Scheduling algorithms using event graph constructs

摘要

We present a framework that allows users to easily develop scheduling algorithms that are both highly customized and powerful. We propose modeling constructs borrowed from the discrete event simulation field to produce high-fidelity shop floor models. We then propose a problem-independent heuristic search methodology (Problem Space Search) that can be applied "on top of' this model to find good schedules. We illustrate this idea using a real industrial problem that requires simultaneous lot sizing and scheduling, and involves alternative routes, sequence dependent setup times, and due dates. We argue that our framework overcomes shortcomings of currently available methods that can be too general, too costly, and/or too simple.