本研究以0-1整數規劃建模，並以先分群後規劃路徑策略分解原來問題。這樣的策略形成兩階段的求解過程：在第一階段以掃描法分配需求點至各鄰近車輛，第二階段則以最佳化軟體求解各車輛之路徑規劃。藉由多種分群組合的方式反覆找出較佳的解答。 Fukang bus is a door-to-door transportation service system in Taiwan established for the handicapped. Such type of service is also known as dial-a-ride transportation system. Though the scale of the Fukang bus becomes larger and larger, the performance of the system is not satisfactory due to the low utilization of vehicles. The purpose of this study is to improve the dispatch and routing of vehicles to enhance their utilization. This study focuses on the dynamic requests of service and hence makes the route planning a dynamic dial-a-ride problem, which also increases the difficulty in solving the problem. This study uses 0-1 integer programming to model the dynamic dial-a-ride problem, and adopts the cluster-first-routing-second strategy to solve the problem. Such a strategy divides the solution procedure to two stages. In the first stage, the sweep algorithm is employed to allocate demand nodes to their neighboring vehicles, and thus forming a set of single vehicle routing problems. In the second stage, an optimization solver is used to solve the route planning of each vehicle. By performing the sweep algorithm repeatedly with different stating nodes, we can find different combinations of demand clusters and then obtain different routing plans. A best solution is finally found from all combinations.