Innovative Uncertainty Analysis Techniques for MDA Simulation Codes

Progress in missile defense relies heavily on large software packages such as BEST. BEST creates scene simulations for modeling sensor response in the battlespace environment. Sensitivity analysis of the package's predictions is essential due to uncertainties associated with the input parameters. However, performing such an analysis is hindered by the large number of parameters and the complex interactions between models. Automatic differentiation techniques potentially provide a means for overcoming these challenges. However, application of automatic differentiation to the large models that comprise BEST can be difficult due to their complex structure (in some cases derived from legacy codes). Other challenges are created by simultaneous software development at geographically distributed sites. Creare proposes to apply an innovative set of algorithms to BEST that harnesses the capabilities of modern methods for automatic differentiation to provide not only uncertainty analysis and parameter estimation capabilities, but also a distributed, object-oriented software development framework. The Creare approach to automatic differentiation-based uncertainty analysis offers a number of advantages in terms of software engineering, quality assurance, project management, and reduced code runtime.