Winter 2015 | VOLUME 57, NO. 4
LEARNING AIRCRAFT BEHAVIOR from Real Air Traffic
Plus • Standards for Data Quality Assurance in ATM Modernization Initiatives • Voluntary Safety Reporting in the FAA’s Air Traffic Organization • What’s Next for Air Traffic Control Training in the FAA?
Innovative technical services & solutions for complicated environments • • • • • • •
Air Traffic Control Training & Operational Support Aviation English Training Operational & Quality Assurance Assessments Airspace Modeling Global Air Traffic Management Safety Management ICAO Recognized Distributive Training Center
ISO 9001:2008 Certified Company ICAO Trainair PLUS Member
www.washcg.com
Winter 2015 | Vol. 57, No. 4
Contents
ATCA members and subscribers have access to the online edition of The Journal of Air Traffic Control. Visit lesterfiles.com/ pubs/ATCA. Password: ATCAPubs (case sensitive).
Published for:
andrea crisante/Shutterstock.com
Air Traffic Control Association 1101 King Street, Suite 300 Alexandria, VA 22314 Phone: 703-299-2430 Fax: 703-299-2437 info@atca.org www.atca.org Published by:
140 Broadway, 46th Floor New York, NY 10005 Toll-free phone: 866-953-2189 Toll-free fax: 877-565-8557 www.lesterpublications.com President, Jeff Lester Vice-President & Publisher, Sean Davis EDITORIAL Editorial Director, Jill Harris Editorial Assistant, Andrew Harris DESIGN & LAYOUT Art Director, Myles O’Reilly Senior Graphic Designer, John Lyttle Graphic Designer, Crystal Carrette Graphic Designer, Jessica Landry Graphic Designer, Gayl Punzalan ADVERTISING
Articles 07 Controller Productivity by the Numbers
A Breakdown of Air Traffic Controller Operational Performance Both Domestically and Abroad By Frank Frisbie and James H. Cistone
10 Learning Aircraft Behavior from Real Air Traffic
By Arcady Rantrua, Eric Masesen, Sebastien Chabrier, and Marie-Pierre Gleizes
16 Standards for Data Quality Assurance in ATM Modernization Initiatives
By Aleksandar Balaban and Charles Chen
24 Who’s in Control?
Book Leader, Quinn Bogusky | 888-953-2198 Louise Peterson | 866-953-2183 Colleen McDonald | 888-953-2194
Securing Commercial Unmanned Aerial Systems Command and Control – A Methodology and Way Ahead
By Rusty Baldwin, Terry Hofecker, and Greg Carter
DISTRIBUTION
32 Voluntary Safety Reporting in the FAA’s Air Traffic Organization
Nikki Manalo | 866-953-2189
© 2015 Air Traffic Control Association, Inc. All rights reserved. The contents of this publication may not be reproduced by any means, in whole or in part, without the prior written consent of ATCA. Disclaimer: The opinions expressed by the authors of the editorial articles contained in this publication are those of the respective authors and do not necessarily represent the opinion of ATCA. Printed in Canada. Please recycle where facilities exist.
Cover image: tigristiara/Shutterstock.com
Air Traffic Safety Action Program
By Ken Myers, Terry M. Biggio, Steve Hansen, and Steve McMahon
38 Paradigm Changes Related to TSAS
Viewed Through the Perspective of the FAA/NASA Operational Integration Assessment
By Kevin E. Witzberger
48 What’s Next for Air Traffic Control Training in the FAA?
Communicating Between Generations By Giora Hadar
Departments 3
From the President
5
From the Editor’s Desk
55
Directory of Member Organizations
The Journal of Air Traffic Control
1
FROM THE PRESIDENT
By Peter F. Dumont, President & CEO, ATCA
2015 was a Great Year – Let’s Do it Again!
A
s I look back on the year that included the 60th anniversary of ATCA’s Annual Conference and Exposition, and as we are closing the books on 2015, I have to tell you I thoroughly enjoyed this past year. Now, I will admit, looking back at the highlights, my new grandson makes the top of the list, but many things we did here at ATCA are a close second. Last March, we completed our third World ATM Congress with the Civil Air Navigation Services Organisation (CANSO), and it was bigger and better than the year before. We had almost 7,000 attendees and 203 exhibitors. One attendee who had not previously been at World ATM said it was like going to an air traffic control super market, with every piece of equipment you could want. ATCA’s partnership with CANSO has been so successful that the CANSO Board of Directors moved its fall meeting to coincide with our ATCA Annual in Washington, D.C. We are currently in discussions to explore additional ways ATCA and CANSO can collaborate. ATCA has also worked hard to become true partners with the Federal Aviation Administration (FAA). Our relationship was fine before, but we knew that we could work together even more effectively. Ed Bolton and Pam Whitley, assistant administrator and deputy assistant administrator for the Next Generation Air Transportation System (NextGen), respectively, reached out to us about ways we could better communicate the NextGen effort. The idea was to synchronize the FAA’s NextGen message with the contractors’ NextGen message. This coordination began in Atlantic City at our Technical Symposium and then blossomed into the full coordinated effort at the ATCA Annual. The FAA’s booth discussed NextGen in its phases of flight, and the exhibitors with NextGen products explained their role using the FAA NextGen structure. The effort was successful and will be the foundation to build an even better coordinated effort in 2016. The Technical Symposium in Atlantic City was also bigger and better than ever before. The FAA opened its doors the day before the symposium for “Tech Center Tuesday,” which was an incredible day to explore the many testing facilities at the Tech Center. For attendees, the FAA offered tours of their fire hangar, new machine for pavement testing, and their numerous labs simulating towers, Terminal Radar Approach Control (TRACON) Facilities, and every other FAA facility. The two-day conference focused on technical advancements in air traffic, commercial space, and NextGen. We had great participation from the FAA throughout the event, including a “fireside chat” with Teri Bristol, Chief Operating Officer of the FAA’s Air Traffic Organization, and Ed Bolton fielding questions from me. Early in 2015, the FAA Program Management Organization (PMO) expressed an interest in getting some feedback from stakeholders on its procurement process. ATCA members joined us in a day-long, in-depth, facilitated discussion about the good, the bad, and the ugly parts of the FAA’s procurement process. The event
did not include FAA staff, so the dialogue was candid and resulted in a comprehensive paper for the FAA. I was very impressed with the time and effort that everyone gave to the event, and ATCA staff summarized the session. Jim Eck, vice president of the PMO, allowed us to interview him for a Journal article on FAA procurement issues, and Jim later organized and moderated a panel at ATCA’s Annual on FAA procurement. We will continue to work with the FAA on providing constructive input on ways we can all help improve the procurement process. This year brought a lot of talk about FAA reform, from panels at our conferences, FAA officials discussing broad general principles, and Congress providing outlines for reform. Here at ATCA, we decided to document some of that discussion and dedicated our summer issue of The Journal of Air Traffic Control to FAA reform ideas. The issue was so popular that we were asked to provide additional copies for other events as well as send it to a university to use in a business class. We expect the discussion for FAA reform will continue in 2016, and we will look forward to utilizing our events and our member expertise to educate and debate proposed solutions. Our ATCA Annual Conference was a great success as well and marked a significant milestone for the association. As I mentioned above, our increased participation from CANSO and the FAA made an already incredible event for networking and policy discussion even more productive and valuable for our members. So, as I turn the page to 2016, I feel so grateful for ATCA’s success – but I know that we could not do all this without our dedicated membership. ATCA members volunteer to organize our discussion panels, spend hours writing or editing articles about technical air traffic issues for our publications, sponsor our events such as the annual Glen A. Gilbert Memorial Award Dinner, serve on committees, and so much more. While I think about everything that ATCA has done this past year, I had to ask myself “Did ATCA promote the science of air traffic control?” because that is the core of what we are trying to do. And I believe the answer is “yes.” I look forward to ATCA’s Cyber Security Day in January and to our fourth World ATM Congress in March. And as we all look around at the state of the world and wonder what we are leaving to our children and grandchildren, I think safer and more efficient air travel is our little part. Safe travels.
Peter F. Dumont President & CEO, ATCA
The Journal of Air Traffic Control
3
CONFERENCES
SCHOLARSHIPS
PROFESSIONAL DEVELOPMENT
CAREER OPPORTUNITIES
ENGAGE, EVOLVE, EXCEL. Professional Women Controllers, Inc. (PWC) is a resource that provides support, training, encouragement, and camaraderie for all air traffic professionals. PWC advocates balancing work and family life, recruiting and retaining excellent employees, developing people, enjoying work, and maintaining a positive sense of community.
For more information, visit www.pwcinc.org
FROM THE EDITOR’S DESK
By Steve Carver Editor-in-Chief, The Journal of Air Traffic Control
Winter 2015 | Vol. 57, No. 4
Formed in 1956 as a non-profit, professional membership association, ATCA represents the interests of all professionals in the air traffic control industry. Dedicated to the advancement of professionalism and technology of air traffic control, ATCA has grown to represent several thousand individuals and organizations managing and providing ATC services and equipment around the world. Editor-in-Chief: Steve Carver Publisher: Lester Publications, LLC
Officers and Board of Directors Chairman, Neil Planzer Chairman-Elect, Charles Keegan President & CEO, Peter F. Dumont Treasurer, Rachel Jackson East Area Director, Susan Chodakewitz Pacific Area, Asia, Australia Director, Peter Fiegehen South Central Area Director, William Cotton Northeast Area Director, Mike Ball Southeast Area Director, Jack McAuley North Central Area Director, Bill Ellis West Area Director and Secretary, Chip Meserole Canada, Caribbean, Central and South America, Mexico Area Director, Rudy Kellar Europe, Africa, Middle East Area Director, Jonathan Astill Director at Large, Rick Day Director at Large, Vinny Cappezzuto Director at Large, Michael Headley
Staff Marion Brophy, Director, Communications Ken Carlisle, Director, Meetings and Expositions Theresa Clair, Associate Director, Meetings and Expositions Ashley Haskins, Office Manager Kristen Knott, Writer and Editor Christine Oster, Chief Financial Officer Paul Planzer, Manager, ATC Programs Rugger Smith, International Accounts Sandra Strickland, Events and Exhibits Coordinator Ashley Swearingen, Press and Marketing Manager Tim Wagner, Membership Manager
Rewarding Outstanding Work
S
ome say that ideas come from a series of events, experiences, or dreams. Whatever you believe about how ideas are formed, in order to propagate them, one has to share them. The Journal of Air Traffic Control provides a great forum for that. 2015 was a banner year for The Journal; each issue contained articles spanning a wide array of interesting and challenging topics, from NextGen to privatization. As is customary at the end of each year, the ATCA Publications Committee chose the top three Journal articles of the year. With so many incredible articles to choose from this year, it was not an easy task. These three Journal papers stood out as the most outstanding and were awarded first, second, and third place: 1. “Integrated Traffic Flow Management” (Q1, Spring 2015) by James Hayes, CSC 2. “Air Traffic Control Restructuring – The What and Why of it All” (Q2 Summer 2015), David Grizzle, Dazzle Partners 3. “Avoiding Clouds Associated with Core Engine Icing” (Q4, Winter 2014), Dr. Julie Haggerty, National Center for Atmospheric Research and Jennifer Black, National Center for Atmospheric Research During our recent 60th ATCA Annual Conference and Exposition, ATCA, for the first time, publicly acknowledged the winners and their employers on a billboard
Cylonphoto/Shutterstock.com
Air Traffic Control Association 1101 King Street, Suite 300 Alexandria, VA 22314 Phone: 703-299-2430 Fax: 703-299-2437 info@atca.org www.atca.org
prominently displayed in the main registration area and listed them in the November 4 edition of the on-site conference newspaper, ATCA Today. It is a great compliment to the writers for all of their hard work. As Editor-in-Chief of The Journal and a member of the ATCA Publications Committee, I want take this opportunity to thank not only the winners of the competition but all the writers and their employers who contribute their time and professional expertise to the writing of papers. Your year-round support has brought continued success to The Journal. As always, if you have an idea for a paper or want to join the committee, please contact Kristen Knott (kristen.knott@atca.org) or myself at tomsomag@aol.com.
Steve Carver, Editor-in-Chief
The Journal of Air Traffic Control (ISSN 0021-8650) is published quarterly by the Air Traffic Control Association, Inc. Periodical postage paid at Alexandria, VA and additional entries. EDITORIAL, SUBSCRIPTION & ADVERTISING OFFICES at ATCA Headquarters: 1101 King Street, Suite 300, Alexandria, Virginia 22314. Telephone: (703) 299-2430, Fax: (703) 299-2437, Email: info@atca.org, Website: www.atca.org. POSTMASTER: Send address changes to The Journal of Air Traffic Control, 1101 King Street, Suite 300, Alexandria, Virginia 22314. © Air Traffic Control Association, Inc., 2015 Membership in the Air Traffic Control Association including subscriptions to the Journal and ATCA Bulletin: Professional, $130 a year; Professional Military Senior Enlisted (E6–E9) Officer, $130 a year; Professional Military Junior Enlisted (E1–E5), $26 a year; Retired fee $60 a year applies to those who are ATCA Members at the time of retirement; Corporate Member, $500–5,000 a year, depending on category. Journal subscription rates to non-members: U.S., its territories, and possessions—$78 a year; other countries, including Canada and Mexico—$88 a year (via air mail). Back issue single copy $10, other countries, including Canada and Mexico, $15 (via air mail). Contributors express their personal points of view and opinions that are not necessarily those of their employers or the Air Traffic Control Association. Therefore The Journal of Air Traffic Control does not assume responsibility for statements made and opinions expressed. It does accept responsibility for giving contributors an opportunity to express such views and opinions. Articles may be edited as necessary without changing their meaning.
The Journal of Air Traffic Control
5
THE INTEGRATED TOWER SOLUTION THAT CHANGES AIR TRAFFIC MANAGEMENT Our NAVCANatm fully integrated suite of air traffic management tools is designed to innovate and simplify controller workstations by delivering mission critical air traffic control information safely and efficiently. Scalable and customizable, this advanced technology offers a complete automation platform to support airport, tower and terminal ATC functions. NAVCANsuite tower automation products provide controllers with an integrated platform for surface management, electronic flight data exchange, departure sequencing and scheduling, surveillance data integration, and enhanced data exchange with external stakeholders. Platform-independent and developed on an open architecture, this system is highly adaptable to any working environment. Experience a solution that is proven worldwide and that controllers and airport operators trust. A system that changes air traffic management. navcanatm.ca
NAVCANatm is exhibiting at the ATCA Annual Conference. Visit us at booth #429
CONTROLLER PRODUCTIVITY
Controller Productivity By the Numbers
A Breakdown of Air Traffic Controller Operational Performance Both Domestically and Abroad By Frank L. Frisbie, P.E., Double F Consulting, LLC; and James H. Cistone, PhD., Sullivan Aviation Services, LLC
U.S.
Europe
Air traffic controllers (2009)
15,770
16,900
Controlled flights (2010)
15.9M
9.4M
Operating cost per flight hour
$410
$723
Flights per controller/year
1,008
556
While the WSJ drew positive and similar comparisons, once broken down the contradictions become more obvious. Just to make the math easier, assume the U.S. controller workforce is 15,000. And, again, for simplicity sake, say that operating 24/7/365 (in three shifts around the clock) that there might be 3,000 controllers on the average shift. Next, assume that the number of controlled aircraft in the skies over the U.S. at any given busy time is 6,000. In this overly
simplistic example, each controller has to handle two aircraft over an eight-hour shift. If there are 5,000 controllers on the busy shift he or she only need to service 1.2 aircraft over the period. See Figure 1. This is an example of “thin slicing” the U.S. air traffic system or looking at it holistically. Another way to thin slice this is to look at the 8,760 hours in a year and using the WSJ report, out of 15.9M flights in the year you would derive an average of 1,815 flights per hour in the U.S. system. Once again, if there are 3000 controllers available then each has to deal with 0.6 planes per hour or around five aircraft per shift. This would be like saying, you drive your car 100 miles per day; take the 100 miles and divide by 24 hours, and you drive the car an average of 4.2 miles every hour. In reality, you drove your car for two hours, probably around 50 miles in each hour, and it sat idle for 22 hours. Complicating factors are supervisors, training time, vacations, labor agreements, sick leave, flow management, uneven traffic density, geographical areas, etc. The analogy nevertheless unmasks a kind of basic truth; that the problem with controller productivity is not the controllers or their lack of effort – it’s a systemic problem. However, the WSJ does not tell the whole story. Figure 1 shows that the U.S. and Europe areas of control are nearly the same, but
Elena Kalistratova/Shutterstock.com
T
he May 4, 2011 edition of the Wall Street Journal (WSJ), extracted from a EUROCONTROL/FAA report,1 compared the record of the U.S. air traffic system to Europe and documented the operational performance of each country’s air traffic controllers. WSJ quoted the following in support of their commentary2:
The Journal of Air Traffic Control
7
CONTROLLER PRODUCTIVITY
Calendar Year 2010
Difference
Europe
USA
11.5
10.4
Number of En Route Air Navigation Service Providers
38
1
Number of Air Traffic Controllers (ATCOs in Ops.)
16 700
14 600
= -13%
Total staff
57 000
35 200
= 38%
9.5
15.9
= +67%
Share of flights to/from top 34 airports
66%
63%
Share of General Aviation
4%
23%
= x 5.5
Flight hours controlled (million)
13.8
23.4
= +70%
Relative density (flight hours per km2)
1.2
2.2
= x1.8
557 NM
493 NM
=-11%
63
20
= -68%
>450
= 509
= +13%
>50
3
Eurocontrol
FAA/ATO
Geographic Area (million km2)
Controlled flights (IFR) (million)
Average length of flight (within respective airspace) Number of En Route centres Number of airports with ATC services Of which are slot controlled Source
US vs. Europe = -10%
Figure 1.
the number of aircraft in the U.S. is almost double, as are the aircraft hours flown. So, the relative density in the U.S. is about twice that of Europe, 2.2 vs. 1.2 (flight hours per km 2). Although, there are 63 Airport Consultant Councils (ACCs) in Europe versus 20 in the U.S. and slightly more controllers in Europe than the U.S. In that case, slightly more controllers are required in Europe to control about half the traffic density. On the other hand, there are more than twice as many ACCs in Europe as in the U.S., and a natural (yet unproven as of now) conclusion is that the increased coordination between ACCs could be a contribution to the lower productivity of the European controllers. If the above hypothesis holds true, then facility and sector consolidation is a likely candidate to increase controller productivity; fewer points for coordination between controllers and less handoffs all contribute to increased productivity. The European Area Control Center segmentation is largely political, but the U.S. Air Route Traffic Control Center (ARTCC) boundaries are largely historical and were originally set due to line of sight limitations in air to ground communications and radar coverage. Communication and surveillance limitations have now been overtaken by technology, so it is conceivable that a single controller, at a single location, could manage 8
Winter 2015
multiple flights from the departure gate to the arrival gate and never have to hand off. Coordination would be accomplished only when some trajectory contention occurs, and much of that contention can be resolved by the automation system (e.g., flight prioritization). So far we have neglected a very significant attribute of the U.S. system, its remarkable and magnificent safety record. Intelligent observers of the ATC scene recognize safety as the one criterion that must not be compromised in the name of efficiency or employee productivity. Nevertheless, we all should not be so besotted with U.S. success in maintaining safety that we close our eyes to opportunities to become more productive or efficient without compromise to safety. Are we so bound by tradition that we cannot envision a better way to enable the controller to be even more productive? That is exactly what the framers of the Next Generation Air Transportation System (NextGen) had in mind when they dubbed the undertaking as transformational. They recognized that undue reverence for the good things in the then current National Airspace System (NAS) would be a bar to move to a new paradigm and they sought to set that restriction aside. Unfortunately, that impediment is still very much with us in regard to controller productivity, albeit cloaked in the mantle of safety.
CONTROLLER PRODUCTIVITY
Sander van der Werf/Shutterstock.com
A renewed commitment to the possibility that there may well be a different way to safely organize and control air traffic is needed.
The good news is that we need not invent a new ATC/ATM system. The NextGen program is producing the NAS Enterprise Architecture (EA) and the work of the now dormant Joint Planning and Development Office (JPDO) has done much to lay out that blueprint. At the heart of that work is the Trajectory Based Operations (TBO) initiative, whereby an aircraft traverses the system along a route precisely defined by the user (and agreed upon by the control authority) in advance. This methodology has the potential to eradicate many of the jurisdictional impediments that dictate the distribution of controllers in the U.S. NAS of today (sectors, altitudes, facility location, “line of sight,” and communications/navigation/ surveillance). Carried to its limit, the aircraft could operate without any controller intervention at all. A renewed commitment to the possibility that there may well be a different way to safely organize and control air traffic is needed. Given a clean sheet (i.e., one that does not a priori dictate where controllers physically work or how/whether they exercise “control” over each aircraft), there is clearly the potential to increase productivity. The U.S. needs to re-open the possibility that one site (with a possible back-up) could be the control point for all high altitude Contiguous United States (CONUS) traffic. All the aircraft are on
instrument flight rules (IFR) flight plans, which are 4-dimensional trajectories (4DT) so they are in a TBO environment. Aided by an advanced automation platform, the “controllers” monitor the aircraft compliance and facilitate the handoffs to/from low altitude conventional control facilities. The potential in this concept comes from two dimensions: 1) it does not necessarily3 depend on new avionics on the aircraft and, 2) it can radically reduce the number of controllers needed to manage the en route airspace. Granted, the automation platform and the compliance monitoring and communications links that would derive from System Wide Information Management (SWIM) and networked surveillance are not yet available. Nevertheless, these pieces are in FAA plans today, but they are not to be arrayed in the manner suggested herein. This “tweak” in the NextGen plan could have significant payoff in facility and personnel costs while maintaining the safety record that is the gold standard. Endnotes
[1.] 2010 U.S./Europe Comparison of ATM-related Operational Performance Final Report – March 2012 [2.] See Figure 1 for full chart from the aforementioned source document. [3.] Depending on the resulting route separation better FMS may be required.
The Journal of Air Traffic Control
9
Learning AIRCRAFT BEHAVIOR From Real Air Traffic By Dr. Rusty Baldwin, CISSP Riverside Research, Dr. Terry Hofecker, Unmanned Science, Inc., and Mr. Greg Carter, Electronic Warfare Associates
W
reasons any machine learning method based on a complete knowledge of the environment is not applicable. With EVAA we present a learning algorithm able to use incomplete data using cooperative multi-agent systems1 to produce autonomous and self-adaptive behaviors for aircraft in a simulated environment. Through a large volume of real flight data we build a network of agents, each tasked to learn a piece of the aircraft’s behavContext and Problems ior. Those agents communicate with each other to build the global The process of creating a scenario for an air traffic generator (ATG) behavior that can be used later for a simulation. is often a tedious task. Many parameters have to be manually entered in a long, iterative process. Moreover, they heavily rely on the flight Learning on Real Flight Data plan to generate the trajectory of the plane. But, after taking off, the When we observe real aircrafts flying, they emit through their plane’s trajectory rapidly differs from its original flight plan. It may ADS-B transponder, at any time, a set of parameters. Those paramebe because of the weather, because the controller gave an ATC order ters match real percepts like latitude, longitude, or speed for example to separate aircrafts, or because he or she gave a clearance to take a and each of them has a value. more direct route because the traffic was light. For those reasons we Table 1 shows the exhaustive list of observable parameters in propose a new way of generating traffic based on behavioral learning. our system. Some of those values are static and cannot be changed The behavior of an aircraft is difficult to simulate because it’s during the simulation, others can and are marked with a “*” in the defined by many parameters related to the aircraft and to its current first column. The last column shows an example of correct value for environmental conditions. The information necessary to realisti- each parameter. (See table on page 12). cally simulate the flight of an aircraft is often either insufficient or While we observe the real traffic we can capture the value of unavailable. In addition, the flight plan might be changed during each of those parameters in a snapshot that we call a situation. An the flight. The whole flight can be full of changes: change of speed, aircraft will fly through many different situations. Each situation is change of altitude, and ATCO (Air Traffic Controller) orders are linked to its previous one (temporally speaking) creating multiple given and actions are taken. For any action of the plane there can be situation vectors. many reasons and we cannot discriminate between them. For these Figure 1 represents the results of the learning on one aircraft hat if we could observe the real world to teach our simulation how to work? There would be no need for physics computations and no need to describe what kind of entities exist in the world. Everything would be observed, learned, and made usable for simulation. This is the goal of EVAA.
10
Winter 2015
AIRCRAFT BEHAVIOR
The Adaptive Multi-Agent System (AMAS)2,3 technology provides a way to deal with unpredictable events (like ATCO tactical orders) that aircrafts encounter during their flight. Those events are the reason very few aircraft follow their original flight plan.
increases linearly with the number of aircrafts observed. It could be a problem but the merging of situations enables us to merge redundant information and increase the speed of the learning process. Adaptive Traffic Generation A traffic generator must be able to compute the location of an aircraft over time. With the situation vector mechanism we are able to compute a set of future locations for the aircraft: if the current location of the aircraft matches the initial situation of a vector it means that the terminal situation of this vector is a potential future situation. “Agentifying” the aircrafts will generate the traffic. These aircrafts start with a specific situation that can be the departure airport (defining latitude, longitude, and altitude) and ready to take off (speed is null, callsign is set), or at a specific 3D position as if it were already flying. From their starting situation a plane can find what its next action will be by comparing its current situation with the initial point of every situation vector in the area. All of those who are sufficiently close in an “Nth” dimension comparison (N being the number of parameters) are candidates and the best situation vector among the candidate is chosen. Now the plane knows exactly what it should do next: • It knows where it should go by looking at the latitude, longitude, and altitude of the terminal situation of the situation vector. • It knows at which speed it must go there by looking at the time difference between the initial and the terminal situation vector. The Journal of Air Traffic Control
Rasch/Shutterstock.com
where each situation encountered by the aircraft is linked to its following and so on until the last situation which gives us an “unary tree.” Nevertheless, it would not be realistic to hope to create a graph with every situation encountered by the plane. Many sections of “unary graphs” can be simplified by removing intermediary nodes if the changes described between the first node and the last node of the section is linear (like an aircraft moving in a straight line). The remaining situations are called situations of interest. Every observed aircraft gives us one “unary tree.” Since we observe multiple aircrafts, some nodes (or situation) can be merged. The merging process is based on the proximity between the situations, if the distance between nodes is below a given threshold they merge with each other. When all the relevant nodes are merged we are left with a directed graph able to guide an aircraft. The nodes in the final graph represent the points where the aircraft has acted. It’s close to the notion of navigation point and we could expect those points to match the waypoints of the flight plan but our results disprove this hypothesis by showing that many aircrafts take shortcuts multiple time during the flight. Hence, there is no perfect matching between waypoints and nodes. (See Figure 2 on page 12). The scalability and usability of this method entirely depend on learning. If learning is not doable on big samples of diverse data then it cannot simulate diverse data and EVAA is not a realistic simulator. The graph in Figure 2 shows that the number of situations of interest
11
AIRCRAFT BEHAVIOR
Situation 1
Situation vector 1
Situation 2
Situation vector N
Situation vector 3
Situation N
Figure 1. “Unary graph” – The results of observing one aircraft
Parameter
Example of Value
Time
26 Nov 2014 12:07:06
Callsign
AF263PE
Latitude
45.66
Longitude
-0.3073
Altitude
25700
Heading
130
Departure airport
BOD
Destination airport
ORY
Type of aircraft
A321
Registration number
393320 F-GMZA
Ground speed
425
Vertical speed
1664
Transponder ID
4e6f657
Squawk
1000
Radar code
F-LFCH2
Table 1. Observable parameters
agents are geographically located on the map. Any simulated aircraft start with an initial situation, which is used to find the first objective of this aircraft. An objective is a situation in which the aircraft “wants” to be. The first step is to find this objective. EVAA provides a way to send a message to any agent in a specific radius of a location. The aircraft sends an objective request, a message containing its current situation, to any situation vector in a radius (R) around itself. Each vector has to decide if its initial situation matches the current aircraft’s situation. This is done with Algorithm 1, which is able to compute a numerical distance between two situations. The scale function put the difference between p(initial) and p(current) on the same scale between 0 and 10. It is necessary because a difference of one unit of heading is not much whereas one unit of latitude/longitude is huge.
1000 800 Number of items
Change
Number of situation of interest Number of situation of interest after merge
600 400 200 0
chungking/Shutterstock.com
0 20 40 60 This method, applied on multiple aircrafts and on long recordNumber of aircraft observed ing of flight, gives us the basis for learning a realistic behavior. An agent is launched for every existing situation vector. Those Figure 2. Situations over number of aircrafts observed
12
Winter 2015
80
100
AIRCRAFT BEHAVIOR Then, if the distance is less than a defined threshold the vector decides that it matches the current situation and sends its initial and terminal situation to the aircraft. Then, the aircraft receives responses to its request and do a certain amount of checks and verifications on every potential vector. Those that don’t pass those tests are discarded. An example of sanity check is: if a vector advice to go from 0 to 36,000 ft in one second this message will be discarded. Also an aircraft will prefer to change it’s heading than its altitude unless it’s close to its arrival airport. Among the remaining vectors, the aircraft will choose the one with the smallest distance to its current situation as its new objective. Once the objective is reached the process starts again with the new current position. By following this process the aircraft is able to follow a realistic trajectory with realistic parameters in the virtual sky of EVAA without any human help whatsoever. The Adaptive Multi-Agent System (AMAS)2, 3 technology provides a way to deal with unpredictable events (like ATCO tactical orders) that aircrafts encounter during their flight. Those events are the reason very few aircraft follow their original flight plan. Using AMAS means that we have to follow a set of principles if we want to benefit from those advantages. • Agent should be autonomous and the network between them should be self-organizing. • Agents should base their decision only on local knowledge. • Agents should cooperate with each other. Not to a point where they would be altruistic but they should try to help their neighbors if it improves the local state.
This functionality is necessary because, in a controlled zone, every aircraft must follow its flight plan unless otherwise said by the air traffic controller. Moreover, any aircraft can be remotely piloted with high-level orders through a generic message-based API. Figure 3 shows an interface we developed to show the capabilities of EVAA. In (1) you can see the list of aircrafts. If one of them is selected then its information are displayed in (3). You can also search a plane using the auto-complete field in (2) and then send it a control order with the panel (4). Results In this section we will compare the results of our simulation with the reality between French national airports.
We saw that the learning phase uses the self-organization principle when situation vectors build their networks on the fly. We use the locality principle when an aircraft only asks its potential objective to the vectors in its neighborhood. The fact that situation vectors, when they receive an objective request, can judge themselves as non-pertinent (and do not send a response) shows cooperation. Supervised Traffic Generation Figure 4. Trajectories of real Figure 5. Trajectories of simulated When using traffic generation you might want the aircrafts to follow aircrafts aircrafts a specific route. In EVAA, aircrafts with their flight plan specified can Figure 4 shows a set of 50 real trajectories (here Toulouse to switch between adaptive mode and flight plan mode with a single click. Paris Orly) and Figure 5 shows a set of 50 simulated flight for the same ADEP/ADES air line. We can see that the trajectories are very similar in shape and that many different kinds of trajectory are available in the simulation providing diversity and realism. Figure 6 shows the altitude profile of 25 real aircrafts between Toulouse and Paris airports. We see that the aircrafts start by
Figure 3. Pseudo pilot interface
Figure 6. Altitude profile of real aircrafts over time
The Journal of Air Traffic Control
13
AIRCRAFT BEHAVIOR
climbing, reach their cruise level, and then descend to reach another flight level to finally land on ADES. Figure 7 shows the altitude profile of 100 EVAA ATG simulated flights. We can see the same succession of phases than in the real world with a little less variety.
form. The coefficient is the same for each route). Those calculations have been made on 500 flights (50 flights for each route). We compared multiple air routes and the fact that the box (difference between first quartile and third quartile) is smaller in simulation than in reality shows a lack of diversity in the flights produced by EVAA. Nevertheless, we can see that our simulation always respects the minimum and maximum boundaries of travel time. In most cases the statistical distributions of simulated flight are included into and statistically close from its real flight counterpart.
Conclusion EVAA is using machine learning, multi agent systems and real trajectories observation to generate behaviors of aircrafts. Since those behaviors are based on what happen in the real world, the resulting trajectories will be very realistic. It is also possible to simulate supervised aircrafts, which will follow their flight plan (or the orders of a pseudo-pilot). Figure 7. Altitude profile of simulated aircrafts over time This enables EVAA to be usable in many situations such as pilot and controller training, generating autonomous surrounding traffic Figure 8 is a box plot of the time it takes for an aircraft to fly generation, fully human-controlled traffic, or any combination you from its departure to its arrival (The time scale for simulated flight can imagine. has been multiplied by a coefficient to fix a problem with our plat-
IM_photo /Shutterstock.com
References
14
[1.] L Panait, S Luke “Cooperative Multi-Agent Learning : The State of the Art”, Autonomous Agents and Multi-Agent Systems 11.3, p. 387–434, 2005. [2.] G Di Marzo Serugendo, M Gleizes, A Karageorgos, “Self-Organisation and Emergence in MAS: An Overview”, Informatica, p. 45–54, 2006 [3.] Jean-Pierre GEORGÉ, Marie-Pierre GLEIZES, Pierre GLIZE (2003). “Conception of adaptive system with emergent functionality: The AMAS theory”
Figure 8. Statistical distribution of travel time for aircrafts on multiple air routes in reality and in simulation
Winter 2015
SAME SYSTEM
NEW LIFE
COST-EFFECTIVE SOLUTION TO ENHANCE ATC SYSTEM LONGEVITY Our Common Terminal Digitizer (CTD) is a cost-effective COTS solution that extends the service life of existing primary and secondary ATC radar systems. Selected by the FAA, this field proven solution digitizes analog data, improves target reporting, data fidelity and provides 6-level weather processing.
To learn more, visit us at ATCA exhibit #341. Š Telephonics
I www.telephonics.com
Standards for DATA QUALITY ASSURANCE in ATM Modernization Initiatives By Aleksandar Balaban, m-click.aero and Charles Chen, Skymantics, LLC
S
tandard ATM data formats for aeronautical, weather, and f light (AIXM, WXXM, & FIXM) are becoming more widely accepted in both the U.S. and EU ATM communities. This is evidenced in the FAA in future System Wide Information Management (SWIM) programs such as Common Support Services for Aeronautical Information Management (CSS-AIM) and Weather (CSS-Wx). In Europe, the European Aeronautical Information System (AIS) Database (EAD) and Airspace Database Repository (ADR) contain aeronautical data to be distributed via SWIM. As more and more data is being on-ramped via SWIM in the form of data publishers and web services, governance of this information becomes critical for managing the information content and quality. Especially in an operational context, the data must be validated as a precondition for safety, security, and efficiency of operations. Since data syntax can be verified and validated through XML schema validation, the issues regarding data quality centers around data content and usage. The validation of correct data sets is essential, especially when exchanged across Flight Information Region (FIR) boundaries. A method has been developed within Eurocontrol for the checking of data content and usage using a semantic vocabulary called Semantics Business Vocabulary and Rules (SBVR), which is an OMG standard (OMG, 2015). Using SBVR, a set of business rules can be defined to enable the validation of data content and ensure data quality within the ATM data models. Furthermore, the translation of SBVR to an ISO-standard, Schematron, enables the validation through existing XSLT processes (ISO, 2006). Joint research is being conducted by the FAA, Eurocontrol, and the Open Geospatial Consortium (OGC) within the OGC interoperability testbed to develop a standard process for automated data validation using current OGC standard services. Standards Lower Costs through Re-Use The standardization and adoption of new ATM data models is improving the way in which producers and consumers are sharing ATM information. Using XML, the producers of data are able to exchange information using System Wide Information Management (SWIM). While SWIM is rapidly decreasing the amount of time required to connect and acquire relevant data, data quality is still a concern. Currently, it is expected that a data producer verifies and validates the information before disseminating it to the general pub16
Winter 2015
lic. However, these checks are currently limited to syntax validation or contain a limited set of business rule validations specific to the airspace in which the data is generated. This may limit the usability of the data by certain consumers, especially those external to the specified airspace domain, even though they may be conducting operations that span across these airspaces. It can be considered to be the responsibility of SWIM to govern the validation of data for each SWIM information region as part of a larger global SWIM data exchange. Within the FAA today, producers are checked for data load, performance, and ability to maintain service level agreements. However, data verification is left to the associated program office supplying this information. For example, the FAA Federal NOTAM System and NOTAM Distribution Service (FNS-NDS) ensure that NOTAMS submitted via its service interfaces are checked via several automated business rules prior to distribution. This process of business rules enforcement is specific to the DNOTAM system, and while the architecture and process can be reused, the specific method may not be easily reusable by other data management systems. With the implementation of SWIM services such as CSS-AIM and CSS-Wx, use of a reusable validation method should be developed to reduce costs and streamline the processes. Methods already exist for conducting business rules validation functions. The Object Management Group (OMG) organization has standardized the use of Semantic Business Vocabulary and Rules (SBVR) for defining a taxonomy and repository of business rules. The International Organization for Standardization (ISO) has
DATA QUALITY
Since data syntax can be verified and validated through XML schema validation, the issues regarding data quality centers around data content and usage.
Semantic Validation Checks Content Based on Business Rules Syntactic validation is based on the structural data check as defined by an XML schema or DTD. This check neither includes “business rules” defined by the regional operation of the system using the data, nor it is capable to express some special application dependent constrains. Validation based on business rules means additional rules are specified by domain experts and using well understandable domain language in order to eliminate or reduce the lack of expressiveness in the data structure description taxonomies such as the XML schema (e.g. cardinalities or data value ranges). Validation is performed by a software component called validation engine. A full semantics-based approach to data validation is a complex but powerful method in which standard taxonomies of aeronautical
data types would be converted into equivalent semantic representation, specified through an equivalent taxonomy/ontology and processed together with accordingly specified business rules. The role of a validation engine would be performed by a computational component called a “reasoned,” responsible for automatic evaluation of data sets, meta-data information, and associated rules in order to identify possible lack of logical consistency in data sets. Business Rules Engine for Rules Validation In the case of data validation and business rules enforcement, these differ between two distinctive approaches, a procedural and a declarative one. In the procedural approach the encoding of validation rules using an adequate programming language is a step-by-step, iterative validation procedure for each assertion for every validation rule. The declarative approach assumes design of high-level “declarative” statements (i.e. business rules), which represent the data validity requirements or are used to describe the fulfillment of certain business rules. In the implementation of procedural enforcement, one could program validation classes with a function per assertion, specify rules containing assertions as higher level functions, and finally, compile and pack them into software libraries. As an output, a number of software modules and libraries would be created carrying different sets of rules. These libraries are then deployed into a validation/rule enforcement application server container. Although there are some benefits in using this method, such as better overall performances and more specificity in cases of very complex rules, the lack of flexibility and low level programmatic approach make this variant rather inappropriate for dynamically changing business driven data validation and business rule enforcement. Another major difficulty with this approach is that maintenance complexity requires well-written and well-maintained software documentation. In the declarative approach, statements are expressed using standardized, high-level description taxonomy (preferably constrained native human language such as SBVR) and stored and managed in a component called a business rule repository. The validation is performed by a dedicated component, responsible for obtaining, parsing, and applying these rules. The term, “declarative,” does not mean the opposite of a procedural approach. Instead, it allows domain experts to describe what the valid aeronautical data are, instead of specifying validation procedures. The Journal of Air Traffic Control
winui/Shutterstock.com
defined a standard for asserting validation rules called Schematron for XSLT. OGC has proven through a series of tasks in the OGC interoperability testbed, that service orchestration using the OGC Web Processing Service (WPS) 1.0 standard in SWIM can provide the necessary business processes for validation of SWIM data. ATM data formats were defined in order to standardize the way in which ATM data producers and consumers exchange information. Adopting standards provided by OMG, ISO, and OGC can reduce the costs of implementation for SWIM technical services such as data validation and improve workflows, ensure data quality, and increase safety.
17
DATA QUALITY
Figure 1. Declarative versus procedural approach (Rules, 2015)
Figures provided by authors
Figure 3. SBVR-to-Schematron validation flow
18
specifications the concept of validation service has partially been specified. In FAA SWIM, there is an infrastructure service (i.e. EMS) classified as one of “common services� containing validation properties, while in the SESAR SWIM similar service has been identified as one of major capabilities of the technical infrastructure. However, it remains unclear, how the validation shall be implemented, deployed, and performed considering that both approaches don’t provide exact service specification, capabilities, and operational guidelines. It could be assumed that the validation was mentioned just for the sake of completeness and is expected to be performed inside of SWIM compatible services, for example, as part of their service end-points. During the aviation thread developments of the OGC Interoperability Testbed-11, two tasks were defined to develop a solution for declarative validation based on standardized, high-level OGC Web Services can Provide Validation validation rules. The first task involved development of an aviation Services for SWIM In both USA and the Europe SWIM initiatives and corresponding profile for defining business rules using SBVR, and the second task The use of Schematron for XSLT is a concrete example how an XML document might be checked against predefined structural validity rules initially expressed in SBVR. Those rules described conditions beyond those stated in XML Schema, such as value ranges of properties, cardinalities of aeronautical entities, constraints in associations between entities and so on. The declarative approach is well suited for business rule enforcement because it provides better support for domain experts to specify high-level rules using a predefined taxonomy using SBVR. The benefit is that all rules are stored in a human readable format for better classification, versioning and monitoring. These rules are then executed by a dedicated software component called a business rules or validator engine using the concepts of Schematron.
Continued on page 20
Winter 2015