Optimization Approaches to the Single Airport Ground Holding Problem

摘要

This paper surveys optimization approaches to the Single Airport Ground Holding Problem (SAGHP). We provide the first in-depth comparison of these approaches, highlight the merits and limitations of each, and evaluate the performance of four models over a large range of reward functions. We also introduce the first sequential evaluation model for the SAGHP. Motivated by current practice in Ground Delay Program planning under the Collaborative Decision Making initiative, the sequential model evaluates planned aircraft acceptance rates (PAARs) over any arbitrary time horizon. We experiment with several methods of choosing PAARs by varying the planning horizon of a static model and compare its performance to that of a dynamic heuristic. We find that research on the SAGHP has either focused on developing high performance, complex dynamic models whose outputs are not consistent with current practice, or toward developing static models that perform less well in theory but can incorporate operational considerations. Our experiments show that dynamic models outperform static models by a large margin in simulation. We find that this performance gap can be reduced by limiting the planning horizon of a static model, which allows it to imitate the ability of a dynamic model to 'wait and see'. The models were evaluated using real arrival and departure schedules from San Francisco International Airport.