Research Information System
Thesis Type: Doctorate
Institution Of The Thesis: Gazi Üniversitesi, Fen Bilimleri Enstitüsü, Turkey
Approval Date: 2017
Student: BURAK AYDOĞDU
Supervisor: BAHAR ÖZYÖRÜK
Abstract:Vehicle Routing Problem (VRP), an important component of Supply Chain Management (SCM), is at the forefront of the operational processes that companies need to improve in an increasingly competitive environment. The cost criterion is one of the most important elements in competition of companies with each other. Companies may encounter additional costs due to the collection of recycled / returnable products due to legal and environmental obligations besides of customers' delivery costs. In real life, it will be a great advantage for the companies in terms of cost to satisfy the requirements to pickup of recyclable / returnable products by vehicles in circulation. In this study, Dynamic Vehicle Routing Problem with Simultaneous Pickup and Delivery (DVRPSPD) is dealt with to satisfy companies' pickup and delivery requirements with minimum cost. Within the scope of the thesis, the periodic re-optimization process is preferred from the periodic and continuous re-optimization process as the re optimization process for DVRPSPD. This process is divided into two phases: the first phase (static problem) and the second phase (dynamic problem) in which re-routing is performed. In the first phase, a mathematical model from literature is used for Vehicle Routing Problem with Simultaneous Pickup and Delivery (VRPSPD). For the second phase, a mathematical model with binary variables for DVRPSPD is developed. The effectiveness of the developed mathematical model was evaluated using the problems in the literature. Since DVRPSPD belongs to the class of NP-hard problem, it is difficult to find the optimal solution in a reasonable time (7200 sec.) for medium and large size instances. For this reason, Simulated Annealing (SA), Local Search (LS) together with Simulated Annealing Tabu Search (SA-TS) and Random İterative Local Search Variable Neighborhood Descend (R-ILS-VNS) algorithms in which hyper-heuristics are used developed to solve the problem. The performance evaluation of the heuristics algorithms are performed using the instances in the literature for the first phase. Results show that SA-TS algorithm gives better results than those of other heuristic algorithms. Finally, the results obtained by the solution of DVRPSPD with the SA-TS algorithm have been evaluated.