A tabu search algorithm for the vehicle routing problem

Abstract The purpose of this study is to develop a new tabu seach heuristic to solve the single-depot vehicle routing problem of a distribution company carrying goods from a depot to a set of dedicated dealers. The heuristic proposes a new neighbourhood generation procedure which considers the scattering pattern in the locations of the dealers. A set of neighbours are generated by first executing a procedure which ignores the clustering of the vertices and then executing a second procedure which takes into account the relative locations of the dealers; then the feasible non-tabu move among these with the best objective function value is implemented. During neighbourhood construction, a combination of traditional improvement techniques are used simultaneously in order to achieve the exchange of vertices. The intensification efforts, on the other hand, are focused upon the hill-climbing behaviour of the search, and the diversification step which is standard in all previous tabu search approaches is not included in this heuristic. Numerical results on well-known benchmark problems indicate that the performance of the algorithm developed in this study is compatible with the other best-known algorithms in the literature. Scope and purpose This paper deals with the design of a heuristic algorithm to solve the vehicle routing problem (VRP) for a well-known distribution company in Turkey, which transports electronic household commodities from various plants to a large number of dealers. The logistics system between the distribution company and the parent company which owns all the plants is operated on a two-echelon framework; the manufactured commodities are first transported from the plant to the main depot in close proximity of each plant by the plant management and then from the depot to the dealers by the distribution company. Thus, it is just necessary for the distribution company to schedule the delivery from the depots to the dealers. Since each plant manufactures a different commodity and they are located far away from each other, the company management has decided to plan each depot independently. The main issue is to design a fast and robust procedure to solve a single-depot VRP for each plant. Since the VRP is a hard, combinatorial problem, and only relatively small instances can be solved to optimality, an approximate approach is prefered to manage the large number of dealers assigned to each depot. Specifically a tabu search approach known to provide good solutions in the VRP context is employed to obtain a practical solution. The algorithm developed in this study is then tested upon the benchmark problems in literature and shown to provide competitive results. The company management was satisfied with theoretical performance and accepted to use it in their daily operation. The early findings in practice also indicate considerable improvements in operational performance.