Practical implementation of an extended VRP model
One of the most interesting problems in transportation networks is the Vehicle Routing Problem (VRP). Given a set of goods, a transportation network with delivery locations and a fleet of transportation vehicles the goal is to carry out deliveries from the central storage facility to all customers incurring minimal costs and subject to certain additional constraints. Computationally this is known to be a hard problem. However, with a computational technology of nowadays, it is feasible to find “good enough” solutions in a reasonable time. The purpose of this paper is to present one such prototype tool, which is focused on the road transportation in Slovenia. Transportation networks are usually modeled by weighted graphs. With development of Geographic Information Systems (GIS) geographic information concerning transportation networks and other geographic information relevant for logistics became available for research and practical purposes. In Slovenia, GIS data on transportation networks are maintained in Cadastral register of the economic public infrastructure. During the efforts invested in optimization of processes of a leading Slovenian energy company, it turned out that the transportation costs represent an important component of a storage facility operation. In our case specific additional constraints have to be taken into consideration such as use of different types of vehicles, with different capacities (trucks and vans), different costs per kilometer and limitation of route duration to 8 hours due to regulations. As an additional constraint, some customers occasionally require delivery at a certain hour of day. With this constraint we have a well known and a well studied problem of VRP with time windows (VRPTW). The initial situation involved a central storage facility from which a fleet of more than 20 vehicles coming from several regions in Slovenia carried out deliveries on a daily basis. A dispatching in the storage facility was manual and the delivery locations were first distributed to route regions. We have developed a practical prototype software implementation which uses the data from the information system of the storage facility, GIS data of the transportation network, integrated together with an open source tool for graphical representation and our own implementation of special purpose optimization algorithms. For optimization we use heuristics based on local optimization upgraded with some meta-heuristics like tabu-search. The system is operational but as it is further tested in practice, we expect yet unforeseen constraints to emerge. The system represents a powerful tool for research in road transportation logistics in Slovenia. The emphasis is set on implementation of algorithms which have to be structured in such a way that incorporation of additional constraints in the future is possible. In our case additional constraints are included in the object function through penalization when particular constraint is not respected.