Online Scheduling: Understanding the Impact of Uncertainty

Abstract

We present a framework to study quality of schedules obtained iteratively and online (real-time) in the presence of demand uncertainty. Using this framework, we carry out a computational study, and make interesting observations. First, we find that uncertainty plays a less important role as a manufacturing facility is operated close to capacity. Second, the choice of the horizon for the online iterations, is dependent on the mean load, but independent of the accuracy of the forecasts. Finally, feedback, in the form of re-optimization, plays a very important role in mitigating the impact of uncertainty. Thus, through the analysis presented in this work, we gain insights that are applicable to all general rescheduling approaches.