Teaching Genetic Algorithms With A Graphical User Interface

• Published in: Association for Engineering Education - Engineering Library Division Papers p. 1.417.1
• Main Authors: Haupt, Randy L, Toussaint, Gregory J, Pack, Daniel J
• Format: Conference Proceeding
• Language: English
• Published: Atlanta American Society for Engineering Education-ASEE 23.06.1996
• Subjects: Antenna arrays; Artificial neural networks; Biological evolution; Biological models (mathematics); College students; Computer simulation; Crossovers; Evolutionary algorithms; Fuzzy control; Fuzzy logic; Fuzzy systems; Genetic algorithms; Graduate studies; Graphical user interface; Matlab; Mutation; Neural networks; Optimization; Parameters; Pendulums; Principles; Problem solving; Software; Students; Teaching; User interface

Over the past several years, genetic algorithms have emerged as a powerful tool for solving optimization problems in engineering. Genetic algorithms model biological evolution on the computer using the principles of natural selection, mating and mutation. Although the subject has been predominantly studied at the graduate level, undergraduate students can easily master the concepts. We have developed MATLAB-based software with a graphical user interface to teach the fundamentals of genetic algorithms. The program allows a user to adjust several different parameter values associated with genetic algorithms including the optimization function, the population size, the crossover rate, and the mutation rate. A user can graphically monitor how these parameters affect the evolutionary path the algorithm takes to find an optimal solution. This approach of teaching students to experiment with genetic algorithms increases their level of understanding and allows them to quickly grasp the essential properties of the method. We will demonstrate how users can execute the genetic algorithm software to solve some sample engineering optimization problems. 1 Introduction Recently, many researchers around the globe have given a considerable amount of attention to a number of methods that use non-traditional system models to solve complex problems. Among these methods, neural networks, fuzzy logic, and genetic algorithms are most prominent. All three methods simulate the dynamics of a system without relying on rigorous mathematical models. To those who are not familiar with the new methods, such approaches may seem inadequate without mathematical rigor, but the techniques are superior to conventional methods in a variety of advanced problems: inverted pendulum control problem using 1 2 neural networks , Sendai train control using fuzzy logic control , and antenna array pattern optimization 3 problem using genetic algorithms . In most cases, however, these new methods have been taught and studied in the graduate schools and have not been easily accessible to wider audiences. To help undergraduate students learn one of these new-proven techniques early in their academic careers, we developed simple and effective graphical user interface software for genetic algorithms. We started this work because we could not easily find a good educational tool to present genetic algorithms to undergraduate students. Our MATLAB-based software allows students to easily explore how genetic algorithms use the principles of natural selection to solve optimization problems. The software allows students to control most of the algorithm’s parameters and observe how they influence the problem solving process. The paper is organized in the following manner. First we start off with a brief description of genetic algorithms. Section three presents the graphical user interface software followed by two sample problem 1996 ASEE Annual Conference Proceedings