The Centre of Excellence in Defence, Risk and Resilience (CEDRR) is inviting you to attend a guest lecture by data scientist Ms Carolyn Chen on "Applying sequential decision modelling to defence investment planning".
The lecture will take place via Teams on May 13th 2026 from 3PM to 4PM.
Admission is free by registering for the event here.
Abstract: Defence planning in a national defence organization is a complex undertaking, involving the prioritization of hundreds of competing projects and the stewardship of billions of dollars in public funds. While decision makers routinely rely on optimization techniques such as integer programming and genetic algorithms to support investment choices, these approaches often overlook two critical realities: defence investments are made sequentially and under significant uncertainty. To address these gaps, this work develops a sequential decision model of the Canadian Department of National Defence’s multi‑gate Project Approval Process. The model is formulated using the Decision Advantage with Reinforcement Learning framework developed by the NATO SAS‑181 Research Task Group. It explicitly captures the dynamic nature of project selection, including the arrival of new projects over time, uncertainty in future budgets, and cost variability associated with delayed or previously approved projects. The application of the model is demonstrated through a comparative analysis of three project selection policies: a ranked‑list approach, a traditional knapsack optimization, and an enhanced knapsack strategy that reserves a contingency fund for future opportunities. Together, these results illustrate how sequential decision modelling can provide a more realistic and robust foundation for defence investment planning.
Biography: Ms. Chen is a defence scientist at Defence Research and Development Canada’s Centre for Operational Research and Analysis. She specializes in optimization and sequential decision processes, and has worked on projects supporting capital investment planning across the Canadian Department of National Defence. She is an active member of the NATO SAS-181 Research Task Group “Exploiting Reinforcement Learning To Achieve Decision Advantage” which has been investigating how reinforcement learning and approximate dynamic programming may be best applied to support defence and security decision making.
NB: The Teams link for attending will be sent to the registred users before the event and will also be visible on the event page 5h prior to the event.
For further information, please contact : cedrr@port@ac.uk