Research
My research sits at the intersection of multi-stage stochastic optimization, distributionally robust optimization, and decomposition algorithms. I formulate optimization models for the sequential decisions that plan and operate infrastructure where the uncertainty a system faces can itself depend on the decisions made along the way. I build decomposition algorithms that solve these models exactly at the scale of a real network, and turn the solutions into policies an operator can run. The goal throughout is infrastructure that is resilient and adaptive by design.
- Methods: multi-stage stochastic optimization, distributionally robust optimization, and decomposition algorithms
- Modeling: sequential decisions with decision-dependent uncertainty
- Applications: resilient and adaptive infrastructure systems, including power networks, inspection, and supply chains
The projects below develop these ideas across infrastructure and logistics networks.
Dynamic investment planning and operation of power systems under wildfire-prone conditions
Multi-stage stochastic optimization models support long-term investment planning and adaptive grid operation under wildfire risk. The models represent uncertainty that evolves with decisions such as line switching, and the resulting solutions become policies for resilient response.
Static and dynamic infrastructure system monitoring under uncertain time-dependent failures
Decision-dependent failure uncertainty shapes which infrastructure components should be inspected and when. I develop decomposition algorithms for dynamic inspection routing and translate the resulting solutions into adaptive monitoring policies.
Risk-averse stochastic optimization for logistic network design and reconfiguration
Distributionally robust and risk-averse optimization models design and reconfigure logistics networks under demand and lead-time disruptions. Decomposition methods make these sequential decisions tractable at realistic network scale.