公車站牌穩健時刻表之設計

近年來各縣市政府積極推動公車智慧型站牌建置，提供乘客掌握公車最新位置及預估到站資訊，進而提升公共運輸使用率及旅客服務滿意度，然而考量到智慧站牌的設置成本及電力供應，於國內所有站牌建置智慧型站牌的難度較高。因此，本研究將設計可靠的站牌時刻表供偏遠地區公車站牌使用，用以取代智慧型站牌建置之方案。本研究依據公車到離站歷史資訊，期望發展一套站牌時刻表模型。此模型將透過控制站牌點的方式，當車輛提早抵達控制站牌點時，需停等至時刻表表定離站時間方可離站；若車輛晚到控制站牌點時，乘客等至車輛抵達站牌方可上車。依據車輛早到及晚到機率，將產生車輛怠速營運成本、車上乘客等候成本及車外乘客等待成本。本研究將提出三種控制站牌點設計情境，情境一公車可停等於所有站牌、情境二公車僅可停等於所有控制站牌點及情境三公車僅可停等於五個控制站牌點。研究選取國內各一條短路線及長路線，分別利用歷史平均法及基因演算法兩種計算方式比較總時間成本，其結果顯示利用基因演算法停等於所有站牌之總成本最小，且此方法應用於長路線的表現效果較佳，故在公車路線旅行時間不確定性下，可適用此最佳站牌時刻表模型。

關鍵字

公車即時到離站；站牌時刻表；基因演算法

並列摘要

Recently governments have engaged in enhancing the service quality in bus systems especially the construction of the real-time bus arrival time display at the bus stops so that passengers can be well-informed of bus arrival time. However, due to the limited resource and the unavailability of power supply, the display may not be able to be widely installed. Therefore, the reliable bus stop timetable in remote areas may be a good alternative, suggesting the importance to determine a reliable and robust timetable for each of bus stops. To do so, the study aims to develop an optimal timetabling model for bus stops based on the historical bus arrival data from Advanced Public Transportation Systems so as to provide the trustworthy bus timetable. The core logic of the proposed model is to determine a reasonable and workable timetable when buses arrive the stop ahead of schedule have to wait at stop till the time comes so that no passengers will miss the buses as they arrive the stop on time. However, if the buses arrive the stop behind the schedule, passengers have to wait for the buses. Based on this, the proposed model is balancing the waiting time cost of at-stop passengers and on-bus passengers as well as the additional delay cost of buses. Genetic algorithms are then used to solve the optimal timetable. Meanwhile, considering the restrictions of reality, buses may not be able to wait at some stops which have no space for bus short-term parking. Three scenarios are compared. The first scenario assumes that all stops along the bus route can be control stops (i.e., buses can wait at stops without interrupting traffic). The second scenario assumes that buses can only wait at control stops. The last scenario assumes a given number of control stops and the optimal location of these control stops along the route are compared. To show the applicability of the proposed model, two case studies on a short and a long bus routes are performed, respectively. Compared with the total cost of timetabling schemes of three travel time quartiles, the total cost of the GAs-based timetable is significantly lower. Additionally, the GAs-based timetable performs even better for the long bus route than the short bus route, suggesting the proposed model can successfully determine the optimal bus-stop timetable towards the total cost minimization under travel time uncertainty.