Abstract
The use of simulation as a computational tool arises in many real-time decision-making settings. Such problems rarely lend themselves to steady-state formulations. As a consequence, the computations typically involve “transient” calculations and the answers therefore depend heavily on how one chooses to initialize the simulation. Typically, the initialization will need to be aligned with the current “state” of the real-world system under consideration. However, in most settings, the state variables underlying the simulation will differ significantly from the real-world state that is observable. In this talk, we will examine this data assimilation issue, discuss some of the statistical and computational challenges, and introduce several relevant algorithms that arise in this context.