Natural disasters and extreme events, such as hurricanes that are characterized by their violence and unpredictability, often result in severe damage as well as many fatalities. One way to assist the vulnerable population and decision-makers during extreme event evacuation is to provide both operational and strategic evacuation guidelines that give specific information on individual route selection, evacuation timing and shelter destination assignment. To this end, in this dissertation, we study three interrelated problems in evacuation planning. In the first study, we consider a strategic evacuation network design problem, that mainly determines open potential shelter locations, transfer nodes, and road segments under uncertainty in the number of people evacuating from their origins. We incorporate uncertainty considerations in a chance-constrained two-stage mean-risk stochastic programming framework where uncertainty is represented by a discrete scenario set in the formulation. In our formulation, the edge capacity constraints are modeled via joint chance constraints. These constraints ensure the feasibility of the routing problem and also penalize the violations in the objective function. In addition, we consider the risk-averse approach to model the evacuation network design problem. The conditional value-at-risk (CVaR) is taken into account as the risk measure as well as the expected total cost in the objective function. In our modeling approach, we control the unfulfilled demand by employing a quantitative approach via incorporation of shortage cost into the mean-risk objective function and a qualitative approach via joint chance constraints. To solve large scale instances, we devise an efficient Benders Decomposition (BD) algorithm with both a multi-cut and single-cut approach for our proposed model. Our numerical results show that the multi-cut approach solves our problem in a more reasonable time compared with the traditional single-cut approach. We apply our model and solution approach on a case study from Central Texas and observe the effects of changing risk parameters on the optimal solution and the trade-off between the expected total cost and risk measure CVaR. Second, in a deterministic setting, we consider an evacuation network design problem under cost and travel congestion for pre-disaster decision making in the case of an imminent hurricane. We employ a time-expanded network to incorporate the worst-case evacuation time for the evacuees. Specifically, we propose a mathematical model that prescribes evacuee routes through the road network to shelter locations by making design decisions on open shelter locations and capacities, road segments utilized in evacuation, and their capacities by making contraflow decisions. To solve our model, we devise an efficient BD framework with convergence enhancements for solving large-scale instances. We design and implement an experimental study to test our BD technique using data from Central Texas. In the last part, we examine our model under conditions of uncertain evacuee choice and speed. That is, evacuees who would choose their shelter locations based on their priorities and might not follow the optimal solution of the optimization model. Additionally, evacuees can speed up or slow down during the evacuation process based on traffic density. Therefore, we create a simulation model to test the outcomes of the strategic and operational models outlined above by also considering operational situations that cannot be handled by an optimization model. Our simulation model illustrates that the optimal solution of the optimization model is obtained based on a trade-off between total cost and total evacuation time.
Dr. Halit Uster
Creative Commons License
This work is licensed under a Creative Commons Attribution-Noncommercial 4.0 License
Mansouri, Naderehsadat, "COST- AND TIME- EFFECTIVE REGIONAL EVACUATION NETWORK DESIGN AND OPERATION UNDER TRAVEL CONGESTION AND UNCERTAINTY" (2019). Engineering Management, Information, and Systems Research Theses and Dissertations. 10.
Available for download on Monday, November 20, 2023