M&S Uncertainty Quantification

A goal of Uncertainty Quantification (UQ) is to use computer simulation of complex systems to make scientifically informed assessments for high-consequence decisions. Because end-to-end empirical data is difficult to obtain for the Ballistic Missile Defense System (BMDS), computer simulation provides the best method for understanding BMDS capabilities against a wide range of threats. Numerica Corporation, in partnership with the Georgia Tech Research Institute, proposes to develop an adaptive, sampling-based framework for quantifying the impact of both aleatoric and epistemic uncertainties within the BMDS. The primary goal of this effort is to construct a simulator-agnostic UQ capability that is applicable to a broad range of modeling and simulation environments. Nonetheless, the Phase I effort would also implement a proof-of-concept prototype in a specific environment, to demonstrate the proposed technology on a problem of interest to MDA. Key innovations in the proposed work include: (i) novel sampling methods for efficiently exploring high-dimensional uncertainty spaces with dependent variables, (ii) non-parametric techniques for quantifying the errors due to finite sample size, (iii) feedback mechanisms to reduce the total number of samples needed for UQ, and (iv) flexible techniques for accelerating simulations through computation re-use.