Hybrid models of solving optimization tasks on the basis of integrating evolutionary design and multiagent technologies

Abstract

The paper is devoted to the problem of building hybrid intelligent systems for solving multi-objective optimization problems. The authors present the definition of a hybrid system, and the main problems and tasks of its development. The main idea is that integration of methods of computational intelligence and multiagent systems (MAS) can be promising and useful for developing intelligent systems. The paper describes the concepts of designing agents, multi-agent systems, and the design process with elements of self-organization (interaction, crossing, adaptation to the environment, etc.). The authors propose a method of forming child agents as a result of the interaction of parent agents, develop various types of crossover operators, and present the idea of creating agencies (families) as units of the MAS evolving. To implement the proposed ideas, hybrid fuzzy-evolutionary models of forming agents and agencies based on the use of fuzzy coding principles are created and described in the paper. The authors developed a software system to support evolutionary design of agents and multi-agent systems for estimating the effectiveness of the hybrid approach. The results demonstrate the effectiveness of the proposed approach.