Adaptive control of genetic parameters for dynamic combinatorial problems

Abstract

The idea of using diversity to guide evolutionary algorithms is gaining interest. However, it is mainly used in static problems or in dynamic continuous optimization problems. In this paper, we investigate the idea on dynamic combinatorial problems. The paper uses a measure for population diversity based on distance from the population-best individual rather than distance between all possible pairs in the population. The measured diversity is used to adjust the mutation rate and the selection probability in a standard genetic algorithm whenever the diversity is found to be excessively low or excessively high. This adaptive scheme aims to retain the algorithm ability to search the solution space even after the population converges prematurely around some suboptimal solution. This scheme also enables the algorithm to persevere after converging around solutions that become obsolete due to environmental changes. Tests on several benchmarks of dynamic travelling salesman problem show that the scheme is promising. © 2007 by Springer Science+Business Media, LLC.