Considerations for selective-fidelity simulation

Abstract

The concept of fidelity in a simulation model has become one of great contention among simulation researchers. While there is some agreement regarding the definition of fidelity in a model, there appears to be little accord about how the fidelity of a model might be measured - or even whether it is measurable at all. Because of the abstract nature of both reality and the representations of reality used for simulation systems, some argue that fidelity is also abstract in nature and can not be measured. Research conducted at IST, however, has yielded a method not only for measuring fidelity in a model, but also comparing the fidelity of different models to be used in a simulation. This ability to compare the fidelity of different models can be used in a practical manner to decide which models would be most appropriate for a new application. Further research at IST has explored the idea of using this ability to compare the fidelity of models to give a simulation system the ability to select the models that it should use based on outside factors such as the computational load on the system. This concept is called Selective-Fidelity Simulation, and is documented in this paper.