Emergency Inventory Pre-Positioning For Disaster Response Operations

Disasters include not only natural disasters, but also man-made disasters. Once a disaster occurs, it will cause a lot of casualties and significant economic losses. In order to significantly reduce or even eliminate losses caused by the disaster, we need to focus on disaster relief operations. Since it is almost impossible to know the timing and magnitude of the disaster, it is very hard to estimate the impact, damage and resource needs exactly in advance; meanwhile, the route capacity is uncertain which means in some cases the whole transportation network is vulnerable and may be unoperational totally. In the context of this, a stochastic problem where randomness arises not only from demand but also transportation capacity in which case the decision process must take account into.This thesis firstly introduces the definition of deterministic model, stochastic optimization model, the "arc-road" algorithm of linear multi-commodity flow and so on, then the deterministic model and two-stage stochastic optimization model are shown based on the previous studies, both of which integrate facility location, inventory management and relief distribution together. We also provide a corresponding two-stage robust optimization model that takes into account of the facility location, inventory management as well as the relief distribution for disaster relief operations, and the robust model with a set of uncertainty defined by the L1-norm is equal to a solvable formulation. Next, an instance of the 2010 Yushu earthquake analysis is introduced to compare the two-stage robust optimization model with the two-stage stochastic optimization model and deterministic model respectively in the truck transportation network to demonstrate the advantage of the two-stage robust optimization model. Furthermore, we consider the situation of helicopter mode of transport in the distribution network which proves the superiority of the two-stage robust optimization model again. Finally, we conclude thoughts and propose the direction of future research.