ECACNEWS - ARTIFICIAL INTELLIGENCE IN AVIATION: the future is now - #74 - ecac-ceac
←
→
Page content transcription
If your browser does not render page correctly, please read the page content below
ECACNEWS
WINTER 2021
#74
European Civil Aviation Conference Magazine
ARTIFICIAL INTELLIGENCE
IN AVIATION: the future is now
CONTENTS
1 EDITORIAL
Intelligent aviation through artificial
intelligence
Rannia Leontaridi
AIR TRAFFIC MANAGEMENT AND AIRPORTS
2 EUROCONTROL has adopted AI to support
aviation
Paul Bosman
6 Smart Digital Tower – transforming air traffic
control operations in Singapore
Kuah Kong Beng
9 Fostering artificial intelligence in ATM
Tanja Grobotek
12 Artificial intelligence and the next generation
of airport air traffic management
Andrew Taylor
15 Schiphol as a tech company
Martijn Schouten, Floris Hoogenboom,
Sebastiaan Grasdijk, Bas Cloin
AIRLINES AND AIRCRAFT
18 Airlines’ data sharing and artificial
intelligence adoption
Jean Ruiz Carpio, Jeroen Mulder
ECACNEWS
#74 – Winter 2021 21 Air France pursues continuous innovation
Publication Director and invests in Big Data solution to improve
Patricia Reverdy flight efficiency
Editorial Committee Laurent Lafontan
Patricia Reverdy, Simona Wist,
Gillian Caw 23 Next Generation Intelligent Cockpit
Editor Single-pilot operations’ potential to increase
Gillian Caw flight safety
gcaw@ecac-ceac.org
Dirk Kügler, Christian Niermann
Designer
Bernard Artal Graphisme NEWS FROM ECAC & JAA TO
Cover: © EvgeniyZimin, Deposit Photos
– Magicleaf, Deposit Photos 27 ECAC Spotlight
Ph: © CEAC
ECAC Behaviour Detection Study Group
ECAC News welcomes feedback Carmen Feijoo
and content ideas from interested
parties. 29 ECAC in brief
Subscription and distribution 35 News from the JAA TO
requests should be made to
communications@ecac-ceac.org
The opinions printed in ECAC
News are those of the authors
alone and do not necessarily
reflect the opinions of ECAC
or its Member States.
Follow ECAC on
or
Visit: www.ecac-ceac.org
Editorial
Intelligent aviation through
artificial intelligence
Rannia Leontaridi
Director General for Civil Aviation, Department for Transport,
United Kingdom, and ECAC Focal Point for Environmental matters
ECAC tois pleased to present the first edition of ECAC News dedicated
artificial intelligence (AI) in aviation. Readers will be intro-
duced to several key aspects, as well as to a series of accounts of successful
implementation of AI by air traffic management operators, airports, airlines,
and aircraft manufacturers.
The reason we chose to dedicate this edition to artificial intelligence was
that more and more data are digital nowadays, and as information technology
evolves continuously, the impacts on the aviation sector are potentially
far-reaching.
At a broader level, the concept of artificial intelligence can be defined as
human intelligence simulated by machines, especially computer systems,
capable of learning and reasoning like a human being.
But why do we need artificial intelligence in aviation? Benefits of
integrating artificial intelligence in the aviation industry include, for example:
deciphering patterns from previous experience, intelligent decision-making
processes, prescriptive and predictive analytics, forecasting, improving air
passengers’ experience, boosting productivity, etc.
We have already seen successful implementation of AI applications in air
traffic management (ATM) and air traffic control (ATC) operations, in airports,
U-space (including the use of intelligent drones) and avionics. Other examples
of the integration of AI in aviation include dynamic ticket pricing, flight
optimisation, real-time forecasts of flight delays or trajectories, crew scheduling,
and predictive maintenance.
More importantly, artificial intelligence has the potential to tackle the two
major challenges aviation faces today: reducing the environmental impact of
air traffic, and improving the ability of the network to scale capacity up or
down depending on the growth or decline of traffic.
Of course, the use of artificial intelligence in the aviation industry brings
a series of opportunities and challenges that can be further discussed and
addressed. It is an area where alignment between strategic leadership and
practical implementation will be especially important. This, combined with
constructive dialogue among stakeholders, can help us overcome the
challenges, and identify and mitigate the associated risks.
Finally, I would like to thank all the contributors to this edition of ECAC
News, and I invite you to read this magazine, hoping it will inspire ideas and
fruitful discussions on the most appropriate ways to integrate artificial
intelligence in the aviation sector.
ECAC NEWS #74 1
AIR TRAFFIC MANAGEMENT AND AIRPORTS
EUROCONTROL has adopted AI
to support aviation
Paul Bosman
Head of the Network Manager, EUROCONTROL
“If you read the newspapers, you might think that Europe has lost the race for artificial intelligence
(AI) and it’s going to be China or the United States. They’re going to dominate us but we do not
think it is necessarily the truth. Our destiny is in our own hands and we have many tools and ways
to work together to really move AI forward. Maybe the only limitation is what we have in our own
heads.” FLY AI webinar on partnership
The FLY AI – a joint
effort in AI for aviation
TpeanoTROL
move AI forward, EUROCON-
– together with the Euro-
Commission and a wide range
of partner organisations (airlines,
airports, air navigation service
providers, manufacturers, EU bodies,
military and staff associations (1)) –
jointly worked on finding ways to
demystify AI with the goal of accel-
erating its uptake in the aviation
sector, including air traffic manage-
ment (ATM).
tion community. The webinars high- bility, more data become avail-
The FLY AI report (2) that came lighted that: able, while companies and indi-
out of this cooperation provides an • AI is not science fiction. Aviation viduals keep control;
overview of the many ways that is currently using mainly data- • we must integrate AI-specific risk
artificial intelligence is already driven AI and is beginning to ex- mitigations in aviation system
applied and assesses its potential ploit symbolic AI; design, and also use AI to help
to transform aviation, notably in • in the foreseeable future, AI can detect cyber threats. We need to
reducing human workload, driving complement and augment human work on new AI applications for
the development of new ATM/ capabilities but will not replace cyber protection in aviation;
U-space services, or increasing the human in ATM; • EASA’s guidance for Level 1 ma-
safety and cyber resilience and in- • partnering in AI can really help chine learning applications struc-
creasing its sustainability. Its action grow your business, accelerate tures the way AI systems in
plan aims to accelerate the devel- your know-how, and help achieve aviation should be designed; and
opment of AI in European aviation successful certification of safety- • EUROCAE WG-114 is preparing
and ATM. This year’s FLY AI webi- critical systems in which AI is common standards for the devel-
nars (3) explored each action area, embedded; opment and certification of AI-
covering partnering, research and • with the advent of common Euro- based aeronautical safety-related
innovation, ATM use cases, safety/ pean data spaces focused on mo- products.
certification, data sharing, cyber
security and training/change man-
agement.
(1) EUROCONTROL, the European Commission, EASA, ACI EUROPE, Airbus, ASD, CANSO, Heathrow
Each webinar reached more Airport, Honeywell, IATA, IFATCA, IFATSEA, the SESAR JU, Thales, as well as military partners
than 4000 people on average, EDA and NATO.
(2) https://www.eurocontrol.int/publication/fly-ai-report
mainly in Europe, testifying to AI’s (3) The FLY AI webinars webpage (https://www.eurocontrol.int/fly-ai) provides access to all webinar
growing importance to the avia- recordings and their presentations.
2 ECAC NEWS #74
EUROCONTROL has adopted AI to support aviation
I AIR TRAFFIC MANAGEMENT AND AIRPORTS
EUROCONTROL
supporting aviation by
operating with AI
Rhasesponding to immediate opera-
tional needs, EUROCONTROL
developed a number of AI-based
applications – and there are more
in the pipeline. They will generate
significant improvements to the
European air traffic control network
through offering new digital ser-
vices that will be embedded in our
future integrated Network Manage-
ment programme (iNM) to the ben-
efit of operational stakeholders. Trajectory Prediction Improvement is in
In early 2019, the FLY AI group operation. What were the key steps to get
identified around 25 AI-based ap- there?
plications in European aviation and
ATM, with some already opera- It took a lot of time and effort to get people to accept intro-
tional (Heathrow Airport, Maas- ducing AI, particularly our safety department, as AI intro-
tricht Upper Area Control Centre, duces a new sort of unpredictability in the system. However,
Network Manager (NM)). Less than in the TPI case the worst thing that can happen is to end up
two years later, EUROCONTROL with the wrong flight prediction, exactly as with conventional
alone is developing over 30 appli- systems today. We are not introducing a new failure mode
cations. Some are already used op- here. In MUAC, the system is augmented by AI with humans
erationally in decision-making tools still at the centre of decision-making processes.
for the NM operational team and
their operational stakeholders (air-
ports, airlines and ANSPs). These
paramount importance. EUROCON-
rapid developments were possible Reducing flight TROL’s Curfew AI module (7) can de-
thanks to EUROCONTROL’s innova-
tion hub initiative and the availabil- delays and their cost tect such flights well before the
ity of EASA’s AI guidance material. restricted period commences and
to business provides improved predictions when
compared to those of the enhanced
Improving air traffic
management efficiency AI delays,
is also used to minimise flight
which impact airlines’
tactical flow management system.
Predictions are performed per air-
punctuality and the passenger ex- craft (i.e. tail number), and can be
perience (the combination of which queried at any time during the day,
TwayhetrolforMaastricht Upper Area Con-
Centre (MUAC) paved the
the operational use of AI at
is covered in Regulation (EC) No 261/
2004 on passengers’ rights), and
which is particularly useful when
tracking curfew-affected aircraft.
EUROCONTROL with its Trajectory may generate night curfews im- The MIRROR operational deci-
Prediction Improvement (TPI) (4). TPI posed at many European airports. sion support tool (8) was used as a
has been supporting Flow Manage- Curfews prohibit take-offs and sandbox during the development
ment Position operators for over landings during certain night-time of the Curfew AI module. It cur-
two years by trawling through the hours; a single minute of curfew in- rently provides NM operational
wealth of data that MUAC has col- fringement can lead to a diversion staff with a real-time view of each
lected, and applying AI algorithms of the flight. Diverting a long-haul aircraft’s flight legs, thus allowing
to support several key integrated flight can cost up to €68 000 (6), and identification of potential knock-on
Flow Managment Position (iFMP) (5) results in significant disruption to delays. The Curfew AI module will
functions to improve prediction of the airline and its passengers. be made available in the coming
flight routes, 4D trajectories, take- Developing effective methods to months to MIRROR’s several hun-
off times, sector skips, and even accurately identify flights at risk of dred industry users following an
hotspots of complex traffic. curfew infringement is therefore of update of the MIRROR interface.
(4) https://www.eurocontrol.int/publication/traffic-prediction-improvements-tpi-factsheet-and-technical-documentation
(5) https://www.eurocontrol.int/sites/default/files/2019-06/factsheet-ifmp-22022017.pdf
(6) https://www.eurocontrol.int/sites/default/files/2021-03/eurocontrol-standard-inputs-economic-analysis-ed-9.pdf
(7) http://www.atmseminar.org/seminarContent/seminar14/papers/ATM_Seminar_2021_paper_15.pdf
(8) https://www.eurocontrol.int/tool/mirror
ECAC NEWS #74 3
AIR TRAFFIC MANAGEMENT AND AIRPORTS I EUROCONTROL has adopted AI to support aviation
BERG), which is open to all network
operational stakeholders.
Going live soon…
Eimprove
arly in the flight planning pro-
cess, flights can be optimised to
network efficiency. PRE-
DICT AI improves optimisation and
prediction by offering multiple alter-
native routes tailored both to avia-
tion stakeholders’ business needs
and the Network Manager’s traffic
flow management criteria. PREDICT
AI retrieves the routes flown by any
given airspace user and makes sug-
gestions for better routings taking
due account of their business
needs as interpreted from historical
data. The model can run on any city
In addition, an application pro-
gramming interface running on
AI as cyber-protection pair and on any aircraft operator on
the network. As a result, the Net-
EUROCONTROL’s cloud infrastruc- assistant work Manager Operations Centre
ture will provide a broader access (NMOC) benefits from more reli-
to the Curfew AI module, in partic-
ular to airlines for integration into
their flight operational centres as a
A s a first innovative action on
cyber for AI, EUROCONTROL’s
European Air Traffic Management
able route predictions, thus in-
creasing network predictability
and stability, which in turn facili-
support tool. Computer Emergency Response tates the adjustment of capacity to
Team (EATM-CERT) has explored demand.
the way AI could increase the cyber
Reducing operational resilience of the aviation system,
Additionally, given that only
around 5% of NOTAMs are relevant
impact of ATFM delay notably for the rapid and reliable
to network operations, the NOTAM
detection of aviation-related docu-
uncertainty ment leaks. In a hybrid approach
list AI module will identity, extract,
and display the most relevant
based on symbolic and data-driven
NOTAMs at the top of the list, al-
T odemand
resolve imbalances between
and capacity, the Net-
AI, they pre-trained an off-the-shelf
“multilingual transformers model” to
adapt it specifically to aviation’s op-
lowing NMOC Operators to deal
with them quickly, freeing up their
work Manager imposes regulations, time for higher value tasks.
delaying flights on the ground. The erational, technical and multilin-
gual contexts (10). Now integrated in PREDICT AI and NOTAM list are
length of the ground delay assigned
EATM-CERT tools, the model can now fully developed, are aligned
to a regulated flight may change
identify if a leaked document is rel- with the EASA guidance on AI cer-
dynamically until departure. This
evant to aviation much more tification and thus ready for inte-
variability in the delay stems from
rapidly than data analysts could. gration in NM systems.
the mechanisms used by the com-
puter-assisted slot allocation system Significant hours of scarce human On the airport side, the D-1 AI-
resources are now being saved in based predictive model will pro-
(CASA) to manage the regulation slots.
addition to freeing staff from te- vide new enhanced predictions of
The FADE AI module (9) has dious and boring work. This deliv- airport regulations using informa-
been developed to reduce this un- ers significant cost and operational tion available prior to the day on
certainty and therefore improves benefits for the aviation commu- which they should apply (hence
airlines’ operations management nity. Cyber risk has increased signif- “D-1”). These are then integrated
throughout the day. Trained on his- icantly during the pandemic and, into NMOC’s morning briefings on
torical data, it is fully integrated in as “AI for cyber” is still a new area for the day of operations. D-1 AI uses
CASA and can predict the trend of aviation security experts, increasing information from airport events in
the delay with an accuracy of 75%. AI cyber collaboration and sharing NM’s “airport corner” and from
FADE reduces the prediction error lessons learned are paramount. NOTAMs, as well as weather and
by up to 63% when compared to This is a key objective for the Net- traffic forecasts to estimate the like-
the model without FADE. work Manager Cyber group (CY- lihood of a regulation being re-
quired at any of the network’s
busiest airports. It already predicts
(9) http://www.atmseminar.org/seminarContent/seminar14/papers/ATM_Seminar_2021_paper_14.pdf the hours when regulations are
(10) https://www.eurocontrol.int/event/ai-cyber-and-cyber-ai more likely - with 84% accuracy.
4 ECAC NEWS #74
EUROCONTROL has adopted AI to support aviation
I AIR TRAFFIC MANAGEMENT AND AIRPORTS
Moreover, it addresses concerns at their own airports. Its deploy- EUROCONTROL is moving for-
about “AI explainability” (11) by pro- ment on EUROCONTROL’s cloud is ward quickly with this wide range
viding the rationale behind the expected at the end of 2022, which of AI applications that are already
prediction, be it atmospheric phe- will accelerate its operational uptake. bringing operational benefits to
nomena, industrial action, works Finally, the mature AI-based the European Network. These are
in progress at the airport, over de- application COAST (12) allows ANSPs just the first steps and, together
mand, etc. to optimise the calibration of their with our FLY AI partners, we look
This information is shared with Final Approach Separation Delivery forward to moving faster to help
the rest of the functions in the European aviation recover in a sus-
tools for safe and efficient Time-
NMOC. The purpose of the predic- tainable and scalable way. ■
Based Separation (TBS) operations,
tion is NOT to take any un-coordi- using machine learning models
nated action but to proactively trained on historical local airport
contact the airport for considering traffic and MET data. COAST is a de-
in a collaborative way the best op- cision-support tool that helps air
tions. traffic controllers to organise the
Although it was initially planned spacing and separation of traffic.
to serve as a decision support tool COAST, therefore, is probably our
for the NMOC Airport Function, D-1 first human-AI collaboration appli-
AI will also be shared with our stake- cation and the only ATM use case
holders, to foresee the impact that supporting the EUROCAE standard
other airports’ daily situations have for AI in aviation.
(11) AI models currently used, essentially deep neural networks, are of such complexity that it is
practically impossible for their creator to understand their inner workings. Explainable artificial
intelligence is a set of processes and methods that allows human users to comprehend and
trust the results and output created by machine learning algorithms.
(12) https://www.eurocontrol.int/publication/eurocontrol-coast-calibration-optimised-approach-
spacing-tool-use-machine-learning
Paul Bosman has been with EUROCONTROL since 1992, working in many different technical and managerial
positions in various locations (R&D centre – France, seconded to Airbus, Brussels headquarters).
Currently, he is the head of the Network Manager / Infrastructure Division, planning, deploying and monitoring
European infrastructure digitising the SES European Sky through activities such as CNS, information management
and overall resilience (cyber, interference…).
He chairs the European Aviation/ATM – AI High-Level Group, which has delivered – amongst others – the FLY AI
report.
ECAC NEWS #74 5
AIR TRAFFIC MANAGEMENT AND AIRPORTS
Smart Digital Tower – transforming
air traffic control operations in
Singapore
Kuah Kong Beng
Director of Special Projects,
Civil Aviation Authority of Singapore
Introduction “ The Smart Digitalisation
Digital Tower
T he provision of aerodrome air
traffic services remotely without
a physical tower may seem novel to
prototype at A ir traffic controllers rely on the
out-the-window (OTW) view in
conventional towers to provide
some but in fact it has been a real- Changi Airport aerodrome air traffic services. The
ity for several years now. The first digitalisation of this view in a digi-
digital tower was approved for
provides real- tal tower environment, commonly
operations at Örnsköldsvik Airport time surveillance known as a video wall, is a paradigm
in Sweden in April 2015. shift for air traffic management.
While remote towers have been of Changi Airport’s The video wall allows the air traffic
implemented for smaller airports, controllers to select and integrate
there were none for high intensity
aerodrome critical data such as aircraft call
airports back in 2014. The Civil operations signs and other essential flight
Aviation Authority of Singapore information to be overlayed on
(CAAS) embarked on a research through more the screens. Consequentially, these
study to evaluate a digital tower data overlays will reduce the
concept for high intensity runway than 100 fixed- “heads down” time on multiple air
operations at Changi Airport. The traffic control (ATC) displays and
research found that there was
position cameras allow for easy flight identification
scope to develop a digital tower so- and pan-tilt- and enhanced situational aware-
lution for Changi Airport which ness, thereby improving safety as a
could reap numerous operational zoom (PTZ) result.
benefits. The development of such More importantly, digitalisa-
a prototype started in early 2018 cameras installed tion of the OTW view also allows
with NATS (Services) Ltd and across multiple the provision of aerodrome air traf-
Searidge Technologies. fic services from a remote location.
locations at the This mode of operation has already
been deployed in Germany; the in-
Smart Digital Tower airport. ” ternational airport at Saarbrücken
prototype has been remotely controlled from
and increase visibility during con- a Remote Tower Centre in Leipzig
ditions such as hazy, low light or more than 400km away. Integra-
Treal-time
he Smart Digital Tower proto-
type at Changi Airport provides
surveillance of Changi
night conditions. The PTZ cameras,
with the ability to deliver up to 30x
tion of two or more control tower
operations at different locations
Airport’s aerodrome operations optical zoom, mirror the function into a single integrated facility is a
through more than 100 fixed-posi- of binoculars in conventional con- possibility. Now, it is a function of
tion cameras and pan-tilt-zoom trol towers. getting the appropriate infrastruc-
(PTZ) cameras installed across mul- The Smart Digital Tower proto- ture in place. When implemented
tiple locations at the airport. The type operations room has three successfully, it is expected to en-
flexible and multiple camera place- controller working positions (CWPs) hance the coordination between
ments around the airport allow vi- and a video wall made up of large multiple control towers, reduce the
sual surveillance of the aerodrome monitors that provide a common complexity of operations and im-
areas that are located far away or situational awareness for the air prove overall safety and efficiency.
blocked by buildings when viewed traffic controllers at their respective Separately, CAAS is also explor-
from the physical control tower, CWPs. ing using speech recognition tech-
6 ECAC NEWS #74
Smart Digital Tower – transforming air traffic control operations in Singapore
I AIR TRAFFIC MANAGEMENT AND AIRPORTS
Smart Digital Tower – video wall with a runway view
nology to create a “digital assistant”
to automate processes and stream-
line operations. The aim is to re-
“ As one of the busiest airports in
place manual inputs, laborious the world under pre-COVID-19 traffic,
data entry and issuing of common
commands via speech recognition
air traffic controllers in Changi Airport
technology. For example, an air operate at a high-paced environment
traffic controller will be able to use
voice commands to provide clear- with approximately 380 000 annual
ances for flight altitudes and as-
signed headings, which will then movements. Usage of video analytics
be transcribed and entered auto- and decision support tools will help
matically into the air traffic man-
agement (ATM) systems. Automatic to alleviate the workload of the
transcription of conversations be-
tween pilots and air traffic con- air traffic controllers. ”
trollers not only allows for more
expeditious post operations inci- deployed across the aerodrome. intensity. As a safety measure, air
dent review but also provides These alerts on the video wall can traffic controllers will have to en-
ample data for greater insights into help to enhance the situational sure there is appropriate separa-
ATM operations. If the develop- awareness of air traffic controllers tion, and more communications
ment of the “Digital Assistant” and improve runway safety. On a with the pilots are also expected
proves to be successful, CAAS may similar note, PTZ cameras can also due to the low visibility conditions.
integrate this feature into the digi- be programmed to do a patterned To address the above challenges,
tal tower environment to augment sweep in the aerodrome to scan for CAAS is studying the feasibility of
air traffic controllers’ operations. objects of interests or specifically using camera surveillance sensor
selected by air traffic controllers to technology and artificial intelli-
track a particular aircraft, vehicle or gence to provide indication to air
Video analytics even humans in the aerodrome. traffic controllers when the tail of
applications These features can help to reduce the aircraft is clear of the runway.
cognitive workload for the air traf- This digital advisory tool will be es-
fic controllers and thus allow them pecially useful during periods of
A s one of the busiest airports in
the world under pre-COVID-19
traffic, air traffic controllers in Changi
to focus their mental resources on
other tasks at hand.
low visibility to advise air traffic
controllers on the runway clear-
Airport operate at a high-paced Another video analytics appli- ance status so that they can make
environment with approximately cation we can potentially leverage quicker and more informed deci-
380 000 annual movements. Usage on is in the indication of runway sions on whether to land or depart
of video analytics and decision sup- clearances. Currently, air traffic con- the next aircraft. If the develop-
port tools will help to alleviate the trollers rely on visual means to de- ment of this feature is successful,
workload of the air traffic con- termine whether an aircraft has it is expected to improve overall
trollers. One potential application vacated the runway prior to allow- safety in the following aspects.
that can be incorporated into the ing another arriving aircraft to land, Firstly, there will be reduction
digital tower environment is the or clearing a departing aircraft for in workload for both air traffic con-
detection of moving objects in the take-off. During inclement weather trollers and pilots in terms of ra-
airfield. This can be achieved via or periods of low visibility, this diotelephony (RTF) exchanges during
video processing of the images ob- could be challenging for air traffic periods of low visibility. Timely
tained from the numerous cameras controllers to operate at their usual issuance of landing/take-off clear-
ECAC NEWS #74 7
AIR TRAFFIC MANAGEMENT AND AIRPORTS I Smart Digital Tower – transforming air traffic control operations in Singapore
CAAS had conducted evalua-
tion on the Smart Digital Tower
prototype and the data showed
that the overall concept of a digital
tower was acceptable for opera-
tions. With the advanced features
and tools, air traffic controllers
were able to provide safe, orderly,
expeditious and effective air traffic
control at Changi Airport, albeit at
low traffic conditions due to
COVID-19. The key areas of evalua-
tion include operational safety, effi-
ciency and capacity, performance
Smart Digital Tower – evaluation of the prototype by air traffic controllers
and limitations, air traffic controllers’
ance is key for safe and smooth op- acceptance of the new concept of
erations, especially in a high inten-
Future plans operations, as well as an assessment
sity airport. With this feature, there of human-machine and human-
human interactions.
will be no need for the air traffic
controller to request the pilot to T echnologies advancement and
digitalisation are increasing at a Singapore has embarked on a
journey toward being a smart nation
report when they are clear of the rapid pace. The changes are seen to
runway. Similarly, the pilot will not be disrupting existing business op- since the initiative was announced
need to report to ATC when the air- erations and redefining the new in 2014 (1). CAAS is similarly on board
craft vacates the runway. This will normal in our society. While COVID- this journey of adopting smart
result in better use of airtime, 19 has decimated air travel, it is a technologies as it is important for
which includes lower probability window of opportunity for air nav- us to stay relevant in today’s fast-
of RTF congestion and timelier is- igation service providers (ANSPs) changing world. CAAS will be
suance of clearances to pilots. like CAAS to invest in research and building upon our past successes
Secondly, it is envisaged that development (R&D) projects that with the living lab and continuing
the information provided by this will enhance air traffic manage- our journey of leveraging on artifi-
tool may reduce the time required ment. The low air traffic demand cial intelligence. We will be evaluat-
for an air traffic controller to form a presents good opportunities to ing the steps required to progress
decision on whether to clear the from a prototype system to the
trial new technologies and allows
next arrival to land. Thus, it is ex- next stage whereby we can deploy
ANSPs to deploy air traffic controllers
pected that there will be a lower digital tower technologies in an
for R&D projects – as providing these
risk and probability of an arriving operational environment. This in-
resources for R&D projects is often
aircraft having to carry out missed cludes various activities such as de-
a challenge to most ANSPs when
approaches during periods of low veloping the relevant safety case,
airports are operating at pre-COVID- training plans as well as deploy-
visibility due to late issuance of 19 levels.
clearances. Increase in aircraft ment strategies.
For CAAS, the Smart Digital With the advancements in
movements and traffic can be pos-
Tower prototype is a living lab technology over the years, it is
sible without compromising safety
and can potentially lead to greater where relevant technological inno- timely for one to evaluate the latest
operational efficiency. vations will be incorporated and technology that could potentially
tested to continuously improve enhance safety and operational ef-
and fine-tune the digital tower ficiency, as well as provide better
system. The living lab allows us to service levels. Where possible, CAAS
determine the suitability and ro- will be harnessing smart technolo-
bustness of the new innovations gies that will give us the competi-
(1) https://www.smartnation.gov.sg/whats-
before commissioning the technol- tive advantage to stay ahead as a
new/speeches/smart-nation-launch ogy in an operational environment. leading ANSP in the digital era. ■
Kuah Kong Beng has served as the director of Special Projects at CAAS since 2016, in which he oversees the
Smart Digital Tower programme. He also advises on air traffic management (ATM) operational issues and safety,
as well as ATM research and development projects.
Having joined CAAS in 1972 as an air traffic controller, Kong Beng brings with him a wealth of experience in air
traffic control and management that makes him a renowned expert in his field. He also actively contributes to
making new ATM solutions a reality for the region.
8 ECAC NEWS #74AIR TRAFFIC MANAGEMENT AND AIRPORTS
Fostering artificial intelligence
in ATM
Tanja Grobotek
Director Europe Affairs,
CANSO (Civil Air Navigation Services Organisation)
We are without doubt living in turbulent times. The COVID-19 pandemic and its concrescence at
large are probably one of the biggest challenges the aviation industry has ever faced. It is equally
clear that we should not allow COVID-19 to stop innovation in aviation, or progress on building our
future digital skies.
W e are now entering a new,
exciting era for aviation in
ination of a number of use cases
where machine learning (ML) has
“ There are
which the speed of change and been successfully implemented in several new
rate of innovation will be faster ATM and identifies the key ele-
than they have been for decades. ments that will be important for and emerging
In order to support and foster AI-based technologies to gain the
innovation, it is essential to engage trust of ANSPs and air traffic con- technologies
with all stakeholders in the aviation trollers (ATCOs). AI use has kept
value chain and reduce the time to spreading rapidly in the ATM indus-
that can have
deployment of new and emerging try; however, I feel we have barely a huge impact
technologies, which are essential started to scratch the surface of the
to achieve safer, more efficient and opportunities and limitations of AI on air traffic
more scalable air traffic manage- within aviation. The essential ques-
ment, as well as to accommodate tion is not whether – or if – these management
increasingly diverse airspace users AI-based applications will be used,
and boost the sustainability of avi- but when they will be in operation. – and artificial
ation.
intelligence is
There are several new and Enablers
emerging technologies that can clearly one of
have a huge impact on air traffic
management (ATM) – and artificial
intelligence (AI) is clearly one of
N ew technologies come with
some associated risk, and as
aviation is a safety critical domain,
those. ”
those. there is a need to examine the use • Safety assurance.
The availability of enormous cases and capabilities of AI care- Safety approaches that rely on ex-
quantities of data coupled with the fully. In order to see successful isting software are based on the
growth of highly effective and effi- applications of AI in ATM, several assumption that the software is
cient computing power has the po- enablers are necessary: fully explainable. However, it is
tential to enable the widespread • Trust needs to be built for the not always true for machine learn-
use of AI in ATM and related fields. use of this new technology. ing, as almost every popular ML
AI is an exciting prospect and offers The aim of Explainable AI (XAI) is method contains a “black box”
opportunities to increase the effi- to “describe the purpose, ratio- part, hardly constraining the ex-
ciency of services provided by air nale and decision-making process plainability. This represents a new
navigation service providers (ANSPs) of the AI tool in a way that can be paradigm for software assurance,
through a variety of new tools and understood by the average per- where we go from certifiable al-
systems. son”. It is essential that ATCOs gorithms to certifiable data sets
In April 2021, CANSO published know why an algorithm has pro- and training programmes. So even
the white paper, Emerging tech- vided a particular answer and the most robust ML algorithm’s
nologies for future skies: Artificial how that answer compares with performance is highly dependent
intelligence, which explores the the manual completion of a task. on the training data set. It might
main enablers and challenges of This will require ongoing training not be a problem in the case of
AI in ATM. It also includes an exam- and monitoring. applications with low safety effect
ECAC NEWS #74 9AIR TRAFFIC MANAGEMENT AND AIRPORTS I Fostering artificial intelligence in ATM
but can be a strong constraint in
the development of safety critical
“ In a human- tional awareness. Proper assess-
ment of this issue is important to
applications. machine collabo- make the ATCO self-confident in
• Certification of these technolo- their new role. Explainability and
gies will be crucial. ration paradigm, proper safety assurance of AI is a
EUROCAE (WG-114) and SAE (G-34) key factor to build trust, and sup-
are guiding the relevant authori- the human roles port human-machine collaboration.
ties to agree industry standards
on aeronautical systems using AI. are developed
These are now in place, making Fields of application
possible certification that the
and implemented
industry can trust.
• Cyber security and ethics will also
in a way that S ome of the areas where AI has
been successfully implemented
be very important. fosters trust and in ATM, in particular related to air
ANSPs can benefit from employ- traffic control (ATC), tower, air traf-
ing AI systems to improve both confidence by fic flow management (ATFM), and
the efficiency and safety aspects ground operations, and where its
of air traffic control. It is unlikely the human in use shows the potential to deliver
machines will replace humans in real benefits, are the following:
such a safety critical domain for the automation CONFLICT RESOLUTION
the foreseeable future. But they
can be deployed to enhance situ- functions. ” ADVISORY
AI can provide tools to help ATCOs
ational awareness and provide resolve (especially) mid-term con-
decision support to increase operators, to better understand flicts, once these are detected. In-
safety and efficiency by enabling the new data-based algorithms deed, ML algorithms can provide
human experts to focus on more paradigm and adjust their ex- optimal solutions as a sequence
complex decisions. pectations to the existing limits. of actions solving the conflict and
In the current AI/machine satisfying a certain number of in-
Human-machine learning applications, the role of tended qualities: for example, en-
the human does not change signif- suring preferred safety distance,
collaboration icantly. However, the introduction having a minimal number of com-
of AI-based models will support mands, reducing fuel consump-
W e foresee a fluid coopera-
tion framework between
humans and AI to minimise opera-
higher levels of automation, more
advanced decision-making tasks
tion or delay, and others. These
methods can learn from ATCO ex-
perience and also explore billions
and there is little doubt that it will
tional risks and improve the overall impact the way work is performed of possible solutions and choose
performance of the ATM system. in the future. the best from among them.
Experience gained during the suc- OPTIMAL SECTORISATION
The roles and responsibilities of
cessful implementation of automa- In the dynamic airspace manage-
human actors in the functional sys-
tion in the cockpit or in the ment domain, ML/AI techniques
tem will change and evolve. There
unmanned traffic management can introduce a significant im-
could be new actors – Joint Human-
world can be used in designing au- provement. The basic idea is to
Machine Systems – working with
tomation in ATM systems. train a machine with traffic data
different look-ahead time horizons,
In a human-machine collabora- airspace scopes and operational and sector configuration and let it
tion paradigm, the human roles are performance indicators from the learn how to re-shape sectors
developed and implemented in a ones that exist today. boundaries in order to better ac-
way that fosters trust and confi- commodate the traffic.
dence by the human in the au- One key message is that keep-
ing the human in the loop is es- DEMAND PREDICTION
tomation functions. The AI is not ML/AI algorithms can be applied
used as a simple tool but it is used sential in a synergic human-machine
collaboration, and contributes to the to improve the accuracy and pre-
to complement and augment human
resilience of the functional system. dictability of traffic demand by
capabilities.
adding to the current calculation
The objective is that human ex- For example, air traffic con- methods the complementary
perts can focus in areas where ad- trollers should always be able to stochastic information deriving
ditional operational perspective is override some procedures in a from historical data intelligent
needed, while the machine can au- given context while still maintain- processing aimed to provide hid-
tomate large and repetitive tasks ing an acceptable level of safety. den patterns between flight plans
that it can handle more accurately It is even more important be- and actual flown trajectories. This
in a scalable fashion. cause applications increasing the kind of AI-augmented demand
It is key that trust in the Joint level of automation to support enables the improvement and the
Human-Machine System is im- ATCOs to handle traffic with higher efficiency of the entire demand and
proved by upskilling human complexity, might decrease situa- capacity balance (DCB) process.
10 ECAC NEWS #74Fostering artificial intelligence in ATM
I AIR TRAFFIC MANAGEMENT AND AIRPORTS
“ The roles and responsibilities of human actors
in the functional system will change and evolve. ”
REMOTE TOWERS change in both UTM and ATM will
AI-based pattern recognition can entail increased levels of automa-
Conclusions
be used to combine multiple video tion and applying AI/machine
tracks of one detected object (sep-
arate tracks on wings, body, etc.).
learning to the strategic, pre-tac-
tical and tactical aspects to enable Tavailability
he development of AI has come
a long way in recent years. The
of enormous quantities
The entire shape of an aircraft or the transition from airspace-based
vehicle can be recognised from dif- to trajectory-based operations. of data allied to the growth of
ferent viewing angles (e.g. back highly effective and efficient com-
DIGITAL ATCO & DIGITAL TWIN puting power has enabled the use
profile, side profile, front profile). It
Powerful machine learning tech- of AI in ATM and related fields, as
can also be used to detect standing
niques can provide digital assis- has been explored in this article.
aircraft or vehicles on the runway
tance to ATCOs in many ways: The use of machine learning and
or holding point, stands or parking
• by a natural human-machine deep learning systems is exciting
position, and continue tracking if
such objects start moving. Or in a dialogue, enabled by automatic and offers opportunities to in-
close-up view (pan-tilt-zoom or speech, natural language pro- crease the safety and efficiency
static hot-spot cameras), to recog- cessing (NLP) and intent recogni- of services provided by ANSPs
nise when no landing gear is down, tion. Thus, ATCO orders as well as through a variety of different tools
or special situations (e.g. smoke or pilot voice can be directly tran- and systems. The expanded use of
engine fire) and to alert the con- scribed into the system, to increase AI in ATM has the potential to bring
troller. It can also be used to filter ATCO efficiency and globally the significant benefits, including: im-
output data of conventional video safety of the dialogue; proving safety by understanding
tracker and remove false bounding • by increasing situation aware- anomalous traffic patterns; improv-
triggered by cloud movements. ness, through an increased preci- ing capacity by supporting deci-
sion of data surveillance; sion-making and understanding
UNMANNED TRAFFIC
• by learning from previous ATCO changes in flows; and finally, bring-
MANAGEMENT
actions and providing decision- ing environmental benefits by
Unmanned aerial vehicles (UAVs)
aid for increased efficiency, AI will optimising flight plans.
with high levels of automation
and advanced features will re- complement ATCOs enabling them CANSO is committed to working
quire AI to drive the development to enhance their capabilities, re- with the research community,
of new unmanned traffic man- ducing their cognitive involvement States, industry and other stake-
agement (UTM)/U-space services in repetitive tasks and allowing holders to ensure Europe’s airspace
and address the ATM integration them to focus on highly value- is fit for purpose now, and in the
challenge. AI will support ATM added tasks. future. ■
actors and increase safety from
flight planning to operations by
providing new solutions for con-
flict detection, traffic advisory and
“ CANSO is committed to working
resolution, and cyber security. The
with the research community, States,
industry and other stakeholders to
ensure Europe’s airspace is fit for
purpose now, and in the future. ”
Tanja Grobotek joined CANSO in 2018 and is responsible for CANSO’s activities and objectives in Europe.
Tanja previously worked at IATA as the regional director safety and flight operations (SFO) – Africa and Middle East,
where she was responsible for providing strategic direction, guidance and leadership for IATA’s safety, flight
operations and infrastructure activities and initiatives. From August 2013 to February 2014, Tanja was also ad
interim vice president Africa. Tanja has significant experience of working with governments and international air
transport institutions across Europe, Middle East and Africa.
Before joining IATA, she was a flow control specialist with South African air navigation service provider, ATNS, where
she established the Central Airspace Management Unit (CAMU). She started her aviation career in the airline
maintenance planning division of Croatia Airlines.
Tanja holds an executive master’s in business administration in aviation management from Geneva University and
a Bachelor of Science degree in air transport engineering from the Transport Science University in Croatia.
ECAC NEWS #74 11AIR TRAFFIC MANAGEMENT AND AIRPORTS
Artificial intelligence and the next
generation of airport air traffic
management
Andy Taylor
Chief Solutions Officer, Digital Towers, NATS
Not long before the pandemic – if you can remember such a time – I attended a EUROCONTROL
artificial intelligence (AI) summit in Brussels at which people from across the industry gathered
(again, remember that?) to attempt to cut through the hype and look at real, practical ways in which
this disruptive technology could help transform how the aviation industry works.
F astworldforward two years and our
has been transformed in a
One figure at that event that re-
ally resonated with me was the fact
Neural networks work by
analysing data sets in order to
“train” and create an understanding
way that none of us could ever that less than 10% of the data pro-
have anticipated. You’d think that duced by the industry in Europe is of what normal operations look
with parts of the industry still cling- actually used. That’s a massive, un- like. Once a period of training has
ing to survival, talk of investing in tapped resource and long term a taken place, the next stage is for
new and what’s often perceived to much more open approach to how outlier or marginal data to be high-
be a risky technology, would be data is stored and shared will be lighted in what is referred to as
further away than ever. needed in order to unleash its la- “anomaly detection”. This ability to
tent value. detect operational events which
But, of all the transformational
are outside normal parameters is a
technologies surrounding our in- It’s the ethos of harnessing the
key differentiator between machine
dustry, to my mind artificial intelli- power of operational data that has learning and the traditional system
gence, especially when coupled been the cornerstone of Searidge development and coding, meaning
with our ever-maturing application Technologies’ work for the past few the time between development
of digital tower concepts, remains years. Searidge began considering and operational deployment can
the best placed to support the the ATM applications of AI by build- be shortened from years to months.
wider recovery in terms of improv- ing on the technical expertise de-
Working together with Searidge,
ing efficiency, safety and resilience. veloped using machine learning
our focus is on using AI and ma-
Those are issues that still matter and neural networks to enhance
chine learning to identify ways of
today and will matter even more tracking and detection capability in supporting controller decision-
tomorrow. image processing. making in order to improve effi-
ciency, resilience and performance.
This might be by using it to simul-
taneously monitor multiple areas
of interest across an airport – like
runway exit points for example –
something that humans simply
aren’t physically capable of doing.
This ability can then be used to
reduce the impact of external fac-
tors, such as weather, by creating a
more predictable operation in terms
of aircraft spacing and runway
throughput. This focus doesn’t re-
duce the importance of people in
the process, but rather looks at
how to support the optimisation of
human performance.
To give a practical example,
Heathrow digital tower lab before the pandemic, we started a
12 ECAC NEWS #74Artificial intelligence and the next generation of airport air traffic management
I AIR TRAFFIC MANAGEMENT AND AIRPORTS
Heathrow digital tower lab landscape
project looking at whether we 27R/09L, identifying over 40 000 ar- incoming radio traffic and respond
could apply a combination of AI rivals in all. Data analysis has by giving aircraft route clearances
and digital tower camera technol- proved hugely exciting, showing and transponder codes, including
ogy to help cut airport weather- AIMEE performed extremely well, being able to interpret and re-
related delays. including being able to identify air- spond to the “read back” from the
We installed 18 ultra-high-defi- craft in poor weather conditions pilot over the radio.
nition 4K cameras on the tower at and in darkness, when the 4K cam- It’s that ability to interpret am-
Heathrow and additional 4K cam- eras were shown to perform better biguity, which might include the
eras at exit points on the airport’s than the human eye. use of non-standard phraseology
northern runway. The images from So successful were the initial or accented English (something
those cameras were then fed live trials that we hope to be able to ex- humans are so good at), that has al-
into Searidge’s AI platform, known tend the work to include more var- ways been the real technical hurdle
as AIMEE. ied weather conditions, giving to the idea taking off, but the
When trained on enough data, further opportunities to refine the unique nature of AI and machine
AIMEE can identify the exact mo- model and test whether the solu- learning means the results of our
ment when an aircraft has safely tion would work in full CAT II/III low non-operational trials have been
left the runway and to then notify visibility conditions. If AIMEE per- very encouraging.
the air traffic controllers. Our intent forms as well as expected, it will AIMEE was asked to interpret
was to prove that when a tower like prove an enormous benefit given the pilot request, check the details
Heathrow’s disappears into low the impact Low Visibility Proce- against the existing flight strip sys-
cloud despite the taxiway and run- dures (LVPs) have on operations at tem and then respond to the pilot
ways remaining clear – known as airports around the world. with the appropriate clearance or
“VIS 2 conditions” – AIMEE could To give another example, auto- request for clarification. Obviously,
inform the controllers that the run- mated voice clearances are some- this is a long way off being ready to
way was free for the next arrival, thing that the industry has dabbled deploy, but our analysis shows that
something that would help recoup with in the past, but the technical with the provision of enough train-
the landing capacity that is lost in hurdles have always seemed insur- ing data, AIMEE performed very
these circumstances. mountable. Not only must a system well when monitoring real pilot
Not only would this reduce de- be able to issue the clearance safely RT transmissions.
lays for the airlines and their pas- and correctly, but it would also If the work continues towards
sengers, but it would also ensure need to be able to understand the achieving something that is one
safety is maintained and reduce the pilot’s response and act appropri- day operationally deployable, it
monitoring workload on the con- ately. could be possible to automate
trollers. This would free them up to Again, here machine learning some of the more routine tasks
make other key decisions, whether and artificial intelligence could pro- controllers undertake, helping re-
that’s in support of capacity growth, vide the key. Using AIMEE, we have duce their workload and again,
resilience, safety or efficiency. been experimenting as part of a leave them free to concentrate on
During our trials, AIMEE contin- non-operational trial, by training the things where their skills and
ually monitored arrivals on Runway the system to successfully monitor training are best employed.
ECAC NEWS #74 13AIR TRAFFIC MANAGEMENT AND AIRPORTS I Artificial intelligence and the next generation of airport air traffic management
These are two radical applica- One thing that was very clear is digital tower applications freeing
tions, and while they are not ready that despite the pandemic, we remain people of routine tasks and allow-
for operational deployment, to me in an era of accelerating change, ing them to concentrate on deci-
it is the combination of the digital with pressure to increase the rate at sion-making and performance. ■
tower technology and artificial in- which new systems can be intro-
telligence that is the real potential duced safely. Technology is going
Find out more at
game changer in terms of airport to play a huge part in the future of
www.nats.aero/digital-towers
performance. ATM, with AI machine learning and
Most digital towers around the world today
have been put up to replicate the view of the
airfield and relocate the controllers. That’s an
application that suits some smaller airports, but
it’s not going to be how we really unlock the full
operational and business benefits. We’ve been
on a mission to rethink how digital towers can be
deployed in order to tackle specific challenges,
and have distilled that thinking into five opera-
tional concepts, or what we call “models”. All are
MODEL 1 Digital Tower in Tower
built on the same software platform, making
A tower within a tower for operating a small airfield
them compatible with the kinds of future AI-
remotely from inside the tower of another “parent”
enabled applications like those I have outlined. airport.
MODEL 2 Remote Digital Tower MODEL 3 Remote Digital Tower+
A fully digital control tower for a single runway airport, A fully digital tower for more complex, mid-sized
which can be either on or off-site. airports, which can be operated within the airport or
from another site.
MODEL 4 Hybrid Digital Tower MODEL 5 Hub Digital Tower
A hybrid digitised tower, ideal for upgrading an ex- A fully digital tower, perfect for replacing a physical
isting physical tower at a larger airport. tower at a major, multi-runway, multi-terminal airport,
or for creating an equally capable contingency.
With nearly 30 years of experience in the air traffic management industry, Andy Taylor has led, developed and
delivered innovative ATC solutions for global customers to increase airport capacity and operational performance.
As chief solutions officer, Andy works across both NATS and Searidge Technologies. He is responsible for the joint
development of the partnership’s digital tower capabilities and delivery models.
Andy is currently leading the NATS/Searidge deployment of the world’s largest digital tower programme at
Singapore’s Changi Airport, together with establishing a network of digital tower laboratories.
14 ECAC NEWS #74AIR TRAFFIC MANAGEMENT AND AIRPORTS
Schiphol as a tech company
Floris Hoogenboom Sebastiaan Grasdijk Bas Cloin Martijn Schouten
Head of Artificial Lead Data Scientist Lead Data Scientist Product Manager
Intelligence
Schiphol is quickly becoming Europe’s leading digital airport. In managing an airport that size,
predictability and insights play a key role. From optimising our passenger experience, minimising
our environmental footprint to just running our day-to-day operations. To achieve this, Schiphol is
building a solid data foundation that feeds many of its cloud-based artificial intelligence (AI)
capabilities.
D ata and AI are a foundational
part of transforming from a tra-
residents informed on the number
of flights they can expect over-
head, we developed a mobile ap-
coast, Schiphol’s runway system is
designed to deal with many differ-
ditional airport into a tech com- ent weather conditions. This means
pany. This involves transforming plication called Notifly. This app that the number of flights on a cer-
the way we store and process data, shows the predicted number of tain route may differ depending on
but even more importantly how we flights per hour in the next 24 hours a lot of complex factors. To predict
act on the insights we develop. based on the user’s location. This this number of flights, a machine
Schiphol is moving from using AI as allows residents to factor in the ex- learning model is trained on a year
a tool to develop insight, towards pected nuisance while planning of historic data. During training, the
using AI as a decision-support tool. their activities. model learns the complex depen-
Ultimately, our goal is to put AI first The reason why an app like No- dency of the number of flights on
in a lot of our processes, striving to- tifly is especially interesting for an multiple factors, including weather
wards total airport management airport like Schiphol has to do with information (e.g. wind direction
and even autonomy. In this article, the runway configuration. Due to and speed), the flight schedule and
we will take you through a few of its location on flat terrain near the generic seasonal effects.
the use cases Schiphol is working
on and explain how these fit into a
larger vision of AI as a transforma-
tional power in how Schiphol is
run.
Notifly
Schiphol Airport, the oldest air-
port worldwide still at its origi-
nal founding place, is situated in a
part of the Netherlands crowded
with businesses but also with resi-
dents. Although we try to minimise
nuisance for residents, people liv-
ing close to flight paths experience
aircraft noise every day. To keep Screenshots of predictions in Notifly app
ECAC NEWS #74 15You can also read
