APPLICATION OF AI IN THE NAS – THE RATIONALE FOR

AI-ENHANCED AIRSPACE MANAGEMENT

Ronald L. Stroup Federal Aviation Administration
Kevin R. Niewoehner Booz Allen Hamilton
Rafael D. Apaza NASA Glenn Research Center
Daniel Mielke Deutsches Zentrum für Luft- und Raumfahrt (DLR) Oberpfaffenhofen (Germany)
Nils Mäurer Deutsches Zentrum für Luft- und Raumfahrt (DLR) Oberpfaffenhofen (Germany)

Abstract formal NAS Horizons (3) (20 to 25 year look

forward) analysis within the International Civil
This paper extends on the initial findings of Aviation Organization (ICAO) was conducted.
“APPLICATION OF ARTIFICIAL These analyses provided the foresight to address,
INTELLIGENCE IN THE NATIONAL AIRSPACE through enterprise analyses, the role of emerging
SYSTEM – A PRIMER” (Stroup & Niewoehner: technologies in the future NAS. These analyses
Herndon, VA; ICNS-2019), and looks at why the provided insight into the longer-term vision to
current technologies, enterprise architecture, and support risk reduction of disruptors that may impact
future program plans may not be enough to address the future NAS. As advances occur in smart surface
persistent operational challenges. This paper further transportation systems as well as the maturing of
explores why emergent operational concepts, decision-support tools in National Defense and
business models, and demand profiles may National Intelligence sectors acceptance of AI in
necessitate Artificial Intelligence (AI)-enhanced critical infrastructure operational technology will
Communications, Navigation and Communications become viable.
(CNS) infrastructure to disrupt current operational The NAS is a complex system-of-systems and may
impediments. European airspace, as well as the appear to be non-deterministic (4). However,
NAS, has similar challenges. Key challenges ANSP’s apply various techniques to bound the
explored in this study include: traffic flow operations to make the NAS appear deterministic.
management of diverse users; UAS Traffic These techniques consist of miles-in-trail, minutes-
Management – Air Traffic Management (UTM- in-trail, low altitude alternate departure routing,
ATM) airspace integration; equitable access to capping, tunneling, ground delay program, time-
airspace; information exchange networks and based flow management, airspace flow program,
airborne-ground interoperability of AI applications. ground stop, adaptive compression, and integrated
Finally, we examine why trustworthiness and collaborative routing to produce a safe, orderly, and
resiliency will be key mileposts on the regulatory expeditious flow of traffic while minimizing delays.
pathway to AI certification.
Is a non-deterministic system like the NAS just a
Background very complex deterministic system that we don’t
know enough about? Can the application of
The NAS is part of the transportation critical emerging technologies like AI and Machine
infrastructure sector (1) and concerns relate to Learning (ML) enable us to know more of the parts
sustaining minimum NAS operation. At issue within to enough degree and their behaviors as well, then it
the broader aviation marketplace, is the continued should be possible to forecast its state with an
search for a solution set to the persistent and increased level of confidence at any time in the
evolving challenges that constrains the NAS future?
capacity, efficiency and resiliency.

In 2017, a Strategic Outlook (2) (5 to 10 year look

forward) analysis that complemented the more

978-1-7281-0649-6/19/$31.00 ©2019 IEEE

Authorized licensed use limited to: Auckland University of Technology. Downloaded on June 05,2020 at 19:57:20 UTC from IEEE Xplore. Restrictions apply.
Since this work started in 2017, some additional
events have reinforced the work:
FAA Reauthorization Act of 2018: (Public
Law 115-254; 10/5/2018), SEC. 548. SENSE OF
CONGRESS ON ARTIFICIAL INTELLIGENCE
IN AVIATION. “It is the sense of Congress that the
Administration should, in consultation with
appropriate Federal agencies and industry
stakeholders, periodically review the use or proposed
use of artificial intelligence technologies within the
aviation system and assess whether the
Administration needs a plan regarding artificial
intelligence standards and best practices to carry out
its mission.” Figure 1 UTM / ATM / ETM Operations
Concept
Executive Order on Maintaining American
Leadership in Artificial Intelligence: February 11, UTM-ATM airspace integration; equitable access to
2019. “Continued American leadership in AI is of airspace; information exchange networks and
paramount importance to maintaining the economic airborne-ground interoperability of AI applications.
and national security of the United States and to
shaping the global evolution of AI in a manner Traffic Flow Management of Diverse Users
consistent with our Nation’s values, policies, and
priorities.” National Institute of Standards and The NAS users are becoming very diverse in terms
Technology (NIST) is leading the effort per the of vehicles, their performance as well as their
Executive Order to develop a plan for federal business models to meet their business goals (see
engagement in AI standards. The standards must Figure 1). This is creating challenges with
build trust and confidence in AI technologies and maintaining or improving capacity and efficiencies
applications to meet the nation’s needs. AI in the NAS. At issue within the broader aviation
leadership will sustain with alignment to the Four marketplace, is the continued search for a solution
Main Pillars of Emphasis: AI for American set to the persistent daily delays and schedule
Innovation; AI for American Industry; AI for the perturbations that occur within the NAS. The delays
American Worker; and AI with American Values. persist in the face of reduced demand for commercial
routings. Every delay represents an economic loss to
Eurocontrol AI Conference: May 23, 2019. (5)
commercial transport operators, passengers,
This inaugural event focused on the how AI is
freighters, and any business depending on the
driving business growth in general and the potential
transportation performance. (6)
for AI to disrupt the aviation business models. The
conference discussed key issues and challenges such This is not unique to the NAS, the UK’s air
as pilotless aircraft and automated Air Traffic navigation service provider National Air Traffic
Control (ATC). Service (NATS) is trying to resolve capacity
challenges at London’s Heathrow Airport and have
Why AI may be leveraged to solve our turned to artificial intelligence to reclaim lost
capacity caused when the airport’s 87m-high air
complex challenges traffic control tower is operating in reduced visibility
There exist complex persistent and evolving conditions. (7) NATS is hoping the AI could be
challenges that will demand we look at applying introduced into Heathrow’s operation later in 2019.
emerging technologies to meet performance goals The first step will be to collect data on some 50,000
and mitigate chronic NAS performance issues. landings to ensure the accuracy of the system. The
European airspace, as well as the NAS, has similar company’s findings will then be presented to the UK
challenges. Key challenges explored in this study Civil Aviation Authority. During the summer of
include traffic flow management of diverse users; 2018, Europe seen enroute delays peak compared to

Authorized licensed use limited to: Auckland University of Technology. Downloaded on June 05,2020 at 19:57:20 UTC from IEEE Xplore. Restrictions apply.
the summer of 2017. The contributing factors were 600
traffic up 4%, lack of capacity by Air Navigation
Service Providers, (ANSPs), overall shortage of 500
ATC staff, and weather. (8)
400
Key Question: Can AI help identify and
reclaim lost capacity and reduce delays? 300 UAVs
Aircraft
UTM-ATM-E-above A (ETM) Airspace 200
Integration
100
The sheer number of drones will require machine-to-
machine autonomy in the current Class G airspace, 0
but these emerging users will not be satisfied with FL growth factorRL growth factor
limited access to other parts of the airspace. This
increasing demand will push into more controlled Figure 2 Projected growth factors of Forward
airspace with its business model and infrastructure and Reverse Link (FL & RL) data throughput
(see Figure 1). According to ICAO, “establishing a
comprehensive sectorial architecture will provide a Information Exchange Networks
secure foundation for air transportation
interoperability.” (9)
Overall worldwide civil air traffic growth is
estimated to be 84 % more civil planes in the air
Key Questions: Can AI help optimize when comparing 2017 and 2040, following
integration and interoperability of diverse users EUROCONTROL’s latest report on the growth
in the NAS? of European and worldwide air traffic (11). With
the increase in flight movements, new entrants,
Equitable Access of Diverse Users to Airspace lower latency and increased data processing
The global navigable airspace is a limited resource. requirements in the future, legacy systems in the
Each ANSP per ICAO is responsible to ensure the ATM will reach their capacity limit (12). This
safety and efficient use of airspace they manage. In poses the question how much data will be
the U.S. the Airspace Access Priorities Aviation required per flight, allowing us to estimate the
Rulemaking Committee is responsible for CNS data capacities of the future global
developing criteria that may be used to consider
airspace. After intense analysis, (11) the
competing requests for airspace access (10).The
ability to manage access of commercial transports, estimated growth of worldwide civil air traffic is
general aviation, unmanned aircraft, and commercial 2.7% per year for manned; 13.1% per year in
space operators as well as existing and emerging unmanned aviation; and 16% per year for civil
business models might constrain future air transport aeronautical data link traffic.
growth. (9) How do we balance the societal,
operator, and end-users benefits while maintaining We see in Figure 2 a projected increase of up to
capacity, efficiency and resiliency of the NAS. a factor of 500 for the UAV’s Forward Link
(FL). This evolved data throughput requirement
Key Questions: Can AI help to identify a is orders of magnitude greater, and drives new
strategy for equitable access of diverse users? data-link requirements to accommodate
projected exponential demand growth by new
entrants. Thus, we conclude, we need a hybrid
approach of new datalinks and new AI
technology to perform smart selection of

Authorized licensed use limited to: Auckland University of Technology. Downloaded on June 05,2020 at 19:57:20 UTC from IEEE Xplore. Restrictions apply.
transmitted information, while handling reduction and improved business management
increases of FL/RL data, and flight movements! processes. At the core of an AI-enabled
infrastructure initiative, is the desire to get insight
Key Questions: Can AI help manage the into system performance to reduce uncertainty.
volume and velocity of data exchange to optimize Predictability improvement lends confidence to
airspace operations? strategic planning and may decrease tactical
adjustments. The introduction of AI-enabled
systems into existing infrastructure and transitioning
Interoperability of Airborne and Ground AI it into existing workflows and workloads will require
Applications addressing challenges in different areas that include
acquisition, workforce culture, systems architectures
The aviation ecosystem connects operations,
and more. Overcoming these challenges will bring
capabilities, and infrastructure across the airborne,
about the beginning of a Cognitive NAS.
airspace, air traffic, and airport domains to support
the future NAS. It includes all stakeholders as well
as UAS and commercial space operations. (13) The Mission Effectiveness is a Key Measure.
application of emerging technologies, such as AI, is The US ATC system requires numerous CNS
a multi-dimensional challenge. Simply applying AI systems to deliver services. As of 2017 the FAA had
to one of the domains may not be enough to leverage about 19,000 communications systems, 2,300
all the benefits. The interdependencies across the surveillance systems and 13,500 navigational
domains, as well as other critical infrastructure systems, 2,400 automation systems, and 2,300
weather systems servicing the NAS. (15) There is a
sectors (e.g., energy, finance, communications)
sizable amount of technical performance data
demand AI application standards across the domains collected on a regular basis for all CNS systems in
in support of mission objectives. (14) service today. NAS maintenance practices and
These applications and their role will be defined performance metrics and evaluation procedures are
across a set of AI use cases linked to mission effective, however, services reductions that impact
shortfall, mission objectives, type (advisory, air traffic operations still occur. A predictive
decision), risk level, and data requirements. A maintenance approach, as one example of a use case,
comprehensive aviation ecosystem to support coupled with preventive services will assist and yield
application of emerging technologies must be NAS performance improvements. The use of
developed. The FAA is continuing to examine with machine learning where the learning set is comprised
external stakeholders exploratory AI concepts such of different data collection events and real-time
as IA operations theory and intelligent CNS (iCNS), monitoring can enable an expert AI system to predict
intelligent NAS (iNAS) and associated AI use cases. outage conditions and report it to technical
operations for corrective investigation.
Key Questions: Can AI help integrate
enterprise applications across the aviation
ecosystem? Data Management
In conventional computer programming, code is
AI-Enabled Infrastructure to Solve developed and data input into the code for execution.
the Challenges In artificial intelligence the inputs to the system is
data plus a set of expected answers. The machine is
Infrastructure is a critical enabler to the delivery of trained rather than programmed. Rather than writing
ATC and ATM services. Many electronic systems code for each specific task, lots of examples are
are used throughout the NAS for airspace collected that specify the correct output for a given
management and their reliable performance is input. The objective is to have the program work for
required to keep the air traffic system operating all cases. In this approach training data is very
safely and on time. AI-enabled infrastructure is a important. Erroneous data and small amounts of it
promising technological improvement that can can yield incorrect results and limit the benefits of
benefit regularity of air traffic services, e.g. outage artificial intelligence. When preparing training data

Authorized licensed use limited to: Auckland University of Technology. Downloaded on June 05,2020 at 19:57:20 UTC from IEEE Xplore. Restrictions apply.
for neural network learning, consideration shall be perception that emerging technologies can result in
given to the following factors: job displacement. Historically the aviation programs
have established qualitative goals (increase capacity,
Relevance - Irrelevant data attributes present a
improve efficiency) at the concept level and
problem in that they can be misleading to the
quantitative benefits are defined during the
learning algorithm and unnecessarily consume investments analyses to support a cost-benefit report.
computational resources. While all attributes are not
relevant, the difficulty comes in identifying which After deployment, many of these quantitative
are useful specially in large data sets. (16) benefits have not materialized for one or more
reasons, typically due to integration losses, and
Consistency - Inconsistent training data and performance slippage. Resiliency is one area where
missing attributes throw off a training algorithm and the FAA Administrator has clearly defined
will lead to a poorly trained intelligent system. It is performance objectives. (20) The application of
important to eliminate redundant attributes to reduce emerging technologies into the NAS will take
unnecessary computational expense. education to energize the workforce that technology
will focus on mundane and repeatable tasks and free
Data noise - Another source of error is noise in up the worker to focus on thinking and reasoning
a data set. Data noise can be in the form of spelling tasks. (21)
errors, or systematic noise where an algorithm is
influenced and swayed in a biased direction. Architecture
Data quality (cleansing, labeling and The complexity and connectivity of our NAS
processing). The volume, protection and velocity of systems as well as the volume and velocity of data
presents an opportunity to apply emerging
data is constantly changing. Much of the testing has
technologies to meet the future NAS needs. The
been done on pristine data in a lab. Current
current NAS infrastructure will be challenged by the
deterministic systems account for this but how about ability to handle the amount of data, the rate of
non-deterministic systems. change of the data and identifying unknown patterns
Data poisoning – Techniques such as model and relationships across the data.
reversion, and classification manipulation are key
challenges to be addressed by risk and data
management policy. (17)

Workforce Skill Set and Culture

In order to support an AI-enabled infrastructure, a
workforce with skill sets associated with computer
science, data science, psychology and domain
expertise will be necessary. (18) Domain expertise
consists of operational knowledge and sound system
engineering practices to understand what AI is doing
and how to maintain awareness of the current modes
and system states. The right skill set will drive trust Figure 3 AI Enabled NAS Architecture
in the system as it relates to trust and resiliency.
Understanding the design characteristics, training The NAS architecture must also be updated to meet
and procedures aspects related to AI systems will these challenges. Figure 3 represents a notional AI-
provide for human-system resilience. (19) enabled architecture. It starts with data sources, both
internal and external that provides insight into all
The cultural challenges stem from limited resources aspects of the aviation ecosystem. That data is stored
and budget to explore these emerging technologies, and prepared to be processed by applying various
the heavy dependency on legacy systems and the data management best practices. The data is sent to a

Authorized licensed use limited to: Auckland University of Technology. Downloaded on June 05,2020 at 19:57:20 UTC from IEEE Xplore. Restrictions apply.
processor as well as the enterprise monitor. The AI- continuously in training. An agile methodology of
enabled processor consists of a Central Processing program management, in which each stage building
Unit (CPU) and Graphics Processing Unit (GPU). on successive stages, is more appropriate for an AI-
The addition of the GPU (22) allows the system to enabled effort. (27)
take large batches of data and perform the same task
The agile approach values adaptability, lean
over and over very quickly. The CPU is still the
development, time, and sustainability to a dynamic
brains of the system but conventionally it can only
environment. The current FAA Acquisition
handle of couple tasks at a time while the GPU can
Management System is predominately a waterfall
handle thousands of tasks at a time and therefore
approach to acquisitions. These waterfall-based
accelerate computational tasks. The enterprise
acquisitions take an extended period and often
monitor (23) provides and end-to-end monitoring
delivers minimal capabilities with outdated
capability. The AI-enabled monitor will apply the
technology. A hybrid approach is worth evaluating
data to combine lower-level system status
to support an AI-enabled NAS.
information and user experiences into service level
indications and higher-level awareness and
operational impact to inform decision-makers across Models/Algorithms
the FAA in a timely manner to assure the capacity, NAS functions that may be accomplished by AI will
efficiency, and resiliency of the NAS. The ability to require different algorithm technologies and
collect, manage, process and monitor requires an implementation approaches. A single learning
architecture robust network and information model will not meet all classes and types of decisions
exchange capability to handle the volume of required to predict and support numerous NAS
information that is needed to support future NAS functions and operations. The NAS will have to
operations. employ a combination of machine learning and deep
learning models tailored to solve specific problems
Safety, Security and Resiliency Aspects and meet different needs. AI development for NAS
implementation can be achieved by conducting
The concept of resiliency takes a risk-based investigation in Data Sciences, Core AI and Applied
approach (24) to address the interdependencies AI. The NAS is a data rich environment and
among today’s complex and critical infrastructure. appropriate data needs to be selected to train ML and
Alignment of protection (safety cyber-security, Neural Network (NN) models. Data Science
information security, physical security, and human investigates the correct features that need to be used
factors) mechanisms, detection and response to train models and evaluates data for bias removal
(monitoring and contingency) methodologies and and noise reduction. Core AI focuses on
recovery (logistics and maintenance) techniques are investigation and development of fundamental
key to improving the NAS resiliency posture. algorithms that solve specific problems. Core
artificial intelligence includes development of
AI has also been shown to be able to detect and/or algorithms with integrated safety features to yield a
mitigate different kinds of attacks on information trusted system. Moreover, it concentrates on
systems. This includes both higher layer attacks fundamental development of NN/ML with safety
(“hacking”) and physical layer attacks (“jamming”) assurance and, in some cases, the ability for machine
(25). AI can help with deep reinforcement learning intelligence to explain its decisions and actions (28).
to gather new information with each incident and A NAS user should be able to interrogate the
help recover quicker. explainable model to understand why a prediction
was generated. The system’s ability to effectively
Acquisition explain decisions will provide an element of trust in
The conventional waterfall methodology in program a safety critical environment. An important question
management may not be sufficient due to changing on explanation is; whether the system can provide a
system requirements, application of models and convincing reasoning on decisions taken. Applied
algorithms, non-deterministic testing, and vague AI concentrates on algorithm application
lifecycle planning. (26) AI-enabled machines are investigation to address specific NAS needs and

Authorized licensed use limited to: Auckland University of Technology. Downloaded on June 05,2020 at 19:57:20 UTC from IEEE Xplore. Restrictions apply.
requirements. It focuses on identifying the optimal Smart routing algorithms, probably enhanced by AI
learning model for a given task and proper model technology, may have the potential to increase the
implementation. The development of Core and network capacity of nowadays systems. However,
Applied AI will result in the implementation of a the current patchwork situation makes it challenging
machine learning enabled trusted service. to optimize the routing – even for an AI (30).
To cope with the increasing demand of data
Why Emergent Technical Solutions throughput required for future aviation,
communication systems are under development such
Must Map to Airspace / Operations as the L-band Digital Aeronautical Communications
Classification System (LDACS) and C-band Digital Aeronautical
Communications System (CDACS); (31), (32) as
Communication – Today and the Future well as the Iridium and Inmarsat satellite
Nowadays communication in aviation is still constellations (see Figure 4).
unthinkable without analog voice radio (VHF audio) LDACS is the future ground-based data link for civil
– a technology that goes way back into the beginning aviation; standardization within ICAO started in
of aviation. Except for minor changes in the channel December 2016 and draft SARPs (Standards and
and frequency allocation, not much has changed over Recommended Practices) were passed by end of
2018. The most important LDACS functionalities
have been proven in multiple flight trials.
CDACS, as a robust C2 Data Link for highly
automated aviation, is one of the key enablers for
bringing disruptive technologies, like AI-driven
systems, into aviation. It is designed with a focus on
physical layer robustness while providing high data
rates, with first concepts already tested successfully
in-flight trials. (33), (34).
Figure 4 Logarithmic Scaled Throughput in Navigation – Today and the Future
kbit/s of Several Communication Systems The VOR and DME are still used as most prominent
used or foreseen in Aviation navigation aid in aviation. And even though GNSS-
the years, although the amount of aircraft in the air based navigation like GPS or GALILEO became part
increased significantly. For the exchange of digital of people’s everyday life, it is still not the primary
data, the VHF Data Link (VDL) is used, e.g. in Mode navigation systems in aviation. The term to use here
2. However, it does not provide any mechanisms for is APNT - Alternative Positioning and Navigation
authenticity checks and its data throughput is very and Timing. Other APNT systems providing at least
limited (29). similar precision to GNSS are not available yet (35).
Overall, two developments are going on: Fully integrated, intelligent CNS also allows for
alternative positioning, navigation, and timing
First, more and more piloting tasks are done (APNT). For example, the ranging functionality of
automatically by computer-based systems on board the communications system LDACS is foreseen to
the aircraft. This is expected to lead to the single- establish a future APNT capability.
piloted aircraft in the future.
Also, Multi-Frequency, Multi-Constellation solution
Second, the field of unmanned aviation is growing between all navigation systems worldwide highly
tremendously (see Background). Both developments improves availability and coverage of navigation
are expected to influence at least parts of the required aids and of navigation capabilities worldwide. For
CNS systems. Thus, the CNS technologies the interoperability of ground-based, air-based and
currently in use will not be able to handle the space-based systems, also new, smart routing
challenges of the future. algorithms, enhanced by AI technology, may be key

Authorized licensed use limited to: Auckland University of Technology. Downloaded on June 05,2020 at 19:57:20 UTC from IEEE Xplore. Restrictions apply.
to interconnect all data from different system, them decades off the pace compared to state-of-the-
uniform their data format and make it available for art cell phone industries.
all entities in need of precise navigation on the spot.
No solution providing all three services at the same
time while exploiting available synergies has been
Surveillance – Today and the Future deployed so far. We call these kind of systems
The most prominent cooperative surveillance system integrated CNS. However, on both sides of the
in aviation is the secondary surveillance radar (SSR). Atlantic, projects are working on bringing modern
It is mostly used in Mode A which is facing problems technology into CNS, namely NextGen in the US
not only but also because of the 12-bit limit of the and SESAR in the EU.
aircraft identifier (36). A more modern cooperative LDACS (32), applies modern digital
surveillance system is Automatic Dependent communications technologies known from 4G
Surveillance – broadcast (ADS-B) (37). Besides mobile radio networks, including cyber security
repealing the limit of the air craft identifier, it (41), and – to the best of the authors' knowledge – is
broadcasts the aircraft’s position on a regular base. the first aviation system so far providing true
In a market space where 18-24 months is out of date, integrated CNS.
aviation CNS infrastructure is decades off the pace.
In the future, we will see more and more systems LDACS has already been tested in flight trials, (42)
being integrated into one another and thus classical and provides:
surveillance approaches will also involve direct • Data rates up to 2,600-kbit/s (32);
point-to-point communication, and surveillance data • APNT capability of RNP0.3 and better
broadcast, and possibly multicast of the positions of
these nodes. Again, to make this multi-hop arbitrary without the need for additional spectrum;
mesh network multicast routing even possible, AI • ADS-C (Automatic Dependent
will play a strong role here. Surveillance-Contract) service;
• ADS-B service within the currently
Spectrum Issues
As every wireless system, a CNS system depends on developed LDACS air-to-air mode (43);
three resources: time, space, and spectrum. and
Especially the latter is a rare, valuable resource that • Non-cooperative surveillance (44);
should be exploited efficiently. AI has proven useful
to support this task multiple times. In (38), an AI- The next step after bringing true integrated CNS to
based algorithm has been used to optimize radio cell the market, by e.g. introducing LDACS worldwide
and frequency planning considering interference. as the air-ground datalink in the Future
The concept of AI-based spectrum sensing, as a first Communication Infrastructure (FCI) in civil
step of increasing spectral efficiency, has been aviation, is to bring integrated intelligent CNS to life.
investigated in (39). This means enhancing CNS with AI, thus to enable
the system to observe all incoming positioning and
Nowadays, aeronautical CNS systems are restricted flight trajectory data from all participants in the
to certain frequency bands only. The mentioned AI- network in real time to furthermore detect potential
based approaches have the potential to enable bottle necks and circumvent these by intelligent
coexisting systems in the same frequency band rerouting of air vehicles. It can also react on sudden
(spectrum sharing) resulting in more useable emergencies, e. g. an ambulance helicopter that
bandwidth for aeronautics. (40) requires an unforeseen flight corridor.
Integrated CNS – Today and the Future
In the past, CNS systems were designed to provide Acceptance through Trust
just one of the three services (either communication AI today supports analytical-type functions
OR navigation OR surveillance). Furthermore, new (research, planning, post-ops analysis) but needs to
systems popped up without outdated systems evolve to support mission support/operational
disappearing. These two factors led to nowadays functions involving safety-of-life challenges. (45)
situation: a patchwork of different systems, most of

Authorized licensed use limited to: Auckland University of Technology. Downloaded on June 05,2020 at 19:57:20 UTC from IEEE Xplore. Restrictions apply.
The future of AI will evolve to mission support (i.e., critical areas and proposes some ideas to drive
priority and response maintenance, traffic flow further discussions. By addressing the application of
management) functions and mission operations (i.e., AI in the NAS now, we can, as a community, make
separation management, sequence management, and informed decisions about the path forward to an
spectrum management) functions. (6) In an effort to intelligent NAS (iNAS).
leverage artificial intelligence technology and
support the certification of AI applications in the
aviation (i.e., intelligent flight control systems, References
intelligent air traffic systems, and intelligent airport 1. NIPP 2013. DHS.gov. [Online] 2013.
systems) ecosystem, sufficient processes and 2. Aviation Strategic Outlook Areas Driving
methods, need to be established. One possibility is a NAS ieeexplore.ieee.org. [Online] ICNS 2018
framework consisting of existing as well as enhanced
verification, validation, metrics, independent 3. Thirteenth Air Navigation Conference.
verification & validation, and trust aspect for AI- ICAO.int. [Online] October 9-19, 2018.
enabled systems. (46) 4. Optimal Control of Non-deterministic
• Verification: build the product right, of AI- Systems for a Computationally Efficien Fragment of
enabled systems should include the Temporal Logic. cds.caltech.edu. [Online] 2013.
following areas: contracts, process, design, 5. Eurocontrol. High-level Conference on
code and documentation. Artificial Intelligence: [Online] May 23, 2019.
• Validation: build the right product, of AI-
enabled systems should include the 6. Application of Artificial Intelligence in the
following areas: items (i.e., models and National Airspace System - A Primer ICNS-2019.
algorithms), the testing environment, and 7. Aviation Daily. [Online] January 28, 2019.
testing (i.e., unit, response times, failure
modes, and utilization). 8. ATW Daily News. Air Transport World
• Metrics for AI-enabled systems may Daily News. atwonline.com. November 8, 2018.
include learning time, recall response, 9. ICAO. ICAO Secretary General ACI World
accuracy, repeatability, resiliency (47), Assemble. icao.int. [Online] April 3, 2019.
integrity (48), and confidence. (49)
10. U.S. DOT / FAA. Airspace Access
• Independent verification & validation of
Priorities. faa.gov. [Online] November 20, 2017.
AI-enabled systems ensures the AI aspects
are thoroughly tested against its complete 11. EUROCONTROL. EUROPEAN
set of requirements. AVIATION IN 2040 EUROCONTROL, 2018.
• Trust of an AI-enabled systems must also 12. FAA. Modernization of U.S. Airspace.
address the privacy (50) of data and the faa.gov. [Online] FAA, June 25, 2019.
transparency (51) of the processing to
mitigate human bias. 13. Enterprise Integration: A Framework for
Connecting the Dots. incose.org. [Online] 2012.
Conclusion 14. NIST AI Standards Workshop - Facilitate
The NAS and European airspace are complex, and development of AI Standards. May 30, 2019.
essential parts of the critical infrastructure sectors.
The disruption of new users, technology and 15. FAA. NAS Operational Facility Inventory
persistent and emerging challenges will continue to Report. FAA.gov. [Online] February 2017.
challenge our ability to effectively manage the 16. NAS resiliency : minimizing the impact on
airspace of the future. The application of emerging NAS end-users. trid.trb.org. May 2, 2019.
technologies, like artificial intelligence, may provide
17. IARPA Seeks to Apply Trusted Computing
capability to leverage opportunities and mitigate
to AI and ML Models. militaryaerospace.com.
risks to reduce the impact on NAS and European
end-users in the future. This paper simply identifies

Authorized licensed use limited to: Auckland University of Technology. Downloaded on June 05,2020 at 19:57:20 UTC from IEEE Xplore. Restrictions apply.
 18. Future of AI in the Defense Agencies 38. Feasibility of Cell Planning for the LDACS
Panel, Genius Machines Symposium. April 2019. https://elib.dlr.de/121009/1/mostafa.pdf.
19. FAA Human Factors Design Standards. 39. Deep Learning Classification of 3.5 GHz
HF-STD-001 . faa.gov. [Online] 2003. Band Spectrograms 2018 arxiv.org/abs/1806.07745.
20. FAA ATO Operational Contingency 40. AI Will Rule the Airwaves. Tillgham, Paul.
Group. FAA Administrator Target December 2017. 2019, IEEE Spectrum (June).
21. Preparing Your Agency for AI and 41. ICONAV. ieeexplore.ieee.org. ICNS
Automation, FCW. Washington, DC : FCW, 2019. Conference, 2013.
22. NVIDIA. What is the difference between 42. LDACS FLIGHT TRIALS ICNS-2018
and CPU and GPU? blogs.nvidia.com. Conference, April 2018. ieeexplore.ieee.org
23. FAA Technical Operations Conops, 43. Design Approach of a Future Air-to-Air
Enterprise Monitoring Capability. 2015. Data Link. ieeexplore.ieee.org. 2018 DASC
24. Infrastructure Resiliency: Derived and 44. LDACS-BASED NON-COOPERATIVE
modified from A Risk-Based Framework. 2013. SURVEILLANCE. ieeexplore.ieee.org. ICNS-2014
25. Jamming and anti-jamming techniques in 45. Panel, Future of AI in Defense Agencies.
wireless networks: a survey. dl.acm.org. 2014. Genius Machines: New Age of AI. April 2019.
26. Artificial Intelligence: the bumpy path 46. NASA. Adapted from V&V of Neural
through defense acquisition, . calhoun.nps.edu.2017. Networks for Aerospace Systems. June 12, 2002.
27. Panel, Governance as a Catalyst of AI. 47. DARPA - Robust AI initiative.
FCW Workshop. April 2019. www.darpa.mil/work-with-us/ai-next-campaign .
28. Explanation and Justification in Machine 48. ML Integrity: Four Production Pillars for
Learning: A Survey. cs.columbia.edu. 2017. Trustworthy AI. forbes.com. January 19, 2019.
29. European Standard. VHF air-ground 49. IBM. Trusted AI. research.ibm.com.
Digital Link (VDL) Mode 2; etsi.org. [2010. research.ibm.com/artificial-intelligence/trusted-ai/.
30. A Survey on AI for Network Routing 50. EU's General Data Protection Regulations:
Problems pdfs.semanticscholar.org. Privacy & Cybersecurity Are Converging hbr.org.
31. LDACS: future aeronautical 51. IBM. Trusted AI. research.ibm.com.
communications for ATM ieeexplore.ieee.org. 2014 research.ibm.com/artificial-intelligence/trusted-ai/.
32. Updated LDACS System Specification.
ldacs.com. SESAR JOINT UNDERTAKING, 2017.
33. C-band digital aeronautical communication Disclaimer
for UAS. ieeexplore.ieee.org. 2017. The concepts presented in this paper are those
of the authors and have not yet been established in
34. FRAME STRUCTURE OF THE CDACS
policy. These views do not represent the official
ieeexplore.ieee.org. 2018
position of the FAA.
35. APNT STUDY. ist.psu.edu. ICAO
NAVIGATION SYSTEMS PANEL (NSP), 2010. Digital Avionics Systems Conference 2019
September 8-12, 2019
36. U.S. Patent: Secondary Surveillance
RADAR. July 6, 1999.
37. Realities and challenges of nextgen ATM:
the case of ADS-B. ieeexplore.ieee.org. 2014.

Authorized licensed use limited to: Auckland University of Technology. Downloaded on June 05,2020 at 19:57:20 UTC from IEEE Xplore. Restrictions apply.