PID controller optimisation using genetic algorithms

Abstract

[Abstract]: Genetic Algorithms are a series of steps for solving an optimisation problem using genetics as the model (Chambers, 1995). More specifically, Genetic Algorithms use the concept of Natural Selection – or survival of the fittest – to help guide the selection of candidate solutions. This project is a software design-and-code project with the aim being to use MATLAB® to develop a software application to optimise a Proportional-Integral-Derivative (PID) Controller using a purpose built Genetic Algorithm as the basis of the optimisation routine. The project then aims to extend the program and interface the Genetic Algorithm optimisation routine with an existing rotary-wing control model using MATLAB®. A systems approach to software development will be used as the overall framework to guide the software development process consisting of the five main phases of Analysis, Design, Development, Test and Evaluation. The project was only partially successful. The Genetic Algorithm did produce reasonably optimal values for the PID parameters; however, the processing time required was prohibitively long. Additionally, the project was unsuccessful in interfacing the optimised controller to the existing rotary-wing model due difficulty in conversion between SIMULINK® and MATLAB® formats. Further work to apply code optimisation techniques could see significant reduction in processing times allowing more iterations of the program to execute thereby achieving more accurate results. Thus the project results suggest that the use of Genetic Algorithms as an optimisation method is best suited to complex systems where classical optimisation methods are impractical.