線性軸輻路網接駁轉運最適整合模式之研究

Abstract

本研究探討線性軸輻路網之接駁╱轉運課題，首先利用遺傳演算法之編解碼技術，分別構建同時分組、逐步分組及種子分組三模式，俾進行路線分組，以決定路線間之接駁轉運關係，並與層級分組模式比較；其次，構建路網最佳化及排班最佳化模式，求解各路線分組之最佳路網型式及班距。目標函數係求解乘客候車成本（含起站、接駁站及轉運站）、乘客懲罰成本（含轉乘、接送）、業者成本（含營運、車隊），以及場站設置營運成本（含接駁站及轉運站）之總和最小化。 因路線分組模式本質上屬NP-hard問題，為瞭解各模式之尋優效度與效率，本研究以亂數產生之200個樣本進行測試，結果顯示種子分組之尋優效度最佳，但效率最差；逐步分組模式之尋優效度次佳，效率亦次佳。同時分組模式因染色體較長，尋優效度最差。以台汽公司60條國道直達路線為實例應用，發現種子分組尋優效果明顯優於其他模式，經接駁轉運設計後，形成1組直達集合（32條路線）、3組接駁集合（6條路線）及7組轉運集合（22條路線），總成本降低5.18%，顯示接駁轉運確能改善現行之直達服務方式。由敏感度分析知，車輛容量與乘客單位轉乘懲罰成本係影響接駁轉運成功與否之關鍵。因此，若能增加車輛容量，以提高接駁╱轉運可能性，並妥為設計接駁站及轉運站之轉乘方式，以減少乘客之不便，將更能加強接駁╱轉運系統之功效。

This study discusses the development and application of an integrated feeder/transfer system model in a linear hub-and-spoke network. At first, genetic algorithms are employed to determine the feeder/transfer relationship among routes by clustering the routes into different partition sets. Three models- simultaneous clustering method (SICM), stepwise clustering method (STCM) and cluster seed points method (CSPM)- basing on different coding/decoding techniques are developed and compared with a conventional statistics clustering model- the agglomerative hierarchical clustering method. Then, an algorithm integrating feeder/transfer location selecting, routing, and scheduling is proposed to solve the optimal network and headway for each partition set. The objective function is to minimize the total sum of passenger waiting cost (at origin ends, feeder stops and transfer stations), passenger inconvenience penalty cost (including change vehicles, pick-up and delivery), operator cost (including operation cost and fleet cost), set-up cost (including feeder stops and transfer stations). Because of the NP-hard character of the routes clustering problem, it is essential to analyze the effectiveness and efficiency of these methods. A testing experiment of 200 randomly generated objects is designed to perform the comparison. The testing result shows that CSPM is the most effective but least efficient method, STCM is second most effective and efficient, SICM is least effective because of its long chromosome. A case study of 60 direct service routes operated by Taiwan Motor Transport Company shows an optimal condition can be reached by clustering these 60 direct-service routes into three feeder sets with 6 routes, seven transfer sets with 22 routes and one direct service with 32 routes. As a result, the total cost is lowered by 5.18%, illustrating that feeder/transfer systems can surely improve a direct service system. The sensitivity analysis indicates that bus seat capacity and change vehicle penalty cost are two key factors in designing a feeder/transfer system. It implies that expanding the bus seat capacity to increase the possibility of feeder/transfer passengers boarding and well designing the transfer arrangement to reduce the inconvenience in changing vehicles would further enhance the performance of an inter-city bus feeder/transfer system.