Aspects of evolutionary construction of new flexible PID-fuzzy controller

Abstract

In this paper a new approach for designing control systems is presented. It is based on ensemble of PID controller and flexible neurofuzzy system with dynamic structure. A hybrid population-based algorithm is proposed to select the structure and its parameters. In this hybridization a genetic algorithm is used to select the controller structure and evolutionary strategy is used to simultaneously select the controller parameters. The proposed approach allows design interpretable control systems based on different control criteria and different controlled object. The proposed controller structure and proposed learning algorithm were tested on typical control problem.