ニューラルネットワークを用いたパイロットの着陸操縦の解析 第２報 ＧＡによるニューラルネットワーク構造の構築

Abstract

A Neural Network (NN) model of a human operation is tried to construct to analyze pilot behavior. The flight data obtained by simulator operation are utilized to teach the NN model. The input data to the model are visual cues from the cockpit and the pilot column deflection history, and the output is the elevator angle. The Genetic Algorithms (GA) optimizes the NN model in order to obtain the appropriate structure and its parameters. It can be seen that the sensitivity analysis and the threshold analysis of the obtained NN model reveal the human pilot behavior.