Modeling and optimization of a multi-echelon multi-objective humanitarian logistics network

Abstract

Humanitarian logistics management has become a very important issue now a day due to increase in different types of disasters in recent years all around the world. When disaster strikes proper management of relief network is essential as any indulgence might be met with severe penalty like loss of valuable human lives. Another important issue is the uncertain nature of disasters as we don’t know where disaster will occur or at what magnitude. In case of a disastrous event, decision making usually involves selection of proper locations for aid distribution centers, distribution of relief goods to different demand nodes, proper vehicle management for transportation of relief goods etc. In this thesis a multi echelon multi objective logistics model will be developed which will attempt to find optimal locations for setting up regional relief distribution centers with minimum setup cost and then it will find the optimal quantity of goods flow among different nodes and stages of the relief supply chain to minimize the relevant distribution cost and while doing that will minimize both the unmet demand and amount of unitized relief goods in presence of demand and supply uncertainty. Hence the model under consideration spans from opening of local distribution facilities, to initial allocation of supplies, to last mile distribution of aid, i.e. all three echelons of humanitarian logistics network. The model will also seek to optimize the number trips that are needed to be made by different types of vehicles for efficient distribution of relief goods among different nodes of the network. At the end a case study will be incorporated in which an example problem of the model will be solved and optimized using metahuristic algorithms to demonstrate the models effectiveness. Two suitable algorithms: Firefly Algorithm (FA) and Bat Algorithm (BA) have been used for this purpose. Results obtained from both algorithms shows Pareto optimality, but Bat Algorithm seemed to be capable of solving the model in shorter period of time than Firefly Algorithm for the same number of iterations.