面對未來激烈競爭與資訊充分傳遞之環境，網路庫存座位管理技術仍有許多待研究改善之處，包括：低票價旅客需求出現次序之假設不符實際、各起迄間座位配置未能彈性共用、無法即時提供座位價值資訊、缺乏對於風險偏好之適度考量等。 本研究由最佳化控制特性著手，將航空公司網路庫存機位管理問題定式為具網路資源之二階段決策問題，利用座位資源具整數且可分離特性，組合單一起迄多重票種之期望邊際座位收益函數，用以構建包含多重起迄之網路，並轉換成純網路求解，得到各旅程可使用之最佳座位配置數。 再利用網路參數分析方法進行事前最佳化運算，發展出包括：起迄間座位配置不共用之巢式訂位限額控制，網路彈性座位配置與起迄內巢式訂位限額混合控制，分群巢式訂位限額控制以及即時重新最佳化控制等四種新的機位庫存管理方法，各種方法均有其優點及創新之處。另根據旅客需求特性，模擬旅客出現訂位之整個過程，構建航空公司訂位之動態決策程序，用以比較各種訂位控制方法之收益，模擬測試結果尚稱理想。 This thesis address the problem of determining nested control policies, when different fare classes and multiple origin-destination itineraries on airline networks do not arrive in sequential blocks. Based on the airline booking process, we formulated this problem as a two-state decision problem with network recourses, which involves one optimal seat pre-allocation master problem and several single itinerary booking control parameters sub-problems. According to the optimal decision control rule of booking control, we first solved the single OD itinerary multiple fare class seat inventory management problems to exclude potentially redundant demand and used network skill to solve the multiple itinerary seat allocation problems. Whole this acts as a reference for booking control. The network sensitivity is then employed to deduce the bunch nested booking limits and real-time re-optimal control parameters within a single itinerary or the entire network. In the case of demand overlay, we adjusted the updating procedure for the related control parameters in robust form and developed a simulation scheme with eight different scenario patterns. The preliminary test results were good.
