Application of Coevolutionary Genetic Algorithm for Acquiring Scheduling Strategy

Abstract

In this study, a variation of coevolutionary genetic algorithm is proposed for acquiring the adaptive scheduling strategies in a complex multi-machine system. The task is to assign each machine appropriate dispatching rules that are harmonious with the rules used in neighboring machines. An ordinary coevolutionary process would not be successful due to the high variability of evaluation possible in this system and the ripple effects among neighboring populations. In the proposed method, an individual strategy in a population takes responsibility according to its role in the first-order change of the system performance. The total fitness of an individual is then evaluated by a statistically weighted sum of its zero-order performance and first-order performance. The performance of the attained adaptive strategies was superior to that attained by more time-consuming statistical search The proposed method can be utilized in a wide variety of GA application problems.