1: \documentclass{article}
2: \setlength\topmargin{0.5in} \setlength\headheight{0in}
3: \setlength\headsep{0in} \setlength\textheight{8.5in}
4: \setlength\textwidth{6.5in} \setlength\oddsidemargin{0in}
5: \setlength\evensidemargin{0in} \setlength\parindent{0.25in}
6:
7:
8: \title{A Dynamic I/O Model for TRACON Traffic Management}
9:
10: \author{
11: Maxime Gariel\thanks{Graduate Research Assistant, School of Aerospace Engineering,
12: maxime.gariel@gatech.edu, corresponding author},
13: \ John-Paul Clarke\thanks{Professor, School of Aerospace
14: Engineering, john-paul.clarke@ae.gatech.edu.}
15: \ and Eric Feron\thanks{Professor, School of Aerospace
16: Engineering, feron@gatech.edu.}\\
17: {\normalsize\itshape
18: Georgia Institute of Technology Atlanta, GA, 30332-0150, USA}\\
19: }
20:
21: % Define commands to assure consistent treatment throughout document
22: \newcommand{\eqnref}[1]{(\ref{#1})}
23: \newcommand{\class}[1]{\texttt{#1}}
24: \newcommand{\package}[1]{\texttt{#1}}
25: \newcommand{\file}[1]{\texttt{#1}}
26: \newcommand{\BibTeX}{\textsc{Bib}\TeX}
27: \usepackage{subfigure}
28: \usepackage{graphicx}
29:
30: \begin{document}
31:
32: \maketitle
33:
34: \begin{abstract} This work investigates the TRACON flow management around a major
35: airport. Aircraft flows are analyzed through a study of TRACON
36: trajectories records. Rerouting and queuing processes are
37: highlighted and airport characteristics are shown as function of the
38: number of planes in the TRACON. Then, a simple input-output TRACON
39: queuing and landing model is proposed. This model is calibrated and
40: validated using available TRACON data. It reproduces the same
41: phenomenon as the real system. This model is used to show the impact
42: of limiting the number of aircrafts in the TRACON. A limited number
43: of aircraft does not increase delays but reduces the controller's
44: workload and increases safety.
45: \end{abstract}
46:
47:
48: \section{Introduction}
49:
50: Airspace, especially Terminal Radar Approach Control (TRACON) area
51: around major airports becomes more and more congested . As the
52: concentration of aircrafts is the highest, TRACON areas are very
53: critical areas for controllers. The objective of this paper is to
54: present a study about the dynamics of the traffic congestion in the
55: arrival process at a major airport. A motivation is to regulate the
56: flows of entering aircraft in the TRACON area in order to reduce air
57: traffic controller's workload and to avoid control tactical
58: difficulties. This is already the aim of the Center TRACON
59: Automation System (CTAS)\cite{CTAS} that provides automation tools
60: for planning and controlling arrival air traffic. Many traffic
61: control techniques and tools are applied to reduce traffic
62: congestion. Ground Delay Program (GDP) \cite{GDP} is used to
63: decrease the rate of incoming flights into an airport by delaying
64: takeoffs, when it is projected that arrival demand will exceed
65: capacity. For en-route control, Traffic Management Advisor (TMA)
66: \cite{TMA} and Multi-Center Traffic Management Advisor (McTMA)
67: \cite{McTMA} are used to control arriving aircraft that enter the
68: Center from an adjacent Center or depart from feeder airports within
69: the Center. Those systems have already proven their capabilities
70: \cite{McTMAresults}. TRACON area control can use programs such as
71: Descent Advisers (DA) or Final Approach Spacing Tool (FAST)
72: \cite{FAST}. Finally, ground operations and efficient runway
73: operation planning \cite{Ioannis} can also improve landing
74: capacities. The aim of those different layers of traffic control is
75: to ensure an arrival flow as smooth as possible. We are interested
76: in adjusting the density in the TRACON area and optimizing airports
77: landing capacities by regulating the incoming traffic flow.
78: Theoretically, runway capacities can be calculated using analytical
79: formulas and nominal separation standards. In this paper, those
80: capacities are determined and analyzed through available TRACON
81: data. Then, we propose an experimental queuing model to highlight
82: airport performance function of the demand. The airport performance
83: is the effective landing capacity, which depends on the number of
84: runway in use. This arrival model is based on a similarly model
85: developed by Nicolas Puget in \cite{pujet}. The model proposed is
86: validated with San Francisco TRACON records. Finally, the model is
87: used to show that limiting the number of aircrafts in the TRACON
88: does not increase delays but reduces the controller's workload and
89: increases safety.
90:
91:
92: \section{Data Presentation and Analysis}\label{section:dataPresentation}
93: \subsection{Data Source}
94: The data used in this article were provided by San Francisco
95: International Airport Noise Abatement Office. The data are the
96: records of all the planes in the San Francisco Bay TRACON for the
97: first 3 semesters of 2006.
98:
99: \subsection{Definitions}
100: The focus is made on the arrival process. Hence, in this article,
101: the term \emph{system} will refer to the TRACON area which is a 50
102: Nm radius circle centered at Oakland Airport. The \emph{inputs} to
103: the system are the planes entering the TRACON, intending to land at
104: SFO airport on 28L/R runways (figure \ref{fig:sfoconfig}), and the
105: \emph{outputs} are the planes actually land ing on 28L/R runways at
106: SFO airport. 28R and 28L runway are only 750ft apart which does not
107: allow simultaneous parallel automatic landings. Hence, if the
108: weather is not good enough for sight landing, only one runway can be
109: used.
110: \begin{figure}[htbp]
111: \begin{center}
112: \includegraphics[width = 0.5\textwidth]{sfomodif.eps}
113: \caption{San Francisco International Airport
114: diagram}\label{fig:sfoconfig}
115: \end{center}
116: \end{figure}
117:
118: Different parameters can be used to define the behavior of the
119: system. The evolution of the outputs (planes landing) can be
120: analyzed as a function of the inputs (planes entering the TRACON)
121: and a function of parameters such as the number of runways in use.
122: But we can also characterize it by the shape of planes trajectories
123: with in the same parameters. Each plane has a different track
124: leading to the runway. The analysis is made on 2D trajectories. For
125: the analysis purpose, tracks can be defined as \emph{straight} or
126: \emph{rerouted}. A \emph{straight} track corresponds to the
127: trajectory of a plane that could go directly from its entering point
128: in the TRACON to the runway threshold just following the way
129: points. A \emph{rerouted} track corresponds to the trajectory
130: of a plane that can not land directly and has to change its
131: trajectory. It does not follow the nominal track and has to make
132: a curved trajectory or even hippodromes (figure \ref{fig:DifferentKindOfRerouting}).
133:
134: \begin{figure}[htbp]
135: \begin{center}
136: \includegraphics[width = 0.5\textwidth]{differentKindOfRerouting.eps}
137: \caption{Different kind of rerouted trajectories}
138: \label{fig:DifferentKindOfRerouting}
139: \end{center}
140: \end{figure}
141:
142:
143: \subsection{Data selection}The analysis presented in this paper
144: focuses on the arrivals at San Francisco Airport. The presence of
145: several airports (Oakland Intl Ap, San Francisco Intl At, San Jose
146: Intl AP and many small airports) in a small perimeter such as San
147: Francisco Bay implies dedicated routes for landing and taking off
148: for each airport in the TRACON (figure \ref{fig:westConfiguration}).
149: Those routes decrease the degrees of freedom to reroute planes.
150: Moreover, for practical reasons such as fuel capacity and for
151: financial reasons, it is required to minimize the number of waiting
152: aircrafts and their waiting time. San Francisco International
153: Airport (figure \ref{fig:sfoconfig}) has four runways arranged in
154: two sets of parallel runways. The terrain
155: configuration (presence of residential areas at west, north and south
156: of the airport) is such that the ``West'' configuration is the
157: most used in order to reduce the noise over residential areas.
158: Actually, planes take off on 01L/R runways, departing over the bay
159: and land on 28L/R runways, maximizing the distance over the water in
160: order to reduce noise. We will use the planes landing in this
161: configuration (i.e on 28L and 28R runways).
162:
163: The routes for landings and takeoffs are well defined, hence we can
164: assume there is no correlation between them: the planes flying to
165: SFO do not interact with planes flying from or to another airport or
166: just transiting.
167:
168: The configuration of SFO airport is such that departures and
169: landings are on separate runways. The runways are crossing but we
170: will assume that taking off planes do not have influence on landing
171: planes.
172:
173: \begin{figure}[htbp]
174: \begin{center}
175: \includegraphics[width = 0.5\textwidth]{westconfiguration.eps}
176: \caption{West Configuration in San Francisco Bay}
177: \label{fig:westConfiguration}
178: \end{center}
179: \end{figure}
180:
181:
182: \subsection{Radar track analysis}
183: Each track is analyzed in order to determine whether it is a direct
184: track or a rerouted one. An algorithm has been developed to identify
185: rerouted tracks from direct tracks. The figure \ref{fig:tracks}
186: presents the results of this algorithm on a given day of arrivals at
187: SFO.\newline
188:
189: \begin{figure}[ht]
190: \centering \subfigure[\emph{Direct} tracks]{\label{fig:tracks-a}
191: \includegraphics[width = 0.4\textwidth]{StraightTracks.eps}}
192: \subfigure[\emph{Rerouted} tracks]{\label{fig:tracks-b}
193: \includegraphics[width = 0.4\textwidth]{ReroutedTracks.eps}}
194: \caption{SFO tracks analysis} \label{fig:tracks}
195: \end{figure}
196:
197: Figure \ref{fig:tracks-a} presents the tracks identified as direct.
198: Almost all the plotted tracks are pretty direct to the runway,
199: considering the routes and way points that the airplanes must
200: follow. Figure \ref{fig:tracks-b} presents the tracks identified as
201: rerouted. All those tracks contain a change of direction not
202: corresponding to the route the airplane should follow for a nominal
203: approach. The rerouting is done either by \emph{smooth} changes of
204: trajectory, e.g \emph{S-turns}, or either by important changes of
205: trajectory, such as hippodromes (Figure
206: \ref{fig:DifferentKindOfRerouting}).
207:
208:
209:
210: \subsection{Data Analysis}\label{subsection:dataanalysis} In this section, we investigate the
211: evolution of the system in two different ways. One is time-based
212: analysis and the other is aircraft-based.
213:
214: \subsubsection{Time based methodology} In this
215: methodology, we divide the time in periods of fixed length during
216: which we will count:
217: \begin{itemize}
218: \item The number of planes present in the system which corresponds to
219: the number of planes that are, at some point of this period, in the
220: TRACON area (figure \ref{fig:timebasedanalysis}).
221: \item The number of planes entering the system during this period.
222: \item The number of planes landing during this period.
223: \item The number of planes rerouted during that period.
224: \end{itemize}
225: It is to note that a same plane can be counted as present and
226: rerouted on several time periods but enters or lands only once.
227:
228: \subsubsection{Time based evolution of the system} Figure
229: (\ref{fig:dailyAnalysis}) presents the results of the time based
230: analysis. The time is on the abscises axis. The positive ordinates
231: axis is a number of flight while the negative ordinates axis
232: represents the proportion of the total number of planes in the
233: system that is rerouted, scaled from 1 to 10. Each bar stands for a
234: 15 minutes time period. The light blue bars represent the total
235: number of aircrafts, the dark blue bars represent the number of
236: planes entering the system, the red bars represent the number of
237: rerouted planes. Yellow indicates that only one runway is in use
238: while green stands for 2 runways.
239:
240: \begin{figure}[ht]
241: \begin{center}
242: \includegraphics[width = 0.6\textwidth, angle = -90]{RerouteAnalysis.eps}
243: \caption{Arrivals analysis February the 12th
244: 2006}\label{fig:dailyAnalysis}
245: \end{center}
246: \end{figure}
247:
248: As shown in figure \ref{fig:dailyAnalysis}, in case of one runway in
249: use, the number of rerouted aircrafts increases with the number of
250: planes in the system. As soon as the second runway opens, the number
251: of rerouted aircrafts drops. In the two runways configuration, even
252: a pick of incoming aircrafts or a large number of aircrafts present
253: in the TRACON does not increase significantly the number of rerouted
254: planes. This methodology is more developed and used in
255: \cite{McTDMABenefits}.
256:
257: \subsubsection{Aircraft based methodology}\label{sec:aircraft
258: methodology def} The evolution presented in this section is based on
259: a ``plane'' point of view. We look at the flow of entering
260: aircrafts, the flow of landing aircrafts, the number of planes
261: present in the system, the number of planes rerouted. Those terms
262: need to be well defined. To calculate flows, a time period $T$ is
263: required. The term ``for each aircraft'' means that the count has
264: been done for all the aircrafts contained in the data used. The
265: total amount of selected aircraft is around 95,000.
266:
267: \begin{itemize}
268: \item For each aircraft, the number of planes present in the system
269: corresponds to the number of airplanes that are, at some point, at
270: the same time in physically present the TRACON (figure
271: \ref{fig:aircraftbasedanalysis}).
272: \item For each aircraft, the flow of entering planes is equal to the number of planes
273: entering the TRACON during a period of length $T$ centered on the
274: entrance time of this plane in the TRACON, divided by the period T.
275: \item For each aircraft, the flow of landing planes is equal to the number of planes
276: landing at SFO airport during a period of length $T$ centered on the
277: landing time of this plane at SFO airport, divided by the period T.
278: \item For each aircraft, the number of rerouted planes that is equal to the number of
279: planes present in the system that have a \emph{rerouted} trajectory.
280: \item For each aircraft, the number of runway used corresponds to an
281: average on a period of length $T$ centered on the plane landing. If
282: one runway is used more than 75\% of the time during this period, we
283: will say than only one runway is in use.
284: \end{itemize}
285:
286: \begin{figure}[ht]
287: \centering \subfigure[Time-based methodology to count the number of
288: plane present in the TRACON]{\label{fig:timebasedanalysis}
289: \includegraphics[width = 0.7\textwidth]{timeBasedMeth.eps}}
290: \subfigure[Aicraft-based methodology to count the number of plane
291: present in the TRACON]{\label{fig:aircraftbasedanalysis}
292: \includegraphics[width = 0.7\textwidth]{aircraftBasedMeth.eps}}
293: \caption{Comparison of time-based methodology and aircraft based
294: methodology} \label{fig:methodologies}
295: \end{figure}
296:
297:
298: \subsubsection{Aircraft based analysis of landing and
299: rerouting}\label{subsubsection:aircraftbasedanalysis} This section
300: presents the results of the aircraft based methodology. The figure
301: \ref{fig:planeBasedAnalysis} plots the average flow of entering
302: aircrafts in the TRACON (green diamonds), the average flow of
303: landing aircrafts (red crosses) and its standard deviation (vertical
304: red lines), and the average number of rerouted aircrafts (blue
305: stars), as functions of the number of planes in the TRACON and of
306: the number of runway in use. The definitions of those term are given
307: in section \ref{sec:aircraft methodology def}. The blue line
308: represents the data frequency (i.e the number of time this
309: configuration happened). The period chosen is 15 minutes, to match
310: with the average time in the system which is 14 minutes 30 seconds.
311:
312: \begin{figure}[ht]
313: \begin{center}
314: \includegraphics[width = 0.5\textwidth]{planeBasedAnalysis.eps}
315: \caption{Aircraft based analysis of landing and rerouting}
316: \label{fig:planeBasedAnalysis}
317: \end{center}
318: \end{figure}
319:
320: \begin{description}
321: \item[One runway in use:] Up to 13 planes present in the TRACON, the flow
322: of entering and landing aircrafts are very close and proportional to
323: the number of planes in the system. The average flow of entering
324: aircrafts is slightly bigger than the average flow of landing
325: aircrafts. Over 14 planes present in the TRACON, the flow of landing
326: aircrafts remains at 31 planes an hour whatever the number of
327: entering aircrafts and the number of planes in the TRACON. The flow
328: of entering planes goes up to 35 planes per hour. Up to 12 planes
329: present in the TRACON, the number of rerouted planes increases
330: slowly. Over 13 planes present in the TRACON, this number increases
331: quickly and reaches 20 rerouted planes for 25 planes present in the
332: TRACON. Over 25 planes in the system, the number of rerouted planes
333: stabilizes around 15. The cases over 25 planes present in the TRACON
334: are isolated cases (happened once or twice in 9 month). The standard
335: deviation of the number of landing planes starts at 2.2 aircrafts
336: and is almost all the time around 4.5 aircrafts.
337:
338: \item[Two runways in use:] The flow of entering aircrafts and the
339: flow of landing aircrafts are almost the same up to 29 planes in the
340: system. The number of landing aircrafts increases almost
341: proportionally with the number of planes in the TRACON. Over this
342: number of planes in the TRACON, the data frequency is very low
343: (isolated cases) so the results are not very relevant. The number of
344: rerouted planes increases slowly and reaches at the maximum 10 for
345: 30 planes in the system.
346:
347: \end{description}
348:
349: \subsection{Airport characteristics} From the analysis of figures
350: (\ref{fig:dailyAnalysis}) and (\ref{fig:planeBasedAnalysis}) we can
351: show characteristics of San Francisco airport. First of all, the
352: analysis gives a runway capacity a little above 30 planes per hour.
353: This can be explained by the fact that the second runway can have
354: been used a few time during the period, increasing the number of
355: landing on one runway and by the fact that the capacity of 30
356: aircrafts is a recommandation. To reach the maximum capacity of
357: landings in one runway configuration, 15 planes in the system are
358: necessary. When this limit is reached, the number of rerouted planes
359: increases very fast. The runway occupation has to be optimized and
360: hence, planes have to be rerouted to arrive at the exact time they
361: have been assigned by the controllers. When the number of planes
362: entering the system is high, most of them have to be rerouted.
363:
364: In the two runway configuration, whatever the number of planes in
365: the system, the maximum capacity is hardly reached. Hence, the
366: number of rerouted planes is not high. This configuration can handle
367: a pick of arrivals or a large number of aircrafts in the TRACON
368: without having to reroute all the planes.
369:
370: \subsection{TRACON characteristics}
371: From the analysis of figures (\ref{fig:dailyAnalysis}) and
372: (\ref{fig:planeBasedAnalysis}), it is clear that over 15 aircrafts
373: in the system, the maximum landing capacity is reached. Letting more
374: planes entering the TRACON leads to reroute them. This increases the
375: TRACON controller workload, increase the density of aircrafts and
376: the airspace complexity \cite{keumjinLee} and decreases the degrees
377: of freedom for rerouting. An certain amount of planes is required to
378: reach the maximum capacity of landings, but over it, it decreases
379: the safety and does not improve performances.
380:
381: \subsection{Pros and cons of both methodologies}
382: \subsubsection{Time based methodology}
383:
384: \begin{description}
385: \item[Pros] This methodology is easy to understand and to interpret.
386: It is a usual way to analyze data. As shown in figure
387: \ref{fig:dailyAnalysis} the evolution of the number of rerouted
388: planes in time is very clear.\newline
389:
390: \item[Cons] The time period is given and the start of the first
391: period is arbitrary chosen: the first minute of the first day of
392: records. As it is arbitrary, some limit phenomenon can appear and
393: change in a slight way the values calculated. Figure
394: \ref{fig:methodologies} shows the impact of this arbitrary chosen
395: time period.
396: \end{description}
397:
398: \subsubsection{Aircraft based methodology}
399: \begin{description}
400: \item[Pros] this methodology gets rid of the problem of arbitrary chosen
401: time slot. The planes in the system correspond to the flying planes
402: that interact which each other: all the planes flying simultaneously
403: in the TRACON. This has a real meaning when talking about density or
404: complexity. This method grants a very easy comparison of the results
405: with a discrete time model. The analysis is exactly the same for all
406: the aircrafts.
407:
408: \item[Cons] For the number of planes in the system, we consider the
409: planes flying in the TRACON. This means we assume that the planes
410: already landed do not have any more incidence on the planes flying
411: and neither the planes outside the TRACON.
412: \end{description}
413:
414:
415:
416: \section{Model Calibration and Validation}
417: A model of the TRACON has been developed. One of the aim of
418: developing such a model is to see if it is possible to reproduce
419: with a simple model the phenomenon we observed on the real system.
420: This model presents similarities with the push-back and departure
421: process model in \cite{pujet}.
422:
423: \subsection{Description of the model} This model is a discrete time
424: input-output model. The model has been developed and used with
425: Matlab.
426: \begin{description}
427: \item[System:] The modeled system is the TRACON with a 2 parallels runways
428: airport. In a nominal configuration (i.e. the 2 runway are open and
429: simultaneous landing are possible), the maximum landing capacity is
430: 60 aircrafts per hour, 30 for each runway.
431: \item[Inputs:] The input to the system is the sequence of entering
432: aircrafts in the TRACON.
433: \item[Outputs:] The output to the system is the sequence of landing
434: aircrafts on the two runways.
435: \item[Time:] The time is divided in 30 seconds slots.
436: \item[Travel time:] When a plane enters the system, a
437: \emph{nominal} travel time is randomly drawn with a probability law
438: describing the configuration of the TRACON (i.e the different
439: possible routes leading aircrafts from the entrance to the runway
440: threshold). This nominal time corresponds to the time the plane
441: would spend in the TRACON if the runway is available without having
442: to wait.
443: \item[Runway opening:] To degrade the system, it is possible to
444: close one or both runways. If a runway is closed none aircraft can
445: land on it.
446: \item[Runway availability:] Once a plane lands, the runway is
447: unavailable for 2 minutes, which is 4 time slot.
448: \item[Runway attribution:] When a plane enters the system and is
449: attributed a nominal travel time, it requests the runway for the 2
450: minutes period corresponding to the end of its travel time. If the
451: runway is available, it can land and the runway becomes unavailable
452: for this period. If none runway is available for this period, the
453: plane has to queue.
454: \item[Queuing:] If the runway is not available, the plane will wait
455: until one runway is available. As soon as one runway is available,
456: the plane lands.
457: \end{description}
458:
459: \subsection{Model calibration} The probability law used to draw
460: nominal travel times is presented in figure
461: (\ref{fig:timeInTracon}). In blue, the travel time for straight
462: trajectories distribution extracted from the data and in red, the
463: probability law used for the model.
464:
465: \begin{figure}[htbp]
466: \begin{center}
467: \includegraphics[width = 0.3\textwidth]{timeInTracon.eps}
468: \caption{Distribution of travel time in the TRACON}
469: \label{fig:timeInTracon}
470: \end{center}
471: \end{figure}
472:
473:
474:
475: \subsection{Model validation} To validate the results of this model,
476: the analysis of the data presented in the previous chapter have been
477: used. From it, the entering sequence of aircrafts has been extracted
478: and used as input for the model. The number of open runways has been
479: extrapolated from the landing sequences in the data.
480:
481: Figure (\ref{fig:modelRealisation}) presents the evolution of the
482: modeled system for a two days simulation. The top graph presents the
483: entering sequence, the second graph presents the runway occupation:
484: a dot for each time slot when the runway is occupied. The first
485: runway which is always open is in red with value $1$ and the second
486: runway has the value $-1$ and is in blue. The green line presents
487: the period where the second runway is open. The third graph presents
488: the evolution of the system. Blue line represents the number of
489: aircrafts in the system and red line represents the number of
490: rerouted planes. The horizontal axis is the time, in 30 seconds time
491: slots. One day = 2880 slots.
492: \begin{figure}[htbp]
493: \begin{center}
494: \includegraphics[width = 0.7\textwidth]{modelRealisation.eps}
495: \caption{Evolution of the modeled system for a two days simulation.}
496: \label{fig:modelRealisation}
497: \end{center}
498: \end{figure}
499:
500: For figures (\ref{fig:planeBasedAnalysisModel}) and
501: (\ref{fig:comparisonModelReal}), the model has run on 250 days,
502: corresponding to 67218 planes, 22428 landing in one runway
503: configuration and 44790 landing on two runways configuration.
504:
505: Figure (\ref{fig:planeBasedAnalysisModel}) presents for the model
506: exactly the same analysis than section
507: \ref{section:dataPresentation}.\ref{subsection:dataanalysis}.\ref{subsubsection:aircraftbasedanalysis}
508: . The figure plots the average flow of entering aircrafts in the
509: TRACON (green diamonds), the average flow of landing aircrafts (red
510: crosses) and its standard deviation (vertical red lines), and the
511: average number of rerouted aircrafts (blue stars), as functions of
512: the number of planes in the TRACON and of the number of runway in
513: use. The definitions of those term are given in section
514: \ref{sec:aircraft methodology def}. The blue line represents the
515: data frequency. For the model, the period chosen is 16 minutes to
516: catch the real flow of this discrete model.
517:
518: \begin{figure}[ht]
519: \begin{center}
520: \includegraphics[width = 0.5\textwidth]{planeBasedAnalysisModel.eps}
521: \caption{Aircraft based analysis of landing and rerouting on the
522: model} \label{fig:planeBasedAnalysisModel}
523: \end{center}
524: \end{figure}
525:
526: To compare the results, figure (\ref{fig:comparisonModelReal})
527: presents the comparison between the model and the real system.
528: Figures (\ref{fig:comparisonLanding1}) and
529: (\ref{fig:comparisonLanding2}) present the comparison of the landing
530: flow between the model and the real system. The blue line represents
531: the average flow of landing aircraft in function of the number of
532: planes in the system for the model and the red line is for the real
533: system. The vertical lines represent the standard deviation. Figures
534: (\ref{fig:comparisonReroute1}) and (\ref{fig:comparisonReroute2})
535: present the comparison of the landing flow between the model and the
536: real system. The blue line represents the average flow of landing
537: aircraft in function of the number of planes in the system for the
538: model and the red line is for the real system. The vertical lines
539: represent the standard deviation.
540:
541: \begin{figure}[ht]
542: \centering \subfigure[Flow of landing aircrafts, 1
543: runway]{\label{fig:comparisonLanding1}
544: \includegraphics[height = 0.4\textwidth, width=0.4\textwidth]{comparisonFlowLanding1runway.eps}}
545: \subfigure[Flow of landing aircrafts, 2
546: runways]{\label{fig:comparisonLanding2}
547: \includegraphics[height = 0.4\textwidth, width=0.4\textwidth]{comparisonFlowLanding2runways.eps}}
548: \subfigure[Number of rerouted aircrafts, 1 runway ]{\label{fig:comparisonReroute1}
549: \includegraphics[height = 0.4\textwidth, width=0.4\textwidth]{comparisonRerouted1runway.eps}}
550: \subfigure[Number of rerouted aircrafts, 2 runways
551: ]{\label{fig:comparisonReroute2}
552: \includegraphics[height = 0.4\textwidth, width=0.4\textwidth]{comparisonRerouted2runways.eps}}
553: \caption{Comparison of landing flow and rerouting between the model
554: and the real system} \label{fig:comparisonModelReal}
555: \end{figure}
556:
557:
558: As shown in figures (\ref{fig:comparisonLanding1}) and
559: (\ref{fig:comparisonLanding2}), the behavior of the model concerning
560: the flow of landing aircrafts given by the model is very close to
561: the real system. When there is one runway in use, the model predicts
562: a flow a little lower than in the real system. For high numbers of
563: planes in the system, the average flow is 30 landings/hour which
564: corresponds to the specification. Notice that the standard deviation
565: of the model and the real system are always very close. When there
566: are two runways in use, up to 25 planes in the system, the model
567: predicts exactly the same behavior than the real system. Over 25,
568: there are some discrepancies but the number of realization is very
569: small. \newline
570:
571: On one runway configuration, figure (\ref{fig:comparisonReroute1})
572: shows that the predicted number of rerouted plane is pretty accurate
573: up to 10 planes in the system. Between 10 and 22 planes in the
574: system, prediction are slightly higher than the real system. Over 22
575: planes in the system, the prediction keeps increasing at the same
576: rate while in the real system, it stabilizes. On the two runways
577: configuration, figure (\ref{fig:comparisonReroute2}) shows that the
578: prediction are accurate up to 20 planes in the system and over 20,
579: slightly higher.
580:
581:
582:
583:
584: \subsection{Analysis of the model results} The maximum number of
585: planes present in the system is greater in the real system than in
586: the model. An explanation could be that there is always one runway
587: open in the model contrary to the real system where sometimes, both
588: runway can be closed. On one runway configuration, the model
589: predicts an always increasing number of rerouted planes with the
590: number of planes in the system. This seems logical, while in the
591: real system, it stops increasing.
592:
593:
594: \subsection{Perspective of evolution}
595: \begin{itemize}
596: \item Once a plane lands, the unavailability time of the runway is
597: always exactly two minutes. This does not take in account the size
598: of the aircraft. In a future model, the unavailability time could be
599: dependant of the size of the aircraft. The size of the aircraft
600: would be randomly drawn with a probability law based on the real
601: distribution of aircrafts. A distribution in three kind of aircrafts
602: is used in \cite{pujet}: medium jets, large jets and heavy jets.
603: \item The queuing is very deterministic. When a plane is queuing, as soon as the runway is
604: available, the plane will land. It does not take in account the fact
605: that planes have to keep moving and hence, if it is crowded, it can
606: be difficult to arrive with such a accuracy. An small uncertainty
607: could be added on the moment a plane queuing lands.
608: \item The model does not include different routes in the TRACON.
609: The time spend in the TRACON is randomly drawn with a certain
610: probability law. This could be more representative if the entering
611: sequence would include several entry points corresponding to
612: different routes with different travel time probability law
613: different for each route. Some route could be longer or more
614: frequent.
615: \end{itemize}
616:
617:
618: \section{Applications}
619: \subsection{Graceful degradation}
620: An application for the model presented above is graceful degradation
621: for air traffic control systems. To define \emph{graceful
622: degradation}, we introduce two systems (1) a nominal system (2) a
623: degraded system. \emph{Graceful degradation} happens when the
624: transition from the nominal system to the degraded system is always
625: smooth and with no event.
626:
627: Degradation can occur for many reasons such as bad weather,
628: breakdown of control systems (e.g radar) or communication systems...
629: In our case, the nominal system is given by the TRACON, San
630: Francisco airport with the two runways in use and the air traffic
631: controllers managing the plane arrivals. The degraded system is
632: still the TRACON but with only one or zero runway in use, and/or the
633: controllers not able to manage the system for some reason. This
634: degraded system must be sustainable. The current way to get a
635: \emph{graceful degradation} of this system is to reroute planes (i.e
636: to modify their trajectories so that they can safely wait before
637: landing) or even some ground slops for short haul traffic (e.g
638: United shuttle from LAX to SFO). The interesting fact about San
639: Francisco airport is that this phenomenon is tightly related to the
640: weather. As the two runways are very close, as soon as the
641: visibility is bad, which occurs very often due to the fog in San
642: Francisco, it is impossible to use them at the same time. The system
643: is hence very often degraded. The complexity \cite{keumjinLee} of
644: the system must remain under a certain level so that if a
645: degradation is to happen, the safety is still assured. A way to
646: reduce complexity is to limit the number of aircrafts present in the
647: system.
648:
649:
650: \subsection{Decongestion of the TRACON}
651: A study about the impact of the landing demand on landing capacity
652: is done in \cite{McTDMABenefits} for Newark Airport. This study
653: shows that when demand is higher than capacity of the runway, it
654: results in a drop of runway capacity. Plausible factors can be the
655: application of traffic flow restrictions and holding. Other factors
656: may include a high workload of air traffic controllers and airspace
657: complexity constraints.
658: \newline To see the impact of a congested TRACON, arrivals
659: simulation have been made using San Francisco demand sequence. The
660: TRACON has been given a limit capacity : no plane can enter it if
661: this limit is reached. As soon as a plane lands, one plane waiting
662: out of the TRACON can enter. \emph{Waiting} time refers to the time
663: the plane were asked to wait outside the TRACON and \emph{rerouting}
664: time corresponds to the time the planes were asked to wait for the
665: runway to be available to land. The queue outside the TRACON is a
666: First-In First-Out queue. The simulations cover 243 days and 66,067
667: aircrafts.
668:
669: Figure \ref{fig:delayComparison} presents the results of simulation
670: for TRACON limit capacity varying from 6 to 12 aircrafts. Black bars
671: stand for rerouting values and white bars for waiting waiting.
672: Figure \ref{fig:delayComparisonMean} presents the average delay and
673: figure presents the percentage of delayed aircrafts
674: \ref{fig:delayomparisonPercent}.
675:
676: \begin{figure}[ht]
677: \centering
678: \subfigure[Mean rerouting and waiting time]
679: {\label{fig:delayComparisonMean}
680: \includegraphics[height = 0.4\textwidth, width=0.4\textwidth]{delay1.eps}}
681: \subfigure[Percentage of delayed flights]
682: {\label{fig:delayomparisonPercent}
683: \includegraphics[height = 0.4\textwidth, width=0.4\textwidth]{delay2.eps}}
684: \caption{Simulated delays function of the number of allowed
685: planes in the TRACON }\label{fig:delayComparison}
686: \end{figure}
687:
688: A too low TRACON limit capacity has two effects. First, the number
689: of planes waiting outside the TRACON is very high and second, the
690: runway capacity will not be reached which results in an inefficient
691: configuration. Then, when the TRACON capacity increases, the mean
692: rerouting time decreases and stabilizes while the percentage of
693: delayed flight slightly decreases. From 9 to 12 planes in the
694: TRACON, the average rerouting time is almost the same, but, as the
695: number aircrafts in the TRACON is smaller, the complexity of the
696: system is smaller. The average rerouting time does not change much
697: because the runway maximum capacity is reached and planes have to
698: queue anyway. As shown in the first study, over a certain number of
699: planes in the TRACON, the number of rerouted aircrafts increases
700: very quickly. If the number of simultaneous rerouted planes is high,
701: the controller workload increases. By reducing this workload, the
702: controller could focus on fewer aircrafts and optimize better the
703: trajectories which would result in an increase of the runway
704: capacity. Moreover, the higher the number of present planes in a
705: given area is, the less is the flexibility for an efficient
706: rerouting process. TRACON are very congested area and limiting the
707: number of aircraft present in those areas would increase the safety
708: without generating extra delays.
709:
710:
711: \section{Conclusion}
712: The analysis of San Francisco bay's TRACON records shows the landing
713: capacity of the airport and the influence of the number of plane
714: present simultaneously in the TRACON on the rerouting process and
715: landing capacity. It is not useful to allow a large number of
716: aircrafts simultaneously in the TRACON to increase performances. A
717: validated TRACON input-output model is proposed, calibrated and
718: validated. It is used to show that limiting the number of aircrafts
719: in the TRACON does not increase the average delay, reduces the the
720: controller workload and increases safety.
721:
722:
723: \section*{Acknowledgement}
724: \addcontentsline{toc}{chapter}{Acknowledgement}
725: This research was supported by THALES. The author wish to thank Bert
726: Ganoug from the Aircraft Noise Abatement Office of San Francisco
727: International Airport pleasantly for providing the records of the
728: TRACON of San Francisco bay.
729:
730: \begin{thebibliography}{15}% maximum number of references (for label width)
731: % Bibliography file
732: % Maxime Gariel
733:
734: \bibitem{CTAS}
735: NASA, \it{Center TRACON Automation System}, http://www.ctas.arc.nasa.gov/project\_description/index.html\#overview.
736:
737: \bibitem{GDP}
738: Michael O. Bally and Guglielmo Lulli, \it{Ground Delay Programs: Optimizing over the Included Flight Set Based on
739: Distance}, 2001.
740:
741: \bibitem{TMA}
742: NASA, \it{Center TRACON Automation System, Traffic Management Advisor}, http://www.ctas.arc.nasa.gov/project\_des- cription/tma.html.
743:
744:
745: \bibitem{McTMA}
746: NASA, \it{Center TRACON Automation System, Multi-Center Traffic Management Advisor},
747: http://www.ctas.arc.nasa.gov/project\_description/mctma.html.
748:
749:
750:
751: \bibitem{McTMAresults}
752: Todd C. Farley, Steven J. Landry, Ty Hoang, Monicarol Nickelson, Kerry M. Levin, Dr. Dennis Rowe and Dr. Jerry D. Welch,
753: \it{Multi-Center Traffic Management Advisor: Operational Test Results},
754: AIAA 5th Aviation, Technology, Integration, and Operations Conference (ATIO), 2005.
755:
756: \bibitem{FAST}
757: NASA, \it{Center TRACON Automation System, Final Approach Spacing Tool},
758: http://http://www.ctas.arc.nasa.gov/project\_description/fast.html.
759:
760: \bibitem{Ioannis}
761: Ioannis Anagnostakis, \it{A Multi-Objective, Decomposition-Based Design
762: Methodology and its Application to Runway Operations Planning}, PhD
763: thesis, September 2004
764:
765: \bibitem{pujet}
766: Nicolas Pujet, Bernard Delcaire, Eric Feron, \it{Input-Output Modeling and Control of the Departure Process of Congested Airports},
767: 1999.
768:
769:
770:
771: \bibitem{McTDMABenefits}
772: Husni Idris and Antony Evans,
773: \it{Benefits assessment of multi-center traffic management advisor for Philadelphia and New York}, AIAA Guidance, Navigation, and Control Conference and Exhibit, 2003.
774:
775: \bibitem{PHLtimebasedmetering}
776: Todd Farley, John D. Foster, Ty Hoang and Katharine K. Lee,
777: {A time-based approach to metering arrival traffic to Philadelphia},
778: AIAA 2001-5241, 2001.
779:
780: \bibitem{keumjinLee}
781: Keumjin Lee,Eric Feron and Amy Pritchett, \it{Air traffic complexity: An Input-Output approach},
782: American Control Conference, New York 2007. To be published, 2007.
783:
784:
785: \end{thebibliography}
786:
787: \end{document}
788:
789: % - Release $Name: $ -
790: