Show simple item record

dc.contributor.authorRawls, Carmen Gloria
dc.identifier.otherbibid: 6397137
dc.description.abstractExtreme events such as hurricanes and earthquakes can strike a community with little or no warning and leave high levels of devastation behind. Emergency response providers require large quantities of resource in the aftermath of such events, but these may be limited because of lack of preparation. In order to provide immediate assistance to disaster victims, essential supplies must be strategically placed before the event so they can be accessible after. The main goal of this research is to develop a large-scale emergency response planning tool that determines the location and quantities of emergency supplies together with the location and capacities of their storing facilities. A two-stage stochastic mixed integer program (SMIP) is presented that designs such an emergency response pre-positioning strategy for hurricanes or other natural disaster threats. The SMIP is a robust model that considers variability in forecasted demand and network unreliability. Due to the computational complexity of the model formulation, a heuristic solution that considers the embedded network structures of the SMIP was devised by combining two methodologies: the L-shaped method and the Lagrangian relaxation. The L-shaped method consists of solving an approximation of a stochastic program by estimating the recourse function using an outer-linearization technique. The Lagrangian relaxation heuristic was added to decompose the first stage problem into a trivial facility location problem and a resource allocation linear program. To further improve the computational capabilities of the algorithm, the Lagrangian relaxation was also used to relax the integrality constraints of the facility location variables. The result was a heuristic method referred to as the Lagrangian L-shaped method (LLSM). Various numerical experiments were conducted to test the computational capabilities of the LLSM. These experiments showed the computational consistency of the method compared to a standard integer program solver (i.e. Lingo). Regardless of variations in the data set provided, the running times of the LLSM are 0.05% to 10.0% of the Lingo running times, while the objective values obtained by the LLSM are within 1% of optimum. Based on the experiments, we are confident that the LLSM can be used as a large-scale resource pre-positioning planning tool.en_US
dc.subjectstochastic mixed integer programmingen_US
dc.subjectemergency responseen_US
dc.subjectresource prepositioningen_US
dc.subjectLagrangian L-shaped methoden_US
dc.typedissertation or thesisen_US

Files in this item


This item appears in the following Collection(s)

Show simple item record