Google+ Dynamic Agent Based Modeling using Bayesian Framework for Addressing Intelligent Adaptive Nuclear Nonproliferation Analysis | Educating the Next Generation of Leaders in Nuclear Security Sciences
Skip navigation


R. Elmore, "Dynamic Agent Based Modeling using Bayesian Framework for Addressing Intelligent Adaptive Nuclear Nonproliferation Analysis," Ph.D. Dissertation, Nuclear Engineering, Texas A&M University, College Station, TX (2014).


Realistically, no two nuclear proliferating or defensive entities are exactly identical; Agent Based Modeling (ABM) is a computational methodology addressing the uniqueness of those facilitating or preventing nuclear proliferation. The modular Bayesian ABM Nonproliferation Enterprise (BANE) tool has been developed at Texas A&M University for nuclear nonproliferation analysis. Entities engaged in nuclear proliferation cover a range of activities and fall within proliferating, defensive, and neutral agent classes.

In BANE proliferating agents pursue nuclear weapons, or at least a latent nuclear weapons capability. Defensive nonproliferation agents seek to uncover, hinder, reverse, or dismantle any proliferation networks they discover. The vast majority of agents are neutral agents, of which only a small subset can significantly enable proliferation. BANE facilitates intelligent agent actions by employing entropy and mutual information for proliferation pathway determinations. Factors including technical success, resource expenditures, and detection probabilities are assessed by agents seeking optimal proliferation postures.

Coupling ABM with Bayesian analysis is powerful from an omniscience limitation perspective. Bayesian analysis supports linking crucial knowledge and technology requirements into relationship networks for each proliferation category. With a Bayesian network, gaining information on proliferator actions in one category informs defensive agents where to expend limited counter-proliferation impeding capabilities. Correlating incomplete evidence for pattern recognition in BANE using Bayesian inference draws upon technical supply side proliferation linkages grounded in physics. Potential or current proliferator security, economic trajectory, or other factors modify demand drivers for undertaking proliferation. Using Bayesian inference the coupled demand and supply proliferation drivers are connected to create feedback interactions.

Verification and some validation for BANE is performed using scenarios and historical case studies. Restrictive export controls, swings in global soft power affinity, and past proliferation program assessments for entities ranging from the Soviet Union to Iraq demonstrates BANE's flexibility and applicability. As a newly developed tool, BANE has room for future contributions from computer science, engineering, and social scientists. Through BANE the framework exists for detailed nonproliferation expansion into broader weapons of mass effect analysis; since, nuclear proliferation is but one option for addressing international security concerns.

See Document

Associated Project(s):

  • Proliferation Resistance Methods Development

  • View Sitemap