Algoritmos genéticos e de evolução diferencial aplicados à otimização de controladores utilizando a norma H2 como critério de desempenho

Abstract

The synthesis of controllers using the H2 norm as performance criterion is usually solved from conditions based on linear matrix inequalities or by solving Riccat equations. However, such approaches or non-convex design criteria are incorporated into the project. In this sense, this work proposes to analyze the efficiency of heuristics by the application of evolutionary algorithms in the synthesis of controls using H2 norm as a performance criterion, in order to simplify theirs solutions. The controllers obtained by genetic and differential evolution algorithms are simulated computationally by MATLAB/Simulink® software, and also tested experimentally. It is also necessary to evaluate the closed-loop physical implementations of this systems and to analyze if there is a need add different constraints to ensure the implementation of the obtained controller. To do so, the control system are applied to a torsional system available in the LACOS laboratory of UTFPR Cornélio Procópio. It is important to emphasize that the implemented algorithms can be easily adapted to any system, if it is know the mathematical model of this one. The algorithms are efficient tools for the synthesis of optimized controllers.