A Comparison of Optimization Techniques for Integrated Manufacturing Planning and Scheduling

摘要

We describe a comparison between Simulated Annealing (SA) Dispatch Rules (DR), and a Coevolutionary Distributed Genetic Algorithm (DGA) solving a random sample of integrated planning and scheduling (IPS) problems. We found that for a wide range of optimization criteria the DGA consistently outperformed SA and DR. The DGA finds 8-9 unique high quality solutions per run, whereas the other techniques find only one. On average, each DGA solution is 10-15% better than SA solutions and 30-35% better than DR solutions.