Summary: | 碩士 === 中華大學 === 運輸科技與物流管理學系碩士班 === 100 === In recent years, urban public library systems have always transported and delivered library materials within their branch systems. Consequently, the materials flow and the demand for delivery services is increasing at rapid rate within their branch system. Such an increase makes the higher delivered cost of library and the longer waiting time of readers. Therefore, this research discuss the library materials distribution problem of urban public library systems which according to Apte and Mason’s (2006) strategy for generating alternate designs of the delivery operations to redesign a new constructing routes idea: (1) the situation at have cross docking, the forward routes and return routes are different, (2) then confirms this idea’s feasibility according to the San Francisco Public library's actual case material.
In this research, a special type of Vehicle Routing Problem, named as the Vehicle Routing Problem for Library System (VRP-LS), is proposed and a simple heuristic method is designed to solve the VRP-LS. This heuristic includes three steps: (1) selecting a specific branch library as the cross-docking, (2) constructing routes to serve branch between main library and the cross-docking branch, (3) improving routes by neighborhood search heuristics. In the first step, we use the Selecting Feeder-base (SF5) method proposed by Chang (2006); in the second step, we utilize the Nearest Neighbor (NN) method to construct the initial routes; in the final step, we used two neighborhood search methods: Or-Opt and Inter-route node exchange, and add Threshold Accepting (TA) to reduce the distance of routes.
In order to verify the feasibility and suitability of the VRP-LS model, this research utilizes the same SFPL instances that are from Apte and Mason (2006). Then, we used the C# computer language to coding the above heuristic method and use the SFPL instances to test and analysis its performance. Through comparing with the SFPL current delivery routes and Apte and Mason’s alternative 2, we found that VRP-LS improves the average cycle time and decrease the total travel distance. This implies that the proposed VRP-LS model is feasible and efficient.
|