ABSTRACT
The path to fast and accurate computer simulation depends on the objective one pursues. In the case of offline applications, the goal is typically to simulate a highly complex system to study its behavior in detail and to ascertain its performance. An example might involve the design of an expensive large-scale system to be built, in which case it is essential to use simulation for the purpose of exploring whether it meets performance specifications and cost constraints under alternative designs and control mechanisms. In such a realm, reasonably fast simulation time is desirable but not indispensable, and one may be more likely to trade off execution speed for higher fidelity. On the contrary, for online applications, we often refer to real-time simulation as a tool for generating efficient models providing predictive capability while an actual system is operating as well as the means to explore alternatives that may be implemented in real time. An example might arise when a system is about to 390operate in an environment it was never originally designed for; one is then interested in simulating how the system would behave in such an environment, possibly under various implementable alternatives (e.g., adding resources to it or modifying the parametric settings of a controller). On the basis of the results, one might proceed in the new setting, abort, or choose the best possible modification before proceeding. In such cases, obtaining information fast from simulation is not just desirable, but an absolute requirement.
