貨櫃運輸公司車輛排程問題之研究-模糊多目標遺傳演算法之應用

Abstract

貨櫃運輸公司負責廠商與貨櫃場間的貨櫃運送，其車輛排程以人工方 式依經驗與直覺做動態的即時調度。但運輸旅次起訖點分散各地，調度的 旅次數又多，造成調度人員重的工作負擔。 「資訊化」是目前企業追求效率、降低成本的趨勢。由於一般的貨櫃運輸 沒有時間限制，且運輸旅次事先已知，本研究建議只考慮距離問題，事先 以電腦做靜態的車輛排程工作，將電腦排程結果供調度人員參考。 由於貨櫃的整裝性質, 使每一旅次由一輛車執行一次, 又車輛由公司出發 執行數個運 輸旅次後回到原公司，且旅次間距離要最少，另為顧及司機 的工作量，而給予車輛的距離 限制，所以此問題具有車輛路線問題的特 性。 本研究將每一運輸旅次視為一點，則點與點間的距離即為運輸 旅次間的空車距離。因 旅次I之訖點至旅次J之起點與旅次J的訖點至旅次 I的起點距離不同，因此代表旅次的點與點之間的距離具有非對稱性。 調度員除欲減少行車里程外，尚欲均衡各車的行駛距離，本研究根據模糊 多目標規劃方法及自行修改的遺傳演算法做車輛排程的程式設計。 執行車輛排程程式，可在一分鐘內輸出排程結果，且結果較人工排程結果 為佳，此研究可供業者參考。 Container transportation companies serve the transporting of containers between firms and container-yards. The routing of the tractor-trailer are dispatched by dispatcher's experience and intuition. The scattered origins and destinations of the transport trips and the countless trips make the work of routing very difficult and burdsome. To purpose the high efficiency and reduce the cost, the companies tend to computerize. Because general container trips don't have time-window constraints and these trips are known beforehand, this study suggests only thinking of route length. The problem of the dispatching of vehicles is solved by computer, and the output of running program can be used as a reference to the dispatchers. Each trip is performed exactly once by one tractor-trailer. The vehicles depart form the company and return to the company after finishing performing trips. In this paper, the dispatcher wants to minimize the total distance of vehicles, and limit the distance of vehicles because of considering drivers' work load. This problem have the property of vehicle routing problem. In this paper each set of nodes represent each set of trips. The distance between nodes is considered as the distance of empty- carrying trips. These is a difference between the distance of the destination point of trip I to the origin point of trip J and the distance of the destination point of the trip J to the origin point of trip I, so the distance between two nodes is asymmetric. Another o bjective of this study is to balance the running distance of the tractor-trailer. In order to solve the combinatorial multiobjective problem, the fuzzy multiobjective programming and the altered genetic algorithm are employed. It takes less than 60 seconds to get the compromise solution of the multiobjective problem by PC/586. This result is better than the dispatcher's and provides a reference for the dispatcher.