Genetic algorithms for scheduling: incorporation of user preferences

Abstract

Genetic algorithms (GAs) are optimization tools that simulate evolutionary processes in order to develop solutions for a wide range of optimization problems. In recent years, they have increasingly gained acceptance in industry, as confidence improves in their abilities to solve complex problems that have previously appeared to be unsolvable. One example of an industrial application for GAs is scheduling, an area that can present many difficulties for the manufacturing industry. Much of the advantage of GAs lies in the flexibility with which they may be implemented. In this work, two practically motivated improvements are made to the basic GA used for schedule optimization, to allow the technique to include additional complexities that arose in an industrial application. The first improvement enables the GA to deal with uncertain information in the factory, and illustrates the ability of the GA to aid scheduling decision-making. The second improvement is the application of the GA to multiobjective optimization, in the form of a multiobjective genetic algorithm (MOGA). In this way, the GA can solve problems with several conflicting or incompatible objectives, and allow the user to interact with the process as it evolves solutions. These variations permit the inclusion of specific user preferences, extending the scheduling choices available, whilst still ensuring that global optimization performance is not diminished. Thus the schedule optimization system becomes more interactive, accurate and effective for manufacturing schedule optimization.