Solving multi-depot vehicle routing problems via hybrid algorithms

Vehicle Routing Problem (VRP) is one of the topics that has been working on nearly for 60 years. The vehicle routing problem was first introduced in 1959 by Dantzig and Ramser as an academic study (Dantzig and Ramser, 1959). VRP, in its simplest form, is the routing problem which aims to minimize the costs by taking arising constraints into account to provide full services to n assigned customers. The Vehicle Routing Problem is a NP-hard combinatorial optimization problem that is quite important for firms when planning route plans in distribution channels and it has an impact on logistics costs to a significant extent. There are exact and heuristic solution methods that are recommended for solving this problem. The algorithms that give the most effective results for VRP are metaheuristic algorithms. In the literature there are numerous researches have examined VRP and they continue to study on it. These studies often take a single depot into account. Between the VRP types, Multi-Depot Vehicle Routing Problem considers the most realistic situations than other types (Crevier, Cordeau and Laporte, 2007). Some companies may be able to serve from multiple depots in different locations instead of serving from a single depot to customers in different locations. In this case, these companies deal with MDVRP. Chronologically applying to metaheuristic methods for solving routing problems has been increased. In this thesis, Genetic Algorithm, Artificial Neural Networks and K-Means clustering algorithm is used as Hybrid Algorithms to solve multi-depot vehicle routing. There are two structures that complement each other's deficiencies in this method: GA and ANN. Genetic Algorithm is effective in global search and ANN works better in local search. NeuroGenetic method aims to combine two methods to find higher performance results. This method has been applied to known benchmark to solve multi-depot vehicle routing problem. Keywords: Vehicle routing problem, multi-depot vehicle routing problem, metaheuristic methods, hybrid algorithms.