Hareket halindeki mülteciler için mobil yardım sağlanmasında lojistik eniyileme

As one of the harshest humanitarian problems in the world, displacement has caused more than 79.5 million people to migrate as of June 2020 according to UNHCR. In this thesis, the aim is to provide an efficient, dynamic, and effective solution approach for the optimization of a logistics problem related to the mobile aid provision for en route refugees during migration. The underlying optimization problem is defined on a time-space network concerning a multi-period planning horizon. Transiting refugee groups move from a node to an adjacent node in each period according to pre-fixed paths on the network. To be served, a refugee group must be at the same node with at least one mobile service facility in the same period. The problem is to find the number of mobile facilities that need to be utilized, in addition to their locations and the routes over time, while providing continual service opportunity with predetermined frequencies. The objective is to minimize fixed and variable transportation costs. We define and study this problem for the first time and prove its NP-hardness. We develop an exact mixed-integer linear programming (MILP) model. However, due to the model's impracticality in larger instances, a faster approach is needed, and hence we develop a problem-specific Adaptive Large Neighborhood Search (ALNS) algorithm. The performances of both solution methodologies are compared using test data that we generated from the 2018 Honduras Migration Crisis. For small-sized instances, both ALNS and MILP find optimal solutions at a similar pace. In medium-sized instances, ALNS finds optimal solutions in shorter durations, whereas the MILP model is not able to find optimal solutions in all instances in a limited time. In large-sized instances, the better performance of the ALNS algorithm is observed, as it finds better solutions in shorter times compared to the MILP model. We further analyze properties of the proposed solution methods and obtained some insights through computational tests.