The Evolutionary Growth of Neural Networks for the Autonomous Adaptive Control System

摘要

The paper clarifies that adaptability of both natural and artificial control systems (CS) is reached through several stages of prehistory, syntheses and existence of the system. The CS ability to adapt in the real-time of control is of great importance. Such ability resides only in a complex CS, consisting of a few special subsystems. The article presents the CS of such sort, called "Autonomous Adaptive Control" system (AAC). At the same time because of the complexity of the AAC system and the set of its parameters, their preliminary optimization is required. Yet such optimization is impossible to execute analytically. The AAC system optimization by the means of genetic algorithms (GA) is proposed. The example of such optimization of the CS for the satellite angular motion stabilization is presented.