Mixed Resolution Modeling Issues for the Battlespace InfoSphere

"IET proposes to develop a decision theoretic design for Mixed-Resolution Modeling. The key concept is to treat a parameter produced by one model/simulation as evidence for the actual value of a parameter required by a second model/simulation. IET willinvestigate the feasibility of constructing an influence diagram that suggests an appropriate value for a parameter given information about the models/simulations. To preserve stochastic fidelity and support real-time and abstracted simulations, IETproposes a random variable representation for parameters. To facilitate parameter matching among applications running at different levels of resolution and how much evidential weight to give a parameter, IET proposes to design a knowledge representationfor metadata associated with a parameter. The metadata will be incorporated into the constructed influence diagram and will include knowledge about the parameter's validity and the context under which it was derived. To support reuse of parametertransformation knowledge, IET proposes to investigate the utility of a knowledge base of Bayesian network fragments that perform parameter credibility assessments, parameter conversions and parameter selection. From such fragments, one could automaticallyconstruct a parameter-specific influence diagram that infers a value/distribution for the parameter. IET will investigate relevant commercial and DoD standards and requirements of the JBI.The market need