Discrete Controller Synthesis Based On Genetic Programming

In this paper, an alternative way of a discrete controller synthesis is introduced. The synthesis is based on a proposed genetic programming algorithm, which is named Block Diagram Oriented Genetic Programming BDOGP. The standard GP structure tree is modified in such a way to obtain a complete block diagram of the discrete controller that satisfies a deadbeat response in a closed-loop system. In one framework solution, the algorithm gives both the block diagram topology and the values to the parameters within the controller structure. A new numeric constant mutation operation is added to the algorithm to strengthen the search for optimal parameters of the BDOGP solutions. Two examples are introduced to validate the use of the proposed algorithm, and for the sake of completeness, the state-space approach design of a deadbeat response is introduced briefly. For a servo system, a comparison between the results of the GP and the conventional statespace solutions shows the accuracy of the GP approach. The second example considers a temperature control in an HVAC system. Key-Words: Genetic Programming, Block Diagram Oriented Genetic Programming, Deadbeat Controller