IEEE Systems, Man and Cybernetics Magazine - October 2019 - 29

the process toward full OA is the development of a new
science and paradigm for autonomy T&E. The proper set
of tests needs to be identified to reach an appropriate
state of assessment on which society and policy makers
can agree. Alternatively, a graceful degradation methodology could be adopted that prescribed an optimal level
of autonomy to reach the adequate and appropriate state.

impact until the 2000s. Therefore, many otherwise useful
ideas were discarded, but the approach turned out to be
the next-greatest outcome in terms of machine-learning
advancement. Machine learning has now been applied
very broadly from image recognition to natural language
processing and is being widely adopted by industry and
government programs. The same construct applies to the
broader technology areas of autonomy and artificial intelligence. It is a challenge to come up with something that is
widely amenable to industry and government.

The Role of Academia in Facilitating OA
Unlike industry, academia is not specifically motivated by
profit and financial growth and can work in close conjuncConclusion
tion with companies to share common ground and collabIn this article, we discussed the challenges and outlook for
orative solutions. For example, academia can perform
maturing autonomy technology from an innovation space
research on the safe operation of AUSs prior to their tranin controlled environments to realsition to industry, involving the
world operations. Legal and moral
prototyping and exploration of the
concerns as well as significant
problem spaces. Academia has the
There are myriad
technological and regulatory barrirequired capability, desire, and
ers that hold back the full-scale
workforce to perform the fundaopportunities for
deployment of a large-scale system
mental research on OA. The hard,
academia to play
were highlighted. A preliminary
unsolved problem is in terms of
definition of OA and its relevance
treating an autonomous vehicle as
a role in in
in autonomous systems was proan engineered system from the
addressing deployable
vided. As autonomous capability
deterministic world and translating
autonomy-related
and reliability increase, the prevathat paradigm to the nondeterminlence of autonomous vehicles in
istic domain while simultaneously
solutions.
our daily sociotechnological lives
providing some insight into the
becomes almost inevitable.
behavior of the AUS. There are
We must accept that the bar
myriad opportunities for academia
for an AUS's performance has to be set at less than
to play a role in in addressing deployable autonomy-related
100%. The evaluation of autonomous systems needs to
solutions. Only a small component of U.S. researchers is
be based on the achievement of acceptable performance
working on autonomy T&E and the development of confor a designed degree of autonomy. However, the bars
cepts of trust in autonomy, indicating the existence of a
for acceptance must be evolved and agreed upon in concritical gap [41], [42]. Academia needs to focus on generatjunction with the policies for ethical employment coning a workforce that is going to have the necessary skills
sidering the specific operational environment and
and capability to provide solutions to some of the probapplications. Therefore, a very strict tailoring mentality
lems in OA.
is required because the behavior of the autonomy algoThere are organizations that work to bridge the gap
rithms in the AUS cannot be generalized. When it comes
between academia and the applicability of academic
to machine-based ethics, for example, a machine's deciresearch products to DoD operational use. They translate
sion making can generally be more predictable than a
the requirements to the academic organizations, mature
human's. This is due to the fact that machines can be
academic products, and apply them to DoD systems. Acaprogrammed to set the specific boundaries for identifydemia must continue to collaborate with those organizaing decisions that need to be made. The social aspect of
tions and industry to quickly transition new advancements
incorporating UGVs and UAVs into the daily human routo the real world. There needs to be clarity regarding the
tine is strongly predicated upon the demonstration and
understanding of operational utility and limitations of the
adherence to strict requirements of policy, safety, and
current technologies and continued research into ways to
availability. Of all the human influences that can get in
assess, adapt, and respond to those limitations. Additionthe way, it comes down to certification and standardizaally, academia must do better in terms of interfacing and
tion. This brings us to the importance of building an
educating the public at large. Basic approaches include
evolutionary definition regarding the specific degree of
outreach to K-12 students and seeking coverage of autonoautonomy, the autonomy itself, and the concept of OA.
my research advances in the local and national news.
There is a huge opportunity for exploring new ideas
Acknowledgment
that companies may not have the incentive to undertake in
We would like to thank Michael Corey from the U.S Air
the absence of a product to sell. For example, deep learnForce Research Laboratory for his insightful suggestions
ing studies began in the late 1980s and had no significant
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