Research On The Models Of Robust Bilevel Programming In Emergency Logistics Optimization

This dissertation goes deeply into studying the optimization of the emergencylogistics system and the connections between the optimization problems in the system,also analyses the relationships between decision variables in each subsystem. Based onthis foundation, the location-allocation problem (LAP), vehicle routing problem (VRP)and the location-routing problem (LRP) will be discussed further. At the meantime, theoperations of the emergency logistics might be influenced by the uncertain factors inthe process of dealing with the emergency. As the demand or the connectivity of theroad networks changes, the mismatch of emergency supplies between the supply anddemand and the mismatch between the route and transportations will happen As aresult, the operation and management of the emergency logistics system are damagedwhile leading to timeliness lag, moreover causing more severe disasters. At present,how to deal with the uncertainties impact of the uncertainty on the emergency logisticsprocess and establishing a robust emergency logistics system, have become a realisticproblem which needs to be solved urgently. Hence, based on the bilevel programming(BLP) and the robust optimization, the mathematical models of the robust bilevelprogramming are established to provide the decision-making for the LAP, VRP andLRP. By the transformation theorems of the uncertain BLP, these uncertain models areconverted to the certain bilevel multi-follower programming, and then thecorresponding hybrid genetic algorithms is designed. The main work and creations areas the following:Firstly, based on the decentralized decision-making, the definition and thetransformation theorems of the robust bilevel programming are proposed to handle thebilevel programming with no shared variables among followers and uncertainparameters, which provides a theory basis for converting the original bilevelprogramming problem with uncertain parameters into the deterministic bilevelprogramming, and then the hybrid genetic algorithm is proposed to solve thedeterministic model.Secondly, the collaborative model is established to determine the emergencystorage centers’locations and the material supply plan on the provincial and municipallevels by the vertical supply and lateral transfer based on the BLP with leader-followerhierarchical structure. Furthermore, considering the uncertain demand, a robust bilevelprogramming model with multistage rescue process is established to determine the two-grade material location and allocation of the province and cities, and thecollaborative model is established under the road networks uncertainty.Thirdly, the vehicle routing problem is an important problem in the emergencylogistics system. Under the uncertainty of the emergency material demand, we maychoose the different monitoring equipment after the network is destroyed in case ofemergency. If different monitoring equipment is adopted, the accuracy of the obtainingdata vary, so the accuracy of the network actual distance will affect the decisions of theemergency logistics system. While the cost of using the network monitoring equipmentis directly proportional to the accuracy of the network actual distance, so the accuracyof the network actual distance will affect the vehicle dispatching and further influencethe rescuing time. Based on the consideration of all factors, the robust bilevelprogramming is established based on the bilevel programming (BLP) and robustoptimization to balance the cost and the time of rescuing and the collaborative researchof the vehicle dispatching problem in the emergency material is discussed as well.Fourthly, the location problem and the routing problem are very important for theemergency logistics systems, which are interacted on each other. Emphasizing thepredominant role of the location problem, a basic robust bilevel programming modelwith uncertain demand is established for the location-routing problem to realize thecooperation of the location and routing problem. In the basic model, the upper levelprogramming aims to achieve the minimization of the total cost from the globalperspective, while the lower level pursues for minimizing the operating cost and thetransportation cost from the vehicle routing. Based on the model, this dissertation willfurther study the LRP and robust bilevel programming.