Sie sind auf Seite 1von 9

A What-if Analysis Tool for Planning Airport Traffic

Dr. Sebastian Timar, Mr. Mark Peters, Mr. Paul


Davis
Architecture Technology Corporation
Campbell, USA
Ms. Mary Beth Lapis, Mr. Ian Wilson, Mr. Paul van
stimar@atcorp.com Tulder
The Boeing Company
Chicago, USA
Dr. Phil Smith mary.b.lapis@boeing.com
The Ohio State University
Columbus, USA
smith.131@osu.edu

Abstract—This paper presents the implementation and performance? How could a DMP be used to manage such
application of a prototype What-if Analysis decision support tool forecast TMIs? How could other traffic management initiatives
for airport traffic planning. The What-if Analysis tool is used to complement the DMP? What should the DMP parameters be?
predict airport traffic performance during a future time horizon DMP parameters can include categories of aircraft, target
with forecast operating conditions and to design Departure queue lengths and queue length thresholds, capacity allocations
Management Programs to mitigate the negative impacts of and time periods of application. Airport traffic management
predicted demand/capacity imbalances. Application scenarios decisions are especially complex if categorical DMPs (e.g.
include dynamic weather imposing ground hold and/or Miles-In- departures to particular fixes) are needed, or if different airport
Trial restrictions on airport departures. We demonstrate the use
departure queues need to be managed differently.
of the prototype for a historical traffic and weather scenario at
Charlotte Douglas International Airport (CLT). Future work This paper is organized as follows. Section II describes the
includes enhancing the capabilities and user interfaces of the tool, role and potential integration of what-if analysis into airport
and researching methods to predict future traffic management arrival, departure and surface traffic management. Section III
initiatives from forecast weather and traffic conditions. presents implementation of a prototype what-if analysis tool.
Section IV presents results of using the prototype for airport
Keywords—departure management programs; integrated traffic analysis and DMP design analysis with historical traffic
airport arrival, departure and surface operations; airport traffic
and weather scenarios at Charlotte Douglas International
modeling; airport traffic performance
Airport (CLT). Sections V presents future work, and Section
VI presents conclusions.
I. INTRODUCTION
The FAA Surface Collaborative Decision Making (CDM) II. OVERVIEW
Concept of Operations calls for managing departing aircraft to
This section summarizes the FAA CDM Concept of
balance airport surface traffic levels with both airport and
Operations, the NASA ATD-2 concept of operations, how
airspace capacity [1]. The NASA Airspace Technology
what-if analysis integrates with these concepts, and scenarios
Demonstration (ATD)-2 concept calls for traffic management
for what-if analysis.
technologies and operations to enable departing aircraft to
absorb delay at the gate, taxi unimpeded on the airport surface,
spend minimum time in the departure runway queue, and climb A. Role of What-if Analysis in Airport DMP Planning
continuously to cruise altitude [2][3][4]. The FAA Surface CDM Concept of Operations [1] calls for
DMPs to meter the entries of departing aircraft into the
A what-if analysis tool supports these goals by enabling airport’s movement area to balance traffic levels with both
airport traffic managers to predict airport departure, arrival and airport and airspace capacity. A Departure Reservoir
surface traffic over a time horizon of forecast operating Coordinator (DRC) is responsible for planning, monitoring and
conditions, identify time periods when traffic demand is revising DMPs. The DRC performs what-if analysis to predict
estimated to exceed capacity, and specify Departure time periods of demand-capacity imbalance and to specify
Management Programs (DMPs) to mitigate the negative effects DMPs to mitigate their negative effects. The NASA Airspace
of the predicted demand-capacity imbalance. What-if analysis Technology Demonstration (ATD)-2 concept calls for
is particularly important for planning DMPs during periods of technologies and operations for integrated planning and
dynamic convective weather. What-if analysis enables the management of airport arrival, departure and surface traffic. A
traffic managers to address important questions such as: How goal is to enable departures to absorb delay at the gate, taxi
will anticipated Traffic Management Initiatives (TMIs), such unimpeded on the airport surface, spend minimum time in the
as Miles-In-Trail (MIT) restrictions, impact airport traffic

978-1-5386-0365-9/17/$31.00 ©2017 IEEE


departure runway queue, and climb continuously to cruise The seven steps of the departure management process are
altitude [2][3][4]. depicted in Fig. 1. Steps 1-5 comprise what-if analysis, and
steps 6-7 comprise departure management. Of the what-if
DMPs are implemented for bounded time periods when analysis process, steps 1-3 comprise airport demand analysis to
runway queue lengths are predicted to exceed a Target assess traffic performance over a forecasting time period, and
Departure Queue Length (TDQL). DMPs assign Target steps 4-5 comprise DMP design analysis to mitigate the
Movement Area Entry Times (TMATs) to departures to control negative effects of the demand/capacity imbalance predicted to
traffic levels in the airport movement area. The TMATs may occur. Specifically, in step 1, the DRC performs the what-if
minimize taxi-out time while meeting the TDQL to maintain analysis of the far-term forecasts (e.g., 45 minutes) of traffic
departure throughput. Aircraft operators comply with TMATs schedule, operating conditions and departure restrictions for
by holding aircraft at the gates and other delay reservoirs in the the traffic planning time period (e.g., 1-2 hours) to assess
non-movement area of the airport. DMPs are planned and airport traffic performance (e.g. throughput, queue lengths and
implemented by a DRC. A what-if analysis capability and taxi times). In step 2, the DRC analyzes the impact of potential
candidate requirements are called for in [1][5][6][7] to assess changes to restrictions, runway rates, and scheduled traffic. In
the impact of changes in departure restrictions, airport or route step 3, the DRC predicts the need for a DMP, and DMP
configurations, traffic schedules or DMP parameters on airport
parameters such as start time and duration. In step 4, the DRC
traffic performance, and to plan or revise DMPs as needed to iteratively proposes and evaluates the effect of candidate
address these changes. DMPs on managing demand/capacity imbalances and meeting
NASA’s ATD-2 is developing traffic management systems airport performance objectives. In step 5, the DRC specifies the
for airports. The systems comprise traffic scheduling and DMP using the selected parameters. In step 6, the DMP
management technologies, decision support tools and specifies TMATs for departures using the near-term forecasts
operations for integrated management of arrival, departure and (e.g., 15 minutes) of traffic schedule, operating conditions and
surface operations to enable ideal trajectories. Ideal departure departure restrictions for the traffic planning time period. In
trajectories comprise delay absorbed at the gate, unimpeded step 7, at the current traffic schedule, operating conditions and
taxi on the airport surface, minimum time in the departure departure restrictions, the ATD-2 system performs integrated
runway queue, and continuous climb to cruise altitude [2][3]. arrival-departure management.
The DMP can support the ATD-2 system by managing and
B. Conditions for What-if Analysis
metering departures to minimize delay in taxi-out and climb,
and facilitating the use of airport surface delay reservoirs to What-if analysis is useful when changes in TMIs impacting
manage traffic. In the ATD-2 concept the departure controller, airport departures, airport or route configurations, traffic
supported by the NASA Spot and Runway Departure Advisor schedules or DMP parameters are forecast to occur. Changes in
(SARDA) [8] or similar tools and operations, can optimize the TMIs include imposing MIT restrictions on departures to
sequencing of departures at the movement area entry points particular fixes, airspaces or destination airports in response to
and the runways in conjunction with the metering implied by weather or traffic congestion. Changes in airport operating
the TMATs. The ramp controllers can optimize the use of conditions include changing the configuration or capacities of
delay reservoirs as per the TMATs and the gate pushback times airport runways or taxiways or terminal airspace departure
they may imply. A potential integration of what-if planning fixes. Changes in airport traffic include changing runways and
with FAA Surface CDM DMPs and NASA ATD-2 traffic fixes assigned to flights, substituting flights, changing the
management is depicted in Figure 1. estimated off-block time or changing the gating plan. DMP
changes include extending or terminating the DMP and
evaluating different queue lengths.
In such dynamic forecast weather, traffic and airport
conditions, what-if analysis helps plan DMPs to minimize the
negative impact on airport traffic performance. Previous
research [9][10][11] has characterized the nature of the
challenges in managing weather patterns and has demonstrated
how decision support tools can improve the use of DMPs. The
what-if analysis would be used to design categorical (fix-,
runway- or destination-specific) DMPs and other traffic
management initiatives for departures subject to the particular
forecast restrictions. What-if analysis enables proactive
implementation of DMPs based on forecast traffic management
initiatives or other conditions, rather than reactive
implementation after the conditions are in effect.

C. Illustrative Example
Fig. 1. Integration of What-if Analysis with Airport Departure Management. As an example, we consider the impact of an actual
weather scenario at CLT shown in Fig. 2. For brevity, we only
address the impact on north-, east- and south-bound departures
via runway 18L after the storm has passed over the airport. At
4:45PM local time, the weather closed the airport and
continued to move east. Scheduled departures from runway
18L were delayed during the departure stop from 4:45PM to
5:15PM, and were impacted for another hour as the storm
moved across the departure fixes of the east gate.

Fig. 3. Block Diagram of What-if Analysis Prototype Components.

Data inputs specify pre-defined parameters for different


typical airport operating conditions. The user interface enables
selecting particular conditions and parameters and tailoring
them to the forecast operating conditions for evaluation, and
presents metrics for assessing predicted airport traffic
performance. The fast-time simulation is used to predict airport
Fig. 2. Example Convective Weather Event Sequence Causing an Airport traffic under the specified conditions without and with the
Departure Stop. DMP. The DMP emulation meters departing aircraft to the
airport movement area as per user-specified parameters.
Under these conditions, what-if analysis could have been Metrics computation enables assessing airport traffic
performed as follows. 30 minutes prior to 4:45 PM, the airport, performance. Each component is described in this section.
Terminal Radar Approach Control (TRACON), Air Route
Traffic Control Center (ARTCC) and the Air Traffic Control A. Data Input
System Command Center (ATCSCC) traffic managers provide
Data input to the what-if analysis tool include scheduled
the DRC with a 2-hour traffic management forecast. The
airport traffic, airport and terminal airspace configuration
forecast anticipates a departure stop for 30-45 minutes and 30
information, and forecast traffic flow restrictions.
MIT for departures via east gate departure fixes LILLS and
MERIL for 2 hours. The DRC performs what-if analysis and 1) Airport Traffic Schedule and Configuration
determines that the quantity of scheduled departures and the The prototype what-if analysis tool takes as input the airline
proposed traffic restrictions will result in excessive departure scheduled airport arrivals and departures for the time horizon
queue lengths and surface traffic levels. The DRC evaluates of assessment. Embedded in these data are the implied
possible DMPs to mitigate the negative impacts on airport operating configuration of the airport and airspace. The data for
traffic: a DMP for south gate departures from runway 18L to each flight include the anticipated origin or destination airport,
address the backlog of flights after the ground stop, and DMPs arrival or departure route, runway, terminal gate, non-
for departures via fixes MERIL and LILLS to address the movement area entry or exit point (spot), gate out or in time,
ongoing MIT restrictions. For departures via fixes MERIL and and other flight information (e.g., aircraft type). The user
LILLS, the DRC arrives at a DMP with a TDQL of 8 with an interface enables specifying the operations rate of each airport
upper threshold of 10 and a lower threshold of 8 to start at 5:15 runway. For prototyping, these data were derived from
PM and end at 6:45 PM. The DMP extends beyond the MIT historical airport traffic and listed in a format compatible with
restriction end time to address the buildup of flights to these the NASA Surface Operations Simulator and Scheduler
fixes. For departures from runway 18L to the south gate (SOSS) simulation software [12]. For actual implementation,
departure fixes subject to the ground stop, the DRC determines data would be obtained from the FAA System Wide
a DMP to start at 5:15 PM and end at 6:45 PM, with a higher Information System (SWIM) Traffic Flow Management
TDQL of 10 because those fixes do not have MIT restrictions. System (TFMS) flight schedule data, airline gate plans and
typical airport operating procedures. Actual implementation
III. PROTOTYPE IMPLEMENTATION should also account for the state of airport arrival, departure
and surface traffic at the start of the forecast time period,
The prototype what-if analysis tool comprises data inputs, a including gate occupancy, movement area traffic levels and
user interface, a traffic simulation, a DMP emulation, and locations, and incoming arrivals and outgoing departures.
assessment metrics. A block diagram of the components of the
prototype is depicted in Fig. 3. 2) Traffic Management Initiatives
The what-if tool must also account for TMIs planned or
anticipated as per forecast conditions. The prototype models
MIT restrictions on departures to particular departure fixes or
destination airports or airspaces. The input restriction data applying in-trail separation at the runway between successive
includes the control point, the start and end times, and the in- departures to that control point to satisfy the MIT restriction at
trail separation. The user interface enables adding, removing or the control point. Based on analysis of CLT operations [14],
altering the restrictions. Modeling is to be extended to TMIs the model applies inter-departure spacing which is 2/3 of the
including Expected Departure Clearance Times (EDCTs), MIT at the control point. This is converted to time assuming a
Ground Delay Programs (GDPs), Airspace Flow Programs takeoff speed of 150 knots [15]. Modeling includes splitting
(AFPs), Severe Weather Avoidance Programs (SWAPs) and the departure sequence to maintain throughput.
others. For actual implementation, restrictions would be
estimated from forecast traffic and weather conditions. The gate node queuing models account for single aircraft
occupancy for the modeled occupancy time (service time).
Departures enter their assigned gates as per their airline-
B. Airport Traffic Simulation scheduled gate pushback times. Arrivals enter their assigned
The prototype what-if analysis tool uses a fast-time gates when the gate is unoccupied and there is sufficient time
queueing simulation to predict airport traffic and performance. to occupy the gate. Tail tracking of arrivals and departures is
The simulation uses a graphical node-link model of the airport not accounted for. This modeling approach can result in
surface and terminal airspace, and parameters representing the significant taxi-in delay and should be enhanced.
service rates at queuing nodes and the transit times of aircraft
along links connecting the nodes. 3) Modeling Parameters
Queue service times for the fix nodes correspond to in-trail
1) Airport Graphical Model separation of 3-nautical miles and a speed of 250 knots.
The graphical model of the airport surface and terminal Runway queue service times correspond to 30 operations per
airspace captures the individual terminal gates, spots, runways hour. Spot queue service times of 5 minutes correspond to time
and arrival and departure fixes as nodes. The nodes are for movement area entry clearance to be requested, granted and
connected by links representing airport surface and terminal acted upon. Terminal gate queue service times correspond to
airspace transit routes. The gate, spot, runway and fix nodes occupancy of 30 minutes.
and their interconnections are synthesized from the input traffic
scenario file. Interactions of traffic at points other than the Links are modeled as the nominal transit times of aircraft
gates, spots, runways and fixes are not currently modeled. between pairs of nodes. For prototyping focused on CLT,
Future enhancements include runway crossing and other values were estimated from previous analyses of multiple days
congestion points, gate groups and other delay reservoirs. of ASDE-X surveillance data for CLT [16]. The gate-spot time
of 4.0 minutes is an average of all the gate-spot transit times.
2) Traffic Queuing Simulation The spot-runway time of 2.0 minutes is the 5th percentile of all
The traffic queuing simulation models each node in the the spot-runway times. The runway-fix time of 14.0 minutes is
airport graphical model as a queue. Nodes are points where the average of values obtained from high-fidelity simulations
traffic flow is constrained. If the arrival rate of traffic to the of aircraft flying the CLT MERIL departure procedure [17] and
node exceeds the effective traffic flow rate that can be met at scaling that average to other CLT departure procedures based
the node, the aircraft form a queue as they wait for entry to, on flight procedure leg distances obtained from National Flight
and passage through, the node [13]. Links are transit segments Data Center (NFDC) data.
between nodes.
Queue parameters for each node include a service time, C. Departure Management Program Emulation
Tservice, to process each flight, and a queue size limit, Qmax. The The prototype what-if analysis tool emulates a user-
service time can represent the time for the controller to clear specified DMP as TMATs to meter the entries of departures
the aircraft to take off or enter the movement area, the into the movement area of the airport. The what-if analysis tool
minimum in-trail separation with the previous flight for safety, then simulates airport traffic under the DMP in fast-time to
for example. The queue size limit specifies the maximum assess the impact of the DMP on airport traffic performance.
number of aircraft that may wait in line for service by the node. DMP parameters include categories of flights, start and stop
This can represent the number of aircraft a taxiway or time, TDQL and its thresholds, unscheduled demand buffer,
hardstand can accommodate. Links are modeled as nominal and many others. The DMP parameters may be varied, and
inter-node transit times, Ttransit. Node models process flights in successive simulations conducted, to find DMP parameters for
first-in, first-out order. A flight, j, is serviced by a node, i, at its implementation in the future time period of interest.
node entry time which is the later of its node arrival time or the
1) Target Movement Area Entry Time (TMAT) Scheduling
exit time of its node predecessor, Ti,jentry = max(Ti,jarrival, Ti,j-
1 The prototype tool implements a DMP emulation. The
exit). Each flight exits the node at its node entry time (i.e., the
time at which it is serviced) plus the node’s service time, Ti,jexit DMP emulation generates a candidate schedule of TMATs for
= Ti,jentry + Tiservice. Delay accrued due to waiting in the queue is departures as per user-specified conditions including departure
Ti,jdelay = Ti,jentry – Ti,jarrival. Link models propagate each flight to rates for each departure runway, applicable TMIs on individual
the next i+1 node in its prescribed route, Ti+1,jarrival = Ti,jexit + departures, and candidate DMP parameters. Each departure’s
TMAT implies a scheduled gate pushback time to absorb
Ti,i+1transit, processing flights in their order of exit from the node.
scheduled delay, which is the TMAT minus the nominal gate-
The runway node queuing models account for MIT spot transit time for its terminal. In turn, the what-if tool
restrictions on departures to particular fixes, airports or simulates airport traffic using the DMP gate pushback times to
airspaces by modeling the typical local controller practice of assess its airport traffic performance impact. The time-based
scheduling methods follow the principles, including the Order explore alternative scenarios and conditions, evaluate
of Consideration algorithm, specified in [18]. The prototype alternative DMP designs, and assess sensitivity of airport
includes a feedback control law which adjusts the movement traffic performance to changing conditions. The prototype
area entry rate of departures used in scheduling the TMATs to implementation is very simple, but sufficient to provide access
maintain the target departure queue length at each runway. to the data of interest. Future work includes extending and
enhancing the interface.
TMAT scheduling initializes with identifying the set of
departures subject to the DMP and the terminal gates, runways
and departure fixes they use. These are scheduling reference F. Verification
points. The estimated times of arrival of a departure flight to its To verify the airport traffic simulation, we verified the
reference points are obtained from its airline-scheduled gate modeling logic and compared simulated airport traffic to FAA
out time and the nominal transit times modeled for the airport Aviation System Performance Metrics (ASPM). We verified
links. The TMAT scheduling initially sorts the flights at each the modeling logic by ensuring the trajectories of the simulated
reference point based on their estimated times of arrival. At flights (described by the sequence of gate, spot, runway and fix
each reference point, TMAT scheduling sorts flights by nodes and simulated times at those nodes) complied with the
increasing time, and locally schedules their times of arrival, queue service rates, queue entry/exit sequencing, link transit
delaying the flights as needed, to satisfy flow restrictions, times, and departure restrictions.
separation minima, runway operations rate or other inter-flight We compared simulated airport traffic to FAA ASPM data
time spacing limits associated with the reference point. for August 8, 2014, a day where CLT consistently operated in
TMAT scheduling ensues with scheduling the time of each the south flow runway configuration with typical departure
departure at each of its reference points, working sequentially restrictions. The modeled traffic schedule was created from
through each reference point, delaying the scheduled time of operations data as described in [17] and captured 19-hours of
arrival of the flight at the reference point as needed to satisfy arrivals and departures. Simulated operations were fewer than
in-trail separation with the previous flight at the reference point ASPM-recorded scheduled operations due to deriving the
as per the applicable constraints. The scheduling algorithm traffic input file from actual operations; flight cancellations and
iterates until converging on scheduled times of arrival at consolidations by aircraft operators inherent in the operations
reference points which yield no conflicts or further delays data are not in the schedule data. Simulated throughput was
while satisfying all applicable inter-aircraft spacing limits. The lower than ASPM due to a conservative modeled runway
TMATs imply a takeoff sequence and schedule at the departure capacity. Average taxi-out time from the simulation compared
runway which satisfies the specified runway operations rate, closely to ASPM. Average taxi-in time from the simulation
departure restrictions (in this case, MIT restrictions for specific was higher due to the simple modeling of gate management.
fixes and time periods) and target queue length, assuming For verification of the DMP emulation, we used the
unimpeded transit to the departure runway queue. prototype tool for other CLT scenarios of high traffic demand
2) Queue Length Control and significant TMIs. We verified the DMP emulation
Throughout the process, TMAT scheduling monitors the specified TMATs such that the runway queue lengths complied
resulting departure queue length at each runway with a DMP, with the TDQL, throughput complied with the runway rates,
and uses a control law based on the error between the actual and departures complied with their MIT restrictions.
queue length and the TDQL to adjust the departure pushback
rate used for scheduling TMATs. IV. PROTOTYPE DEMONSTRATION
We evaluated and demonstrated the what-if analysis tool
D. Traffic Performance Analysis using multiple historical traffic and weather scenarios for CLT.
The prototype tool evaluates airport traffic performance We present the results for one particular scenario.
indicators for the simulated traffic to assess the impact of
demand/capacity imbalances on airport operations, and to A. Conditions
estimate the impact of DMPs to mitigate those impacts. Time-
Conditions comprised a schedule of arrivals and departures
series and aggregate metrics for departures and arrivals for
representing aircraft operator scheduling intent, and traffic
each runway are evaluated. Metrics for departures include per-
restrictions representing management initiatives to address
runway throughput, runway queue length relative to the TDQL,
weather events. Modeling used operator Out-Off-On-In
average taxi-out times and gate-out delay. Metrics for arrivals
(OOOI) data for a month of operations at CLT to estimate
include per-runway throughput, taxi-in time, and gate-in delay
pushback and landing times, ramp control procedures for a
(due to gate holding of departures). The user interface presents
major aircraft operator at CLT to obtain gate-spot assignments,
time-histories as bar or line charts of data in 15- or 60-minute
Aircraft Situation Display to Industry (ASDI) data to estimate
bins, and presents aggregate metrics in tables.
arrival and departure fixes, historical weather data, and data
from NASA describing ATCSCC TMIs to model restrictions.
E. User Interfaces
The prototype tool includes a user interface for specifying Airport traffic conditions represent August 8, 2014
the scenario, configuring the DMP, and assessing airport traffic operations at CLT discussed in the Verification section. The
flow performance. The goal is to provide the DRC and scheduled pushback times of departures exhibited 9 banks
stakeholders with an ability to have situational awareness, occurring at approximately 1.5 hour intervals, peaking at 25 to
35 departures in a 15-minute time period, depicted in Fig. 4.
Our demonstration evaluated operations from 2:08 PM to 5:58
PM, local time. During this time, scheduled departures
exhibited two 45-minute banks of approximately 20 pushbacks
per 15-minute period, bounded by the dashed lines in Fig. 4.

Fig. 5. Airport demand analysis predicted departure throughout, runway


queue length, taxi-out time and gate delay by runway.

In this case, the TDQL (Target Departure Queue Length)


was specified somewhat arbitrarily. For actual operations, the
TDQL would be specified to maximize airport throughput,
Fig. 4. Modeled Quarter-hour Scheduled Pushbacks for CLT on August 8, while avoiding excess traffic, as per forecast runway
2014 and 4-hour Period Selected for What-if Analysis. configuration and other operating conditions [19][20][21].

We identified multiple weather days of interest for CLT in 2) DMP Design Analysis
August 2014, corresponding to convective weather at CLT or We used the prototype tool to perform DMP design
proximate facilities, resulting in significant restrictions on analysis. While numerous other DMP types and parameters
departures, including MIT restrictions, Calls For Release may be specified, for example multiple coincident categorical
(CFRs), and Expected Departure Clearance Times (EDCTs). DMPs for all departures to different fixes and/or runways, we
For this demonstration, we considered traffic restrictions for implemented a single DMP for all departures. Automated
August 18, 2014, where heavy rainfall near CLT resulted in analysis of the predicted queue lengths proposed the DMP start
MIT restrictions that impacted 180 departures, CFRs that at 3:00 PM, before the runway 18L queue exceeded the TDQL
impacted 41 departures, and EDCTs that impacted 2 of 3 aircraft, and end at 7:00 PM, after the runway 18C queue
departures. During the period from 2:08 PM to 5:58 PM, local drops below the TDQL. Our analyses of other periods at CLT
time, the restrictions comprised 15 MIT for departures via the of significant departure push and departure restrictions
MERIL fix from 2:28 PM to 5:10 PM. This impacted 40 of the determined that initiating the DMP prior to the restrictions
159 departures in this time period. Of the 40 departures, 36 avoids queue buildup and enables consistent and efficient
departed runway 18L, and 4 departed runway 18C. traffic performance throughout the time period of restricted
traffic flow. Results with the DMP are depicted in Fig. 6.
B. Results
The prototype what-if analysis tool was used for airport
demand analysis and DMP design analysis for the specified
conditions at CLT. Results are presented in this section.
1) Airport Demand Analysis
The prototype tool was used to perform airport demand
analysis for the 4-hour period of scheduled departures under
the prescribed departure restrictions. The airport traffic state at
the start of the prediction period did not account for prior
traffic; in practice the prediction should initiate from the
current or estimated traffic state. Analysis of scheduled
pushbacks identified quarter-hour time periods where their
quantities exceeded airport departure capacity. Predictions of
airport traffic movement estimated the departure queue lengths
for runways 18L and 18C would exceed 3 aircraft, reaching 11
and 12 aircraft respectively, and average quarter-hour taxi-out
times would peak in excess of 30 minutes. Throughput would Fig. 6. DMP design analysis proposed departure throughout, runway queue
saturate at the prescribed 30 operations per hour. Results are length, taxi-out time and gate delay by runway.
depicted in Fig. 5.
The results show that, with the DMP, throughput saturates Regarding tool enhancements, Fig. 7 depicts a notional
at the prescribed 30 operations per hour (7-8 operations per interface for modeling forecast TMIs, planning DMPs and
quarter-hour). For runway 18L, the departure queue length evaluating the impact on airport departure and arrival traffic
comes close to the TDQL of 3 aircraft, peaking at 4 aircraft, performance for baseline (B) and alternative (A1) operating
and average quarter-hour taxi-out times reduce to 15 minutes. plans.
The reduced taxi-out delay is shifted to gate departure delay,
which reaches 10 minutes for runway 18L departures, and
almost 12 minutes for runway 18C departures.
For runway 18C, however, the queue length is predicted to
reach 7 aircraft and average taxi-out time reaches 25 minutes.
This is because runway 18C is shared with arrivals, and
arrivals land to runway 18C in the subject time period. To
account for the shared runway use, the DRC could reduce the
departure rate of the DMP for runway 18C to release fewer
departures to runway 18C.
What-if analysis predicted the changes in aggregate
performance of airport departure traffic as a result of
implementing the DMP listed in TABLE I.

TABLE I. IMPACT OF DMP ON PERFORMACE OF DEPARTURES BY


RUNWAY
Departure Traffic Performance
Runway & Average Maximum Average Average Average
DMP Throughp Queue Queue Taxi-out Gate
Condition ut Length Length Time Delay
(Aircraft) (Aircraft) (Aircraft) (Minutes) (Minutes)
No 22 12 3 17.1 0.0
18L Fig. 7. Notional interface for what-if analysis tool to predict airport traffic
Yes 22 4 1 12.6 2.9 performance and design DMPs.

No 20 14 4 18.6 0.0 This depicts the example scenario of weather passing over
18C
Yes 20 10 3 16.3 3.0 CLT, calling for a departure ground stop on runway 18L
ending at 2115 ZULU, and 30MIT for departures via fixes
MERIL and LILLS, from 2115 to 2215. The departure queue
What-if analysis also predicted the changes in aggregate lengths are predicted to exceed 16 aircraft in the 15-minute
performance of airport arrival traffic as a result of the DMP. time periods from 2115 to 2230. The DMP reduces the
Average arrival throughput for runways 18R, 23 and 18C were predicted queue lengths to 16 aircraft, within the thresholds of
predicted to remain at 20 aircraft, 17 aircraft and 7 aircraft, +20/+10 aircraft. Queue length, off-out and off delay metrics
respectively. ON-to-IN times for arrivals from runways 18R, indicate significant improvement in airport traffic performance
23 and 18C were predicted to increase by 1.6 minutes, 1.0 under the alternative. Situational awareness is enhanced with a
minutes and 2.1 minutes, respectively, if departures were held planview display of the traffic color-coded by applicable
at their gates during the DMP and gate assignments for arrivals restriction, depicted in Fig. 8.
were maintained.

V. FUTURE WORK
Future work for what-if analysis includes developing
methods for estimating traffic management initiatives from
forecast weather, developing and enhancing what-if analysis
tool prototypes with improved traffic prediction capabilities
and DMP planning features, and further exploring what-if
analysis operations including stakeholder collaboration in
situational assessment and DMP planning.
Regarding conditions forecasting, proactive designs and
implementations of DMPs will have to be developed based on
forecasts of traffic management initiatives. Methods are needed
to estimate traffic flow management restrictions and other
conditions from weather forecasts as input to the what-if
analysis, and to use what-if analysis to design DMPs to
accommodate those forecasts and their uncertainties.
which could impact airport traffic performance. Airport traffic
modeling should capture surface characteristics which impact
traffic flow and timing, including arrival and departure runway
and route resources, arrival-departure interaction points, traffic
control points and delay reservoirs while enabling rapid
evaluation. DMP emulation requires DMP parameter
configuration, algorithms for TMAT computation, and
modeling of traffic control as per the TMATs. Interfaces for
specifying airport conditions, traffic demand and traffic
restrictions are useful for exploring and assessing the impact of
possible deviations from current or anticipated operating
conditions, and designing responses to mitigate their effects.
Algorithms for recommending DMP parameters, and
accounting for the uncertainty in forecast data and parameters,
can significantly enhance the decision support of the what-if
analysis tool.

ACKNOWLEDGMENT
We thank Savita Verma and Yoon Jung of NASA Ames
Research Center for their support in conducting this work.
Fig. 8. Notional interface for what-if analysis tool for situational awarenesss
of airport traffic. Flights color-coded by applicable departure restrictions. REFERENCES
[1] Federal Aviation Administration. 2012. “US Airport Surface
The interface highlights in yellow the locations of departing Collaborative Decision Making (CDM) Concept of Operations in the
aircraft assigned to the MERIL departure fix. The DRC can Near-Term.”
allocate these aircraft to delay reservoirs in the non-movement [2] National Aeronautics and Space Administration. 2015. “A Concept for
and movement areas of the airport as needed to satisfy TMATs Integrated Arrival/Departure/Surface (IADS) Traffic Management for
and the departure restrictions. the Metroplex, Airspace Technology 2 Concept of Operations
Synopsis.”
Future work includes enhancing the what-if tool modeling [3] National Aeronautics and Space Administration. 2016. “Airspace
and analysis capabilities. This includes enhancing airport and Technology Demonstrations (ATD) Project Industry Day, ATD-2 IADS
airspace representation, in particular to capture delay reservoirs Metroplex Traffic Management Overview Brief.”
(e.g., hardstands and taxiway segments used for holding [4] Saraf, A., Timar, S., Shen, N., Idris, H. 2015. “Preliminary Queueing
aircraft) and user interfaces and utilities to support managing Analysis of Integrated Departure Operations in Metroplex Systems,”
Proceedings of the 34th Digital Avionics Systems Conference (DASC),
their use. Future work also includes enhancing the tool to Prague, Czech Republic.
perform sensitivity analysis of traffic performance to variations [5] Metron Aviation. 2012. “Surface Collaborative Decision Making
in forecast operating conditions and DMP parameters. Concept, HITL 01 Summary, August 21-23, 2012.”
Extending and enhancing interfaces for aircraft operators and [6] Metron Aviation. 2012. “Surface Collaborative Decision Making
other stakeholders would be helpful for common situational Concept, HITL 01 Summary, June 26-28, 2012.”
awareness and collaborative traffic planning. [7] Metron Aviation. 2013. “Surface Collaborative Decision Making, HITL
7 Summary, January 15-17, 2013.”
VI. CONCLUSIONS [8] Jung, Y., Montoya, J., Gupta, G., Malik, W., Tobias, L., Want, H. 2011.
“Performance Evaluation of a Surface Traffic Management Tool for
This paper presents a prototype what-if analysis tool for Dallas/Fort Worth International Airport.” Proceedings of 9th USA
airport traffic planners to predict airport traffic performance Europe Air Traffic Management Research and Development Seminar
(ATM2011), Berlin, Germany.
over forecast operating conditions and design DMPs to
mitigate the negative effects of predicted demand-capacity [9] Smith, P.J., Weaver, K., Fernandes, A., Durham, K., Evans, M.,
Spencer, A. and Johnson, D. 2012. “Supporting distributed management
imbalances. The tool is demonstrated for an historical traffic of the airport surface.” Proceedings of the 2012 AIAA Digital Avionics
and weather scenario at CLT to assess airport traffic Systems Conference (DASC), Williamsburg, VA.
performance, specify the start and end times of a DMP for [10] Fernandes, A., Smith, P.J., Durham, K. and Evans, M. 2015.
airport-wide departures, and evaluate the resulting traffic “Individual Problem Representations in Distributed Work.” Proceedings
performance. of the 2015 International Symposium on Aviation Psychology, Dayton,
OH.
General requirements for what-if analysis include the [11] Smith, P.J., Fernandes, A.B., Durham, K., Evans, M., Spencer, A.,
following. Airport demand analysis requires representing Beatty, R. Wiley, E. & Spencer, A. 2011. “Airport Surface
airport traffic and weather scenarios and modeling airport Management as a Distributed Supervisory Control Task.” Proceedings
of the 2011 AIAA Digital Avionics Systems Conference (DASC),
traffic movement. DMP design analysis requires specifying the Orlando, FL.
parameters for one or more DMPs, and emulating the DMP to
[12] Windhorst, R.. 2012. “Towards a Fast-time Simulation Analysis of
assess its impact on airport traffic performance. Scenario Benefits of the Spot and Runway Departure Advisor.” Proceedings of
modeling includes anticipated airport arrivals and departures, the AIAA Guidance, Navigation and Control (GNC) Conference,
airport operating conditions and traffic management initiatives Minneapolis, MN.
[13] deNeufville, R., Odoni, A. 2003. Airport Systems: Planning, Design and [18] Wong, G.. 2000. “The Dynamic Planner: The Sequencer, Scheduler, and
Management. McGraw-Hill, New York, New York, pp. 819-850. Runway Allocator for Air Traffic Control Automation.” NASA/TM-
[14] Lohr, G. 2016. “Air Traffic Operations at the Charlotte Douglas 2000-209586.
International Airport (KCLT).” [19] Anagnostakis, I., Idris, H., Clarke, J.-P., Feron, E., Hansman, R. J.,
[15] “Airliner Takeoff Speeds.” Odoni, A., Hall, W. 2000. “A Conceptual Design of A Departure
http://www.aerospaceweb.org/question/performance/q0088.shtml. (As Planner Decision Aid.” Proceedings of the 3rd USA/Europe Air Traffic
of 19 September 2016). Management R&D Seminar, Napoli, Italy.
[16] Saab Sensis Corporation. 2012. “Systematic Assessment of Surface [20] Welch, J., Bussolari, S., Atkins, S. 2001. “Using Surface Surveillance
Optimization Functions—Methods and Metrics Report, for the To Help Reduce Taxi Delays.” Proceedings of the 2001 AIAA Guidance,
Adaptation of a Surface Management Tool to Multiple, Capacity Navigation and Control (GNC) Conference, Montreal, Canada.
Constrained Airports Project.” Contract NNA11AC50C. [21] Nakahara, A., Reynolds, T. 2011. “Analysis of a Surface Congestion
[17] Architecture Technology Corporation. 2016. “Methods of Increasing Management Technique at New York JFK Airport.” Proceedings of the
Terminal Airspace Flexibility and Control Authority, Option Year 1 11th AIAA Aviation Technology, Integration and Operations (ATIO)
Final Report.” Contract NNA14AC42C. Conference, Virginia Beach, VA.

Das könnte Ihnen auch gefallen