Evolutionary Based Controller Design

title = "Evolutionary Based Controller Design",

author = "Ivan Sekaj",

booktitle = "Evolutionary Computation",

publisher = "InTech",

year = "2009",

editor = "Wellington Pinheiro dos Santos",

chapter = "13",

month = oct,

keywords = "genetic algorithms, genetic programming",

isbn13 = "978-953-307-008-7",

URL = "

http://www.intechopen.com/download/pdf/pdfs_id/10935",

DOI = "

DOI:10.5772/9620",

bibsource = "OAI-PMH server at www.intechopen.com",

language = "eng",

oai = "oai:intechopen.com:10935",

URL = "

http://www.intechopen.com/articles/show/title/evolutionary-based-controller-design",

abstract = "In this chapter evolutionary based design/optimisation approaches has been proposed for controller design of continuous-time process control. Parameters of controllers with fixed defined internal structure are designed as well as controllers with a-priori unknown internal structure and its parameters. The presented approaches minimise a cost function, which comprises closed-loop system simulation and performance index evaluation. In this way the controller design is transformed into a search problem in the n-dimensional parameter space. The design/optimisation can be carried out for complex systems and control structures of various types. The main and practically the only limitation of the approach is the time consuming computation (compared with conventional approaches) due to thousands up to ten thousands closed-loop simulations needed by each design procedure. From the point of view the user, on the other hand, the design method is simple to use. It transfers the design effort from the experienced human designer to the computer. The design approach is simple to extend to robust controller design, for which two different methods have been proposed. In addition statistical robustness measure has been introduced, which can be considered as an objective tool for robust controller performance comparison. Next, the design idea has been extended also to a multi-objective design task, where the objective is the search for the Pareto-optimal set of solutions. From these solutions the designer can choose the representative, which is the most appropriate in the particular case. Finally the design goal was extended from a fix defined controller structure with unknown controller parameters to the search/optimisation of the unknown internal structure of the controller. From that reason, the Genetic Programming has been used. The proposed evolutionary-based methods can be used for design of various types of controllers for various system types (linear, non-linear, stable, unstable, SISO and MIMO, fuzzy, neural, etc.). The only condition of this approach is that there exists a simulation model of the designed closed-loop. In the future, this design approach will be extended for solving very complex design tasks in the process control area like complex MIMO control systems for non-linear continuous-time systems and for robotic applications using parallel evolutionary algorithms.",

size = "22 pages",