The NASA Langley UQ Challenge on Optimization Under Uncertainty


NASA missions often involve the development of new vehicles and systems that must be designed to operate in harsh domains with a wide array of operating conditions.  These missions involve high-consequence and safety-critical systems for which quantitative data is either very sparse or prohibitively expensive to collect. Limited heritage data may exist, but is also usually sparse and may not be directly applicable to the system of interest, making uncertainty quantification extremely challenging.

The NASA Langley Research Center has developed a UQ challenge problem in an effort to focus a community of researchers towards common goals. The challenge features key issues in UQ using a discipline-independent framework. While the formulation is indeed discipline-independent, the underlying application features characteristics present in nearly all real-world missions.

To meet the goals of this challenge, NASA is seeking responses to this challenge problem to address the following:


Register for the 2019 UQ Challenge

Each participating group must first register. To register please send us an email using the “Contact Us” link. The email must contain the names and email addresses of your group members.


2019 UQ Challenge Document


Accepted responses to the challenge will be part of a dedicated session of the ESREL 2020 conference to be held in Venice, Italy, from November 1-6, 2020.


Update (October 8, 2020): The results of this challenge will be considered for publication in a special edition of the Mechanical Systems and Signal Processing journal.


Questions and Answers (updated as new questions arrive)

Q1: When ranking the epistemic parameters in B.1, what is meant by ‘improving the predictive ability of the computational model’? Should this ranking determine which parameters most strongly affect the model output (without taking into account the observations) or which parameters should be reduced to most closely match the data?

A1: Both to some extent. Each epistemic variable is currently prescribed as an interval. If you reduce the width of the interval of any of such variables, the spread of the predicted output will be consequently reduced. We want to identify the epistemic variables leading to the largest reduction in the output’s spread. Note however that we do not know the extent by which the interval limits will change in B.2. 

Q2: For the second problem (B.2), can you elaborate on what exactly will be provided in the refined UMs after defining the uncertainty reductions to the set E?

A2: We will give you intervals with a smaller width that are fully that contained in the original intervals, e.g., the original interval [0,2] for any chosen variable will be reduced to [a,b] such that a>=0, b<=2, b-a<2