Exploring the Job - Shop Search Space with Genetic Algorithms

In this thesis, I designed and implemented a new approach of applying Genetic Algorithms to the Jop-Shop Scheduling Problem. In my implementation, I created a population mechanism which uses genocide to maintain solution diversity, and an elite population of superior solutions to improve search time and overall results. My chromosome representation assigns priorities to each operation to be scheduled, relying on forcing techniques to generate feasible schedules. I also introduced stepped genetic operators, so that the GA can control how much genetic information each parent contributions to the generation of a child. Finally, my implementation was tested using well known benchmarks, and the results indicate my approach is as successful as the best known GA methods in the literature. Thesis Supervisor: Patrick H. Wiston Title: Ford Professor of Engineering