IEEE Systems, Man and Cybernetics Magazine - October 2019 - 27

OA is more prevalent in the
This is critical as we consider
automotive industry. OA in the
autonomous systems that incorpoOnce available,
transportation industry has been
rate manned-unmanned teaming
autonomous cars are
subject to much criticism due to
in the presence of manned forces.
the associated risks to human life
A policy approach that expands
expected to reduce
as autonomous capabilities are
the definition of autonomous systhe number of road
still not developed to the level
tems must include trust requirethat communities can place comments. Therefore, expanding skill
deaths by orders
plete and absolute trust in them in
sets and training policies to in of
magnitude.
terms of reducing life risks. But
clude evaluation methods from
those risks exist even with humans
other disciplines, such as human
driving vehicles [35], [36]. Once
factors and psychology, will be
available, autonomous cars are expected to reduce the
required. The T&E community is struggling very hard to
number of road deaths by orders of magnitude.
develop effective modeling and simulation capabilities for
Many of the current approaches to develop autonomous
the comprehensive testing of AUSs. Specifically, the abilicapabilities are very data intensive, which may require a
ty to inject constructive and/or virtual targets into live
long time period to collect the crucial information. Generenvironments will be required to evaluate AUS responses
ally, training such systems is performed based on simulatto dangerous scenarios in a safe and efficient manner.
ed training data, but achieving comparable performance
Finally, T&E for cyberautonomy attacks needs to be
when deployed in the real world is extremely challenging.
addressed. In other words, with all of the nuances, the
ramifications of the requirements for cyber-resilience
need to be identified. This is a must-have in terms of the
Nontechnical Challenges
evaluation of an autonomous system in response to adapNontechnical challenges in terms of OA are mostly related
tive and intelligent adversaries.
to policy making and ethical practices for the use of AUSs.
The users of autonomous systems may tend to over-rely on
OA: Challenges, Barriers, and Players
autonomy. There are examples of people blindly following
The concept and related issues of OA suggest that technolGPS guidance and falling asleep behind the wheel while
ogy development alone cannot provide comprehensive
paying reduced attention to the act of driving itself. There
solutions that lead to operational deployment of AUSs.
have been multiple examples where pilots using autonoRather, a multipronged strategy is required to develop an
mous assistance were oblivious to their aircrafts' functionunderstanding of the technical and nontechnical challengality without such assistance. There is a specific need to be
es. Figure 3 illustrates some of the major obstacles to OA.
cognizant of the appropriate occasions for autonomy usage
and the identification of helpful cues that keep users from
losing their facet knowledge [16], [27] and preserve their
Technical and Technological Challenges
intuition of how to handle certain situations. Therefore,
The critical technological challenges of OA differ across
industries. For example, the operations in C2 involve communication between fighter pilots, vehicles on the field,
and air traffic control to create situational awareness and
Quantifying
understanding relative to the intent of a current mission. A
the Degree
transition is occurring in aviation from GPS-based navigaof Autonomy
tion to automatic dependent surveillance broadcast [33],
Technical and
[34]. It is a critical piece of sharing the intent and explainScalability
Technological
ing what must be done to maintain safe separation and the
Challenges
desired routing. The challenge is the ability to interface
T&E
and interact with autonomous systems in ways that are
Procedures
suitable within operational environments/constraints.
Challenges
There are no well-developed means for robots and other
Policy Making
AUSs to understand human intent or to communicate with
humans in natural ways. Shared intent and a shared lanEthical
Nontechnical
guage need to be evolved to fully enable people and AUSs
Practices
Challenges
to communicate in operational settings. Technologies are
improving through the course of time. We are now able to
talk to digital personal assistants, such as Siri, Bixby,
Safety
Alexa, and others; yet engagement in terms of speechbased interaction and fulfilling partnership roles with
humans remains a challenge.
Figure 3. The challenges for OA.
O c tob e r 2019

IEEE SYSTEMS, MAN, & CYBERNETICS MAGAZINE

IEEE Systems, Man and Cybernetics Magazine - October 2019

Table of Contents for the Digital Edition of IEEE Systems, Man and Cybernetics Magazine - October 2019