空勤組員公平性排班模式之研究

Abstract

在航空運輸系統中，人力成本為航空公司第二大直接營運成本，佔總成本的10~15%。但在排班各種法規限制、工會協定下，組員班表指派(Rostering)本身即為一高複雜度問題，若再加上組員之偏好及公平性考量，將更難有效率的求解。 本研究國內首次對組員班表指派問題進行研究，主要採用Day and Ryan於一九九七年十月最新提出的「公平性系統(Equitability System)」之分析途徑，從組員之觀點出發，依組員對不同行程屬性之偏好，構建一明確、整体性公平性成本函數；並在總公平性成本最小之目標下，構建一公平性排班模式。 本研究提出一啟發式解法，將組員班表指派問題分成三個模組來構建：長勤務組合、差勤及短勤務組合行程指派模組。長勤務組合指派模組首先列舉各組員所有之候選長勤務組合行程，後構建成集合分割問題進行求解，以指派組員最適當之長勤務組合行程。差勤行程指派模組係透過分階段重複列舉與求解組員子排班問題，以決定組員差勤型態及休假日。短勤務組合行程指派模組是以每日構建啟發式解法(day-by-day)，決定組員差勤日之實際執勤內容。 本研究依國內某一最大航空公司實際營運資料建立應用個案，依其一月份航線之實際勤務組合及其實際飛勤作業的法規限制，進行模式測式及結果分析。此個案合計共有144位組員及1282勤務組合。其結果顯示，若以一般化之公平性排班模式進行求解，其問題之變數高達1024，此龐大的規模幾乎不可能求得最佳解，而利用本研究所構建之啟發式解法，分成三個模組依次求解，可有效率的獲得一高品質近似解。 本研究所指派之組員班表相對於原指派結果，在各組員行程分配公平性成本標準差減少了68%，其班表之公平性總成本則降低6%，顯著的提高組員班表指派之公平性。而組員行程中之屬性分配公平性方面，各組員行程之待命時數、不支薪時數、飛行航段及差旅時間之分配標準差均較原指派班表有一成以上的改善，亦顯著提高分配公平性，而飛行加給則和原指派班表之分配沒有顯著的差異。因此，本研究所提出之公平性排班架構，可提高組員班表指派之公平性。

The aircrew rostering problem entails the assignment of crew numbers to planned rotations. Airline companies have the monthly task of constructing personalized monthly schedules (rosters). In a recent paper, Day and Ryan [6] proposed an “Equitability System (ES)” approach for the rostering problem of flight attendants working for short-haul airline operations. The ES approach differs from the traditional “Bidding System (BS)” approach in that the ES approach uses the flight attendants preference and the duty’s quality index instead of the airline’s operating cost as the system performance measure. This study extends the ES approach to both short and long-haul airline operations, and presents a heuristic framework to solve the aircrew rostering problem based on three modules: long-haul pairing allocation, workday allocation, and short-haul pairing allocation. The long-haul pairing allocation module involves complete enumeration of all possible long-haul pairing rosters for each crew number over the roster period, and then the solution of a set partitioning optimization by the CPLEX computer packages to determine a best quality feasible long-haul pairing roster. The workday allocation module first involves the generation of many on-duty rosters, each of which consists of a duration of workdays or off days, consistent with the long-haul pairing module solution for each crew number over a subroster period and then the solution of the problem optimization to determine an optimal feasible subroster. These two steps of generation and optimization are repeated for each subsequent subroster period until a full feasible on-duty-roster for each crew number over the complete roster period. The short-haul pairing module uses the ‘day-by-day’ heuristic algorithm to decide crew workday’s duty. We applied the ES heuristic models to a case study of one of the biggest airlines in Taiwan. A case considered 144 pursers which is in the base of Taipei and 1282 rotations in the one month scheduling period. The results show that the standard deviations of equitable cost among crew rosters are decreasing 68% and the rosters total equitable cost are decreasing 6%, clearly increase the quality of solution. The use of three modules reduces the combinatorial complexity, which has over 1024 variables resulting in problem that can be solved efficiently. The heuristic models lead to efficient construction of good quality legal rosters.