NEUROAVTOPILOT OF THE LONGITUDINAL CHANNEL OF THE AIRCRAFT

Abstract

The use of Classical Control Theory leads to necessity to consider linearized models, to use linear laws of control, the simplified data that reduces the practical importance of the results received. Accordingly, the problem of aircraft control systems creation is actual that uses nonlinear algorithms of control. After two decades of almost full oblivion, interest to artificial neural networks has quickly grown up for the last several years. Experts from such areas as technical design, Philosophy, Physiology and Psychology, are intrigued by the possibilities given by this technology, and are in search for its applications in directions of their interest. Such a revival of interest has been caused by both theoretical, and applied achievements. For today, the working prototypes of regulators of various types based on “purely” neural networks are developed. For research of dynamic characteristics of the aircraft in a longitudinal control channel, under influence of the autopilot on it, using programming language Java, the autopilot has been simulated that supplements the model of the aircraft. It was offered to include neuro-autopilot in a altitude control contour, that is the included neuro-autopilot covers the main functions on working out control signals on altitude stabilisation during transition to other altitude similar to developed autopilot (without speed control). To study the influence of neuro-autopilot on transient processes of model of the longitudinal channel of the aircraft, two situations are simulated: descent at the preset altitude, and ascent at the preset altitude, while making comparison of transient processes by altitude between autopilot control and neuro-autopilot control. Analyzing the transient processes graphs, we can make a conclusion that the airplane under neuro-autopilot not only reduces a static error, but essentially makes smoother the transient by altitude. More stable and smooth transient process improves the safety of air transportation. The conducted research shows that the approach to control of complex, nonlinear, dynamic systems in situations of uncertainty which is realised with the use of adaptation, gives a chance to a control system to adapt to change of a current situation, including the supernumerary ones.