| This research investigates the construction of routes for local delivery of packages. The primary objective of this research is to provide realistic models to optimize vehicle dispatching with random customer locations and demand, while maintaining driver's familiarity with their service territory. Driver familiarity results from visiting service areas repeatedly. With increasing familiarity with their service territory, the driver's performance will increase due to increased ease in finding addresses, and increased efficiency of organizing routes. At the same time, it's also very important to have flexibility to re-optimize dispatch plans by adjusting the number of vehicle routes according to changing customer locations and demand.; The objective of increasing driver familiarity tends to make routes or service territories fixed. On the other hand, the objective of increasing flexibility to optimize the number of routes tends to assign vehicles/drivers to variable routes or service territories. To balance the tradeoffs of these two objectives, the proposed research develops concepts of “cell”, “core area” and “flex-zone” for large-scale vehicle dispatching problems, and explicitly considers the value of driver familiarity as well as dispatching flexibility. Further, a two-stage vehicle routing model is developed: Strategic Core Area Design model and Operational Cell Routing model.; The system testing results have shown that the core area method performs favorably compared with the variable routes method. On average, the core area method cuts the driver utilization by about 5% and the total duration about 3%, and it significantly increase the dispatching consistency, by about 32%. The values of the core areas and flex-zones are further confirmed by the fact that the core area method is only 3% (number of drivers, total duration and distance combined) above the lower bound of total routing cost. |
