Fluid Policies, Reoptimization, and Performance Guarantees in Dynamic Resource Allocation

In many applications, decision makers must repeatedly allocate limited resources across randomly varying subsystems. This structure arises in a wide variety of applications, such as network revenue management, online advertising, assortment planning, marketing, managing energy systems, and multilocation inventory management, to name only a few. These problems are challenging, as the constituent subsystems may vary over many states. In this paper, we develop easy-to-compute decision rules for such problems and provide a rigorous theoretical analysis of these methods, showing that they often perform remarkably well. We illustrate these methods on examples from multiwarehouse inventory control, and these numerical examples underscore our theory.