Pilot Model Using Neural Networks

Abstract

Motion of aircraft controlled by the pilot is decided depending on the characteristics of man-machine system. Although analysis and study using the mathematical model of the aircraft including the control system is usually done, the method for making mathematical model of pilot, which is necessary for analysis and study of man-machine system, has not been established. Although there was an example of the method for making mathematical model of pilot using transfer function (Kato, A., et al., J. Jpn. Soc. Aeronaut. Space Sci., 50 (2002), pp. 409–415), it was thought that there was a possibility that a more accurate, more flexible pilot model was obtained if neural networks (NN) was applied. Therefore, various studies were done concerning pilot model to which NN is applied. As a result, it has been understood that pilot model with better performance is obtained comparing with the case of applying the transfer function. Moreover, it has been understood also that versatile pilot model, which can deal with various conditions by one model, is obtained by applying NN and studying the control results in various conditions.