Modeling and Analysis of an Airport Departure Process

This paper develops and implements an analytical queueing model for an aircraft departure process. The model is formulated using data collected at LaGuardia Airport in June 1994. Departure demand is represented by a nonhomogeneous Poisson process, and service times are modeled as appropriate mixtures of exponential stages. Transient analysis of the resulting Markovian system is performed to produce a time-dependent plot of anticipated departure delay. The model and analytical results yield useful insights for airport capacity estimation and departure delay prediction. ROWING congestion at major U.S. airports has created a need for better understanding of factors that can lead to aircraft departure delays. This article presents an analytical queueing model that captures key variables in the aircraft de- parture process and offers an approach toward airport capacity estimation and departure delay prediction. The model is de- veloped and demonstrated using data collected from New York City' s LaGuardia Airport (LGA) during a single week in June 1994. LGA was selected because the airport has characteristics common to many U.S. facilities, including the frequent occur- rence of signie cant departure delays. Like many airports, LGA is cone gured with two intersecting runways. Normal operating procedures assign a primary de- parture runway, which can be viewed as the single server in a queueing system. Several dee nitions are useful in formulating an appropriate queueing model. Departure is synonymous with service completion, which occurs when an aircraft completes takeoff and clears the runway environment sufe ciently for an- other aircraft to be granted takeoff clearance. Service demand occurs when an aircraft enters the departure queue after leav- ing a passenger gate (pushback) and taxiing to the runway. Departure delay is the difference between service demand time and the initiation of service (clearance for takeoff ). Roll-out time is the total time between pushback and takeoff clearance (sum of taxi time and departure delay ). Initial analysis of the LGA activity data reveals that roll-out time varies greatly with time of day, primarily because of time- dependent variation in the frequency of scheduled pushbacks. The queueing model is consequently designed to capture var- iability in pushback rates and runway service time to produce a time-dependent plot of expected roll-out time. While this type of transient analysis has been explored for some aspects of airport operation, the departure process has received rela- tively little attention. There are various published investiga- tions of airport arrival (landing) processes. 1 › 6