Identification of Simulation Semantic Descriptors

For simulations to become ubiquitous, and to maximize utility, they must be capable of providing precisely the results desired - and the results must be valid in the context of the application. To meet these requirements in an effective and economic manner implies extensive reuse of models and the ability to either compose or tailor simulations to meet specific needs. To accomplish this with a high degree of confidence, and to ensure that the simulation is valid in the context of the application, an understanding of the model semantics and their semantic interdependencies is required.

This initiative is the first systematic investigation into the semantic definition of models and simulations. Identification of these semantics will provide a structure for the definition of the level of compatibility between models, will assist in verification and validation, especially with respect to federations, and will lay the groundwork for composable simulations.

This research has the following specific goals:

	Develop an initial set of semantic descriptors for models and simulations
	Establish an infrastructure that will be used to test the adequacy of these descriptors
	Identify procedures for associating the values of semantic descriptors with the range of applicability for the models or simulations

The initial set of semantic descriptors will be a prototype set sufficiently large to test the adequacy of the descriptors to capture the semantic interdependencies of a selected set of models. This set will demonstrate the feasibility of identifying semantic descriptors and provide an initial demonstration of their value.

In order to test the semantic interdependencies between models, an infrastructure in which to conduct such tests must be created. This infrastructure must be flexible to easily accommodate any set of models to be tested. Additionally, it must provide for a high degree of automation to make statistically based sensitivity analysis practical. The infrastructure that will be developed as a part of this effort provides those qualities through its combination of sensible agent and statistical validation technologies (defined later). This infrastructure will be well suited to support not only the future testing of additional semantic descriptors, but also for investigations into questions of composability and reuse.

Semantic descriptors can also be used to identify the range of applicability of models and combinations of models. Using the prototype set of defined descriptors, this research will then identify procedures that can be employed to identify limitations on the reliable applicability of models based upon the values of those descriptors. This research will therefore develop a template by which appropriate semantic descriptors are identified and validated, and their implications to model employment identified.