Abstract
Mishandled baggage is one of the major issues in airports. Mishandling of baggage is ranked second in the complaints category by the Air Travel Consumer Report in 2016. In US, passengers are compensated at least $500 per mishandled bag. A large number of mishandled baggage may result in passenger dissatisfaction and is also costly to the airports/airlines. To transport passengers’ baggage from check-in counters to the unloading zones, baggage handling systems (BHS) are widely used in airports. BHS is an automated system with a set of conveyors to transport and sort most of the baggage to the unloading zones. If the unloading zones in BHS are not fully occupied, baggage will be handled by BHS; otherwise, ground crews have to manually sort the baggage which is more labor intensive, costly and has higher error rates. However, the number of baggage and their variation are usually unknown when the BHS was constructed. The efficiency of using BHS unloading zones becomes important especially for airports with an increasing number of passengers. To use BHS, airports assign each flight to an unloading zone. One of the major challenges is to consider the future uncertainty in the flight to unloading zone assignment. Not every flight will be on time due to uncertainties such as mechanical problems or weather changes. The latest flight arrival status may change in any minute and it is not possible to re-plan accordingly. Therefore, a robust plan with constant stability in performance with respect to future uncertainty realization is needed. In this research, a robust optimization model is developed to solve the flight to unloading zone assignment problems with consideration of the future uncertainty. We develop the mathematical model which considers operation requirements for avoiding baggage mishandling and generates an optimal assignment of flights to unloading zones. Our results are evidently superior to that of current airport assignment (by human experiences; taking Taiwan Taoyuan International Airport Terminal Two as case study) in reducing the amount of unassigned flights, 50% on average. Considering data uncertainties, we construct a robust optimization (RO) model using two-stage programming method to generate a robust plan. The computation speed of RO model is accelerated by implementing the progressive hedging algorithm (PHA). The results of RO model are proved to be significant by Sample Average Approximation (SAA) and indicate that RO model is superior in performance to deterministic model (about 10% improvement) and to current assignment (about 100% improvement). Validation of their results are done through simulation modeling.