Consortium for Enabling Technologies and Innovation 2

Thrust Area 1

Data Science and Digital Technologies in Nuclear Security and Nonproliferation

Highlights of Thrust Area 1

The nuclear nonproliferation community today is greatly benefitting from the advances in data science of the past seven decades, especially in the domain of machine learning. However, most of the applications have focused on ‘narrow-AI’—highly specialized problems, often with hard-coded expert knowledge but lacking in sufficient input data. There is a need for a broad range of models, signatures, and methods to not only be able to discover signs of early or ongoing proliferation activities, but to have predictive capabilities of indicators of foreign nuclear program activities. This is vital for real-time and prognostic situational awareness in multi-modal dynamic domains.

TA1 projects present opportunities and identify challenges in data science for nonproliferation. On the opportunity side, we as a community are already using machine learning for a few tasks best suited for computers—data classification, visual learning, anomaly detection and swarm control to name a few. There are several data science challenges that ETI 2.0 will address. The ETI 2.0 team will coordinate projects from the proliferation detection domain subarea with the testbeds and digital twins cross-cutting areas. The TA1 scope addresses the challenges applying AI-assisted technologies in nuclear nonproliferation including synthesis and applications of AI in data analysis, machine learning, digital twins, and computer vision/augmented/virtual reality. Given the critical mission of DNN R&D, we aim to address the issues of robust AI in nonproliferation from every angle, including use of various sensors, networks, testbeds, and digital twins.

Thrust Area Lead: Pavel Tsvetkov, Texas A&M University

Universities: GT, TAMU, U. Wisconsin, VCU

National Laboratories: ANL, LLNL, LANL, ORNL, PNNL

Projects:

  • Project 1 – Robust AI data collection and processing for proliferation detection domains
  • Project 2 – AI-assisted signature discovery for high-fidelity detection of anomalies
  • Project 3 – AI-assisted automatic risk-based detection of cyber-physical threats
  • Project 4 – Synthesis of augmented reality (AR) and AI with digital twin technologies