Genetic Algorithms Working in Dynamic Environments

Abstract

<p>Genetic Algorithms (GAs) are search methods based on principles of natural selection and genetics. GAs attempt to find good solutions to the problem at hand by manipulating a population of candidate solutions.</p> <p>Each member of the population is typically represented by a single chromosome, the chromosome encodes a solution to the problem, the initial population is generated randomly, GAs are often used as optimizers, and the fitness of an individual is typically the value of the objective function at the point represented by the chromosome. The individuals with better performance are selected as parents of the next generation. GAs create new individuals using simple randomized operators that resemble crossover and mutation in natural organisms. The new solutions are evaluated with the fitness function, and the cycle of selection, recombination, and mutation is repeated until a user defined termination criterion is satisfied.</p> <p>In the real world, we always encounter the problems that need to be solved in a changing environment. This means that our algorithm needs to be dynamic or even adaptive to the changing environment.</p> <p>In this thesis, we will mainly deal with the adaptive GAs that have a new genetic operator called transformation instead of traditional crossover.</p> <p>In our study, we use a dynamic problem generator to create a dynamically changing landscape and study the behavior of transformation based GA in different parameter settings, such as: transformation rate, mutation rate, segment replacement rate.</p>