Framework for studying online production scheduling under endogenous uncertainty

We propose a framework for studying online production scheduling in the presence of endogenous uncertainties. We address uncertainties in (i) processing times; (ii) batch yields; and (iii) unit operating status. First, we illustrate how uncertainty can result in infeasibilities in the incumbent schedule and propose a model for systematic schedule adjustment to restore feasibility in the absence of new scheduling inputs. In this model, we define variables to track and penalize changes between the new and old schedule. Second, we discuss the different probability distributions for the three uncertainties that we consider in this work and how the parameters for these distributions change with sampling frequency. Third, we present a formal procedure for carrying out closed-loop simulations and evaluating closed-loop performance in the presence of these uncertainties. Finally, using this framework we draw useful insights for the design of online scheduling algorithms in the presence of the above three uncertainties.