques based on the above approach are presente d. In the first technique, the helicopter system actively enforces opt imality by augmenting the novice's control inputs by amounts necessary to satisfy desired performance criteria. The second technique uses re laxed performance criteria that are not initially optimal, but approac h optimality in a graded fashion, based on the learning curve of the s tudent. Adaptive neuro-controllers, together with a critic model, are used to implement the adaptive helicopter system. The results using si mulated student models verify the approach adopted, and show that the adaptive neuro-controllers allow the helicopter system to adapt to the novice's learning curve.