目前實務上之共乘配對多仰賴人工經驗或簡單邏輯進行安排，缺乏系統化與最佳化之規劃方法，導致車輛資源利用效率有限，亦難以同時兼顧乘客需求與營運效率。若能透過最佳化模式建構多起迄點之共乘配對架構，將有助於提升旅客運輸需求之服務精準度，並改善整體系統之運輸效率。尤其南投縣屬於以觀光發展為主之地區，假日期間常因大量旅遊需求造成道路壅塞與公共運輸接駁不便等問題。若能成功推展觀光計程車共乘制度，不僅可提升遊客旅運便利性與滿意度，亦可減少私人運具使用所造成之交通與環境衝擊，進而促進觀光與交通運輸之永續發展。過去有關計程車共乘或車輛共乘配對之研究，多以都會區通勤或市區運輸需求為主要探討對象，較少針對觀光地區之旅次特性進行分析。然而，觀光旅次具有需求集中於特定時段、起訖點分散、乘客人數與服務偏好差異較大等特性，與一般都會區通勤型共乘需求有所不同，因此較難直接應用於實際觀光地區之共乘規劃。 本研究針對預約式觀光旅次，從計程車行營運業者之角度出發，建立一套系統最佳化之觀光共乘配對架構，並透過多重人流時空網路與多重車流時空網路，描述乘客與車輛於不同時間與空間節點間之移動情形。由於本研究所建構之模式屬於整數多重網路流動問題，問題規模隨著乘客需求、車輛數量與時空節點增加而快速擴大，屬於NP-hard問題。為提升大規模問題之求解效率，本研究發展混合隨機性與確定性固定鬆弛的多階段啟發式演算法，並搭配CPLEX數學規劃軟體進行求解。最後，本研究透過隨機產生不同特性之測試範例，模擬實務中可能出現之觀光共乘配對情境，並進行範例測試與結果分析，以評估所建構模式與演算法之求解績效及實務應用可行性。;In current practice， taxi-pooling matching is often arranged based on manual experience or simple decision rules， lacking systematic and optimization-based planning methods. This may limit the utilization of vehicle resources and make it difficult to balance passenger needs with operational efficiency. If an optimization model can be used to construct a taxi-pooling matching framework with multiple origins and destinations， it can help improve the accuracy of passenger service and enhance overall system efficiency. For tourism-oriented regions such as Nantou County， high travel demand during holidays often causes traffic congestion and insufficient public transport connections. A well-designed tourist taxi-pooling service can improve travel convenience， enhance tourist satisfaction， reduce the reliance on private vehicles， and promote sustainable development in tourism and transportation. Previous studies on taxi-pooling or vehicle-sharing matching have mainly focused on urban commuting or city-based travel demand， while relatively few have considered tourist trip characteristics. Tourist trips are often concentrated in specific time periods， involve dispersed origins and destinations， and vary in group size and service preferences. These characteristics make existing urban taxi-pooling methods difficult to apply directly to tourist areas. This study focuses on reservation-based tourist trips and develops a system-optimized tourist taxi-pooling matching framework from the perspective of taxi operators. Multiple passenger-flow and vehicle-flow time-space networks are constructed to describe passenger and vehicle movements across different spatial and temporal nodes. The proposed model is formulated as an integer multi-commodity network flow problem. As passenger demand， fleet size， and the number of time-space nodes increase， the problem scale expands rapidly， making it an NP-hard problem. To improve the solution efficiency for large-scale problems， this study develops a multi-stage heuristic algorithm combining random relaxation and local variable fixing， and integrates it with CPLEX for model solution. Finally， randomly generated test instances with different characteristics are used to simulate practical tourist taxi-pooling scenarios. Computational experiments and result analyses are conducted to evaluate the solution performance and practical applicability of the proposed model and algorithm.
