Join us in welcoming Juan Borrero, an assistant professor from the Industrial Engineering and Management Department at Oklahoma State University, as he shares his research.
A Robust Optimization Approach to Enhance Community Resilience under Tornado Hazards
Catastrophic tornadoes cause severe damage and threaten the overall well-being of many communities across the US. Recent research has shown that the resilience of existing structures to tornadoes can be enhanced by implementing simple retrofitting strategies, such as improving the cover and nailing patterns of roofs. Whereas at the federal level retrofitting is now seen as a valid tool to prepare for disasters, agencies at the local level rarely consider retrofitting in their emergency preparedness plans. In this talk, we consider a decision-maker (a local emergency management agency or a public-private consortium) who seeks to allocate resources in retrofitting and recovery strategies to minimize the population dislocation in an urban area due to an uncertain tornado. As tornado paths cannot be forecast reliably, we model the problem using two-stage robust optimization: retrofitting decisions are made in the first stage, and recovery decisions are made in the second stage after observing the tornado. We assume that tornado paths can be represented as line segments and that a location is affected by the tornado if it is sufficiently close to the line segment, which results in a mixed-integer non-linear uncertainty set. To solve the problem, we use a decomposition column-and-constraint generation algorithm that requires solving a challenging two-level optimization subproblem at each iteration. As this subproblem cannot be tackled by standard ‘dualize and combine’ techniques, we design a decomposition branch and cut (DBC) algorithm that adds two types of constraints on the fly. Particularly, the initial constraints of the master relaxation and the feasibility check of the DBC are constructed by exploiting the geometric properties of the uncertainty set. Numerical results are reported using real data from Joplin, MO, which show that there can be reductions of up to 20% in worst-case population dislocation by investing $15 million in retrofitting and recovery. The results also show that our approach can outperform random retrofitting policies by margins close to 20% and that the model does not suffer from over-conservativeness. Moreover, we show that the population dislocation exhibits a considerable ‘diminishing returns’ behavior with respect to budget: beyond $15 million, the reduction in population dislocation is not significant.
Juan S. Borrero is an Assistant Professor in the School of Industrial Engineering and Management at Oklahoma State University. He holds a B.Sc. in Mathematics and an M.Sc. in Industrial Engineering, both from the University of Los Andes, Bogota, Colombia, and holds a Ph.D. in Industrial Engineering from the University of Pittsburgh (2017). His research interests are mainly in the area of optimization under uncertainty. Methodologically, his focus is on bilevel, robust, network and mixed-integer optimization, probability, and stochastic processes. From his doctorate onwards, he has focused on sequential hierarchical decision-making problems with uncertainty and learning, including applications such as smuggling interdiction, defender-attacker problems, and commit or defer problems. More recently, he has worked in applications such as preparedness and response against tornado hazards, UAV routing for intelligence, surveillance, and reconnaissance missions, supply operations under uncertain rippled attacks, and influence problems in social networks, among others. His research has been funded by ONR, AFOSR, and by an NSF CAREER award.
https://ncsu.zoom.us/j/91749207591?pwd=UnhkSldaQ290RG0rYitSVVZZK0lDUT09
Meeting ID: 917 4920 7591
Passcode: 303411
