SBIR-STTR Award

As the use of complex simulation representations becomes more widespread, it is imperative that their accuracy be precisely defined. At the same time, however, the very nature of their increased complexity progressively prohibits the use of conventional methods of measuring or assuring such precision. The statistical validation methods proposed by VisiTech would complement and augment the traditional verification and validation methods. Those methods would still be required at the more fundamental level such as the individual production rule for knowledge-based models or the constituent model used in complex simulations. The statistical methods developed under this SBIR would be used to measure performance when these more fundamental constructs are combined in ways that result in non-deterministic aspects during execution.

Keywords:
Simulation Validation Performance Function Statistics

The performance of simulations, especially distributed simulations, can be very dependent on certain critical operational factors such as processor loading, available network bandwidth, or aspects of the scenario being represented. To date, there is no method of characterizing the expected performance of a simulation as these conditions vary. The purpose of this effort is to develop a system that will automatically generate such characterizations. Once a simulation has been characterized, the system can also be used to provide confidence limits on the validity of the representations provided under the specific operating conditions of any simulation run.

Keywords:
Simulation Validation Confidence Limits Modeling Error Limits