Extreme Control Systems Based on a Combination of Search Optimization Algorithms

Abstract

The authors consider approaches to development of the extreme control systems (ECS) based on a simultaneous combination of different algorithms of search for a fitness function optimum. Such systems require a method for integration of the results of different algorithms. As a distinctive feature an algorithm can have high performance under certain conditions, developed methods of configuration, effective means of estimation of the future, current and past system states. Sometimes a system solution for a control problem was no good. This was a result of an excessive simplification of the search algorithm, e.g. carried out in order to facilitate the proceeding system design and performance of the structural transformations of the control system loops. In the authors' opinion, the approach based on integration of different algorithms, first of all, is aimed at preservation of high quality extreme control of a system. Weighted summation of the separate control signals is the traditional way to integrate the algorithms into the control systems. This method is a "compromise" between the controls proposed by different algorithms. However, in this paper the authors consider another way available for the extreme control systems. They examine ways of integration of the search algorithms as exemplified by the systems, which include not only the well-known algorithms with developed methods of design and algorithms of high efficiency, but also a complex configuration in terms of specific tasks. The authors demonstrate that in certain cases the methods of integration under consideration allow us to achieve a higher quality of the search processes than each algorithm taken separately. They show possible methods to organize configuration for such systems. They suggest high quality performance based on separate search algorithms, at the first stage, and a further "manual" parameter of configuration.