Large-Scale Evacuation Network Model for Transporting Evacuees with Multiple Priorities
Abstract
There are increasing numbers of natural disasters occurring worldwide,
particularly in populated areas. Such events affect a large number of people causing
injuries and fatalities. With ever increasing damage being caused by large-scale natural
disasters, the need for appropriate evacuation strategies has grown dramatically.
Providing rapid medical treatment is of utmost importance in such circumstances. The
problem of transporting patients to medical facilities is a subject of research that has
been studied to some extent. One of the challenges is to find a strategy that can
maximize the number of survivors and minimize the total cost simultaneously under a
given set of resources and geographic constraints. However, some existing mathematical
programming methodologies cannot be applied effectively to such large-scale problems.
In this thesis, two mathematical optimization models are proposed for abating the
extensive damage and tragic impact by large-scale natural disasters. First of all, a
mathematical optimization model called Triage-Assignment-Transportation (TAT)
model is suggested in order to decide on the tactical routing assignment of several
classes of evacuation vehicles between staging areas and shelters in the nearby area. The
model takes into account the severity level of the evacuees, the evacuation vehicles’
capacities, and available resources of each shelter. TAT is a mixed-integer linear
programming (MILP) and minimum-cost flow problem. Comprehensive computational
experiments are performed to examine the applicability and extensibility of the TAT
model.
Secondly, a MILP model is addressed to solve the large-scale evacuation
network problem with multi-priorities evacuees, multiple vehicle types, and multiple
candidate shelters. An exact solution approach based on modified Benders’
decomposition is proposed for seeking relevant evacuation routes. A geographical
methodology for a more realistic initial parameter setting is developed by employing
spatial analysis techniques of a GIS. The objective is to minimize the total evacuation
cost and to maximize the number of survivors simultaneously. In the first stage, the
proposed model identifies the number and location of shelters and strategy to allocate
evacuation vehicles. The subproblem in the second stage determines initial stock and
distribution of medical resources. To validate the proposed model, the solutions are
compared with solutions derived from two solution approaches, linear programming
relaxation and branch-and-cut algorithm. Finally, results from comprehensive
computational experiments are examined to determine applicability and extensibility of
the proposed model.
The two evacuation models for large-scale natural disasters can offer insight to
decision makers about the number of staging areas, evacuation vehicles, and medical
resources that are required to complete a large-scale evacuation based on the estimated
number of evacuees. In addition, we believe that our proposed model can serve as the
centerpiece for a disaster evacuation assignment decision support system. This would
involve comprehensive collaboration with LSNDs evacuation management experts to
develop a system to satisfy their needs.
Subject
Large-scale optimization modelmodified Benders' decomposition
natural disaster evacuation plan
geographic information system
mixed integer linear programming
Citation
Na, Hyeong Suk (2014). Large-Scale Evacuation Network Model for Transporting Evacuees with Multiple Priorities. Master's thesis, Texas A & M University. Available electronically from https : / /hdl .handle .net /1969 .1 /152810.