Multiobjective Optimization for Controller Design
Author(s): HUANG Liang, WANG Ning, ZHAO Jin-Hui
Pages: 472-477
Year: 2008 Issue: 4
Journal: ACTA AUTOMATICA SINICA
Keyword: Controller; tissue P systems (TPS); multiobject optimization; Pareto optimality; evolutionary algorithms;
Abstract: Controller design using the multiobjective opti-mization method is cousidered, in which the objectives and con-stralnts for designing controllers are analyzed and improved.In order to satisfy the objectives synchronously, a new multi-objective optimization algorithm is introduced into designing an optimal PID controller, inspired by tissue P systems. The controller's parameters are coded and evolved according to the rules of the associated membrane. The unique design is that the whole population of the structure is divided into several subpop-nlations to decrease the computational complexity. Simulation results show that the algorithm converges fast and the solutions form a precise front and distribute uniformly. The controllers designed by the variant of P systems have satisfactory perfor-manos. Moreover, the analysis of solutions shows that the new algorithm is suitable for studying the relationship between per-formance and tuning parameters. The proposed method is useful for designing and evaluating different controllers.
Pages: 472-477
Year: 2008 Issue: 4
Journal: ACTA AUTOMATICA SINICA
Keyword: Controller; tissue P systems (TPS); multiobject optimization; Pareto optimality; evolutionary algorithms;
Abstract: Controller design using the multiobjective opti-mization method is cousidered, in which the objectives and con-stralnts for designing controllers are analyzed and improved.In order to satisfy the objectives synchronously, a new multi-objective optimization algorithm is introduced into designing an optimal PID controller, inspired by tissue P systems. The controller's parameters are coded and evolved according to the rules of the associated membrane. The unique design is that the whole population of the structure is divided into several subpop-nlations to decrease the computational complexity. Simulation results show that the algorithm converges fast and the solutions form a precise front and distribute uniformly. The controllers designed by the variant of P systems have satisfactory perfor-manos. Moreover, the analysis of solutions shows that the new algorithm is suitable for studying the relationship between per-formance and tuning parameters. The proposed method is useful for designing and evaluating different controllers.
Related Articles