HYBRID MULTI-AGENT METHOD FOR OPTIMIZATION OF FUZZY COMPUTER SYSTEMS

Abstract

Recently, intelligent computer systems based on fuzzy logic and soft computing are used quite effectively to solve a wide range of complex applied problems in various fields of human activity. One of the promising areas of modern research in the field of artificial intelligence is the creation and testing of bioinspired multi-agent and evolutionary methods for the synthesis and optimization of fuzzy automatic control systems (ACS) and decision support systems (DSS). This paper is devoted to the development and research of a multi-agent method of parametric optimization of fuzzy systems (FS) based on hybrid improved grey wolf algorithms. The proposed method allows optimizing the parameters of fuzzy computer systems more efficiently, compared to the basic and improved grey wolf methods. This method uses group hunting and dimensional learning-based hunting strategies, as well as local search strategies based on algorithms of gradient descent and extended Kalman filter, which significantly reduces computational costs and increases the rate of convergence to optimal solutions when optimizing parameters of complex fuzzy systems. To study the effectiveness of the developed hybrid multi-agent method in this work, the synthesis and parametric optimization of the fuzzy flight control system for the unmanned aerial vehicle (UAV) is carried out. In particular, the optimization of the vectors of normalizing coefficients, adjustable parameters of linguistic terms (LT), and weight coefficients of the rule's consequences of the rule base (RB) for the altitude controller of the fuzzy ACS for the UAV is performed. The obtained results of the comparative analysis confirm the significant advantages of the developed hybrid multi-agent method of parametric optimization, as well as the feasibility of its application for the synthesis of different types of ACSs and DSSs based on fuzzy logic.