Incremental Scheduling: Using Scheduler with Multiple Problem Sub-models

In many real scheduling applications, the scheduling model is not static. There may be an initial problem to solve and then a series of changes such as resource breakdowns or the arrival of new orders to schedule that must be incorporated. With such an application, especially one of reasonable size, it is unrealistic to re-solve the entire problem from scratch. Such a re-solve may take too long, especially if the schedule is in the process of being executed, and the new solution may have little in common with the original solution.

Similarity between consecutive solutions is important when a schedule has external impact. For example, if the original schedule was used by an inventory manager to arrange the delivery of raw materials, radically changing it can have costly repercussions through-out a company's supply chain.

The goal of this chapter is to demonstrate how Scheduler can be applied to a situation where jobs arrive incrementally and the desire is to incorporate the new jobs into the existing scheduling, while satisfying the twin objectives of achieving high schedule-quality and minimizing changes to the existing schedule. To do this we make significant use of the modeling capabilities of Concert Technology, maintaining multiple models of parts of the overall scheduling problem. The communication among these models forms a significant part of the solution to our problem, and therefore this chapter serves as an example of how to solve scheduling problems by combining the facilities of the Scheduler Engine with the modeling capabilities of IBM® ILOG® Concert Technology.