Issues in the complementary use of simulation and optimization modeling

Cahiers Leibnitz n°211Simulation models and discrete optimization models are oftentimes used together in a variety of ways. In this paper, we discuss the issues that modelers must address in cases where simulation models are used to test a discrete mathematical programming optimization model's performance in a stochastic environment. The issues arise during validation of simulation models, when checking agreement between deterministic optimization results and simulation models operating under deterministic conditions. In our case, the issues are derived from validating simulation models that are used to test the performance of scheduling and resource allocation models (integer and mixed-integer programming optimization models) under various types of uncertainty. While the concerns we describe are from our work in the logistics domain (cross-docking operations), they are relevant to a wide variety of problem domains. In addition to describing the issues, we offer suggestions on how modelers might address the concerns