隨著電子商務快速發展，都市地區之包裹配送需求持續增加，零擔運輸（Less-than-Truckload， LTL）亦逐漸成為重要之貨運配送型態。然而，目前多數都會區零擔貨運配送仍仰賴人工經驗進行車輛調度與路線規劃，容易造成配送效率下降與車輛容量利用不佳等問題。本研究以物流業者之角度出發，探討考量貨車容量都會區零擔貨物挑選暨運送排程規劃問題，考量顧客需求服務時窗、車輛容量限制、包裹組合情況、延遲服務與放棄服務等因素，以最小化負利潤為目標，結合一般性網路流動技巧與數學規劃方法，發展一套整合車流與物流概念之一般性時空網路模式，以降低問題規模並提升排程效率。 由於本研究問題屬於大型含額外限制之整數網路流動問題，難以直接利用數學規劃軟體於合理時間內求得最佳解，因此本研究進一步發展混合拉氏固定鬆弛多階段車輛分解法以及具包裹分配之車輛分解法兩種啟發式演算法，並結合隨機大型鄰域搜尋演算法，以提升大型問題之求解效率。最後，本研究透過範例測試與敏感度分析驗證模式與演算法之可行性。研究結果顯示，本研究所提出之模式與演算法能有效縮短大型問題之求解時間，並提升車輛調度效率與營運彈性，可作為物流業者進行都會區零擔貨運配送規劃之參考。;With the rapid growth of e-commerce， the demand for parcel delivery in urban areas has continued to increase， making Less-than-Truckload (LTL) transportation an increasingly important freight distribution service. However， vehicle dispatching and route planning for urban LTL freight distribution are still largely based on manual experience， which may result in low delivery efficiency and poor vehicle capacity utilization. From the perspective of logistics operators， this study investigates the urban LTL freight selection and delivery scheduling problem under vehicle capacity constraints. The problem considers customer time windows， vehicle capacity limitations， parcel consolidation opportunities， delayed deliveries， and unserved requests. The objective is to minimize negative profit. By integrating generalized network flow techniques with mathematical programming， this study develops a generalized space-time network model that combines both vehicle flow and logistics concepts， thereby reducing problem size and improving scheduling efficiency. Since the proposed problem is a large-scale integer network flow problem with additional constraints， obtaining optimal solutions directly through mathematical programming software within a reasonable computation time is difficult. Therefore， this study further develops two heuristic algorithms， namely the Hybrid Lagrangian Relaxation-Based Multi-Stage Vehicle Decomposition Method and the Vehicle Decomposition Method with Parcel Assignment， to improve the solution efficiency for large-scale problems. These methods are further integrated with a Random Large Neighborhood Search algorithm to improve the solution efficiency for large-scale problems. Finally， numerical experiments and sensitivity analyses are conducted to evaluate the feasibility and effectiveness of the proposed model and algorithms. The results show that the proposed approaches can effectively reduce computation time for large-scale instances while improving vehicle dispatching efficiency and operational flexibility. The proposed model and algorithms can serve as a useful decision-support tool for logistics operators in urban LTL freight distribution planning.
