Autonomous vehicles can learn from aviation automation, which perfected autopilot systems but still requires pilots due to risks. Two plane crashes resulted from pilots relying too heavily on autopilot and lacking manual flying skills. Self-driving cars are being developed, like Tesla's Level 3 system, but most roads cannot support full automation. Before widespread adoption, regulations must ensure drivers can resume control and address risks of over-reliance on automation from aviation lessons.

1. From the Sky to the Ground: How Autopilot in Planes will help Self-
Driving Cars
G. Mei Raley
November 13, 2016

2. Abstract:
Self-driving cars are no longer an idea seen in pop culture created through visual effects. In the past ten
years the development of technology to achieve full autonomous vehicles has grown exponentially.
However, before autonomous vehicles can be completely implemented into society, we must look
towards automation in aviation as a precedent for precautions to take. Using lessons learned from
aviation, potential challenges in autonomous vehicles are considered.

3. Introduction
Self-driving cars are nothing new. In fact, they have been around a lot longer than most
people think. The idea of autonomous vehicles gained widespread attention during the 1939
World’s Fair at GM’s Futurama exhibit.[11]
The concept of robotic cars slowly developed in the
following decades until the 1980s. The first autonomous cars appeared in 1984 with Carnegie
Mellon University’s Navigation Lab project. Also in that same decade, Ernst Dickmanns, while at
Bundeswehr University of Munich, developed the Eureka Prometheus Project in conjunction with
Mercedes Benz. Since then, multiple companies including Google, Tesla, and General Motors
have been working on the technology to truly perfect the self-driving car. Google has had multiple
successes with their technology and says autonomous cars will be available to the public in 2020.
Elon Musk says Tesla can do it by 2017.[12]
To fully realize a self-driving car, it is important to
look at other areas of transportation where automation has been successfully applied in order to
help develop the technology and risk assessment for cars.
Aircraft Automation
Gyroscopic Autopilot
Nine years after he first powered flight by the Wight Brothers, aviation pioneer Lawrence Sperry,
developed the first autopilot system. On June 18, 1914, during the Concours de la Securité en
Aéroplane (Airplane Safety Competition) in France, Sperry unveiled his device, a gyroscopic

4. stabilizer.[2]
By controlling all three axes of an
aircraft – yaw, pitch, and roll – the device
allowed Sperry to successfully fly the aircraft
with his hands off the controls, even allowing
him to balance on the wings of the machine.
Since then, the gyroscopic autopilot system has
grown to act as the basis for many aircraft
control systems.
Automatic Flight Control System
A complete autopilot system is called the
automatic flight control system (AFCS). At the
heart of the AFCS is an operating system designed to gather the intelligence needed from all parts
of the plane in order to control it and to correspond with the electronics systems for
communications, navigation, collision avoidance, and weather.[17]
With lightning communication
speeds, the AFCS is able to make multiple changes across multiple surfaces and make adjustments
as needed, many times faster and smoother than a human pilot could.
Autopilots in aircrafts exist to relieve pilots during long periods of time, or to take over
when more precise handling is required, such as times of low visibility. However, they are not
perfectly autonomous. Even the best autopilots do not automate the taxi and take-off phases of
flying and 99% of all landing procedures are still manual.[16]
The US Federal Administration of
Aviation (FAA) regulations state that at least two crew members must remain in the cockpit at all
times, and either the pilot or co-pilot must always monitor the controls, even if they are not actively
handling them.[3]
Overall, autopilots are not systems that completely replace pilots. Patrick Smith,
Figure 1: The three axes of an aircraft
and the set of controsl used to manage
each one.

5. an active commercial airline pilot since 1990 said, “Pilots have some to rely on a somewhat
different skill set over the past 40 years or so. But that’s not to say that hands-on skills have become
any less important.[16]
” Two aircraft disasters highlight this statement and the effect of two
phenomenon: de-skilling and mode-awareness/mode-error. Deskilling occurs when the skills of an
experienced person deteriorate or atrophy due to disuse. In the case of newer pilots, it refers to the
lack of learning specific skills that used to be taught. Mode awareness is being conscious of the
current and future state of automation, and understanding how the automation should act. The
following two accidents acknowledge the roles both deskilling and mode awareness caused.
Flights 447 and 214 Accidents
On June 1, 2009, all 228 people aboard the Airbus A330 Flight 447 were killed.[4]
Junior
co-pilot Pierre-Cédric Bonin was manning the controls when the aircraft entered an airspace
occupied by storms. Although he was a fully trained pilot with nearly 3,000 flight hours, his
experience was predominantly run on autopilot. Therefore, his lack of skills and training in manual
flight caused him to make a few critical errors that led to the ultimate crash. This is an example of
deskilling, where the younger pilot did not acquire the necessary manual skills to successfully
operate the aircraft.
Four years later on July 6, 2014, Asiana Flight 214 crashed into the seawall at San
Francisco International Airport just short of the runway.[1]
Part of the automatic landing system
wasn’t working that day, so the pilot Captain Lee Kang Kuk was forced to manually approach.
Auto-throttles should have kept the plane at 137 knots, but the plane was only going 103 knots
which led to a rapid loss of altitude. Kuk was new to flying the 777 aircraft and felt unsure in his
actions. In this case, it was the pilot’s dependence on an autopilot system, that system’s

6. malfunction (mode-error), and his deskilling that caused the accident. From these two accidents
and countless others, we can learn about how to better implement automation in cars.
Types of Autonomous Cars
Although true 100% autonomous vehicles have been created and successfully tested, they
have not yet been administered to the public.[5]
Semi-autonomous cars have been around for
decades, with varying degrees of automation in primary control functions. The National Highway
Traffic Safety Administration (NHTSA) defines five different levels of automation in vehicles.[6]
Level 0 is No Automation where the driver is in complete and sole control of the vehicle at all
times. Level 4 is full Self-Driving Automation where the driver is not expected to or unable to take
over controls at any time. The following examples utilize either Level 3 or Level 4 automation.
Google: Level 3 Automation
Google has been in the development of autonomous cars since 2009.[13]
Google’s self-
driving car uses lasers to create an extremely detailed permanent 3-dimensional picture of the roads.
This allows the system not to have to rely on GPS technology so it can spend more power focusing
on moving objects. Sensors placed around the car detects objects and classifies them based on size,
shape, and movement pattern. A test driver is present in all vehicles with the ability to take over
controls (steering wheel, gear change, pedals, etc.) if needed.
Tesla: Level 3 Automation
In 2014, Tesla rolled out its AutoPilot system to all new Model S vehicles.[15]
In
conjunction with adaptive cruise control and lane departure warning, this system created semi-
autonomous drive and parking. In conjunction with FAA regulations, However, AutoPilot does
have some restrictions. It requires the environment to meet certain conditions before the autonomy

7. can be engaged. These include clear lane lines, a relatively constant speed (no tight turns or hills),
a sense of cars around you, and a map of the area you’re traveling through.[8]
Although Tesla has
seen a huge success in its AutoPilot system, more than half the nation’s roads may be beyond its
capabilities. Also, weather, construction zones and detours, and an inconsistency in traffic control
devices adds further complications.[7]
Ford: Level 4 Automation
Ford has been working on their
autonomous vehicle for the past ten years,
but recent news has shown their
commitment to achieving Level 4
automation. CEO Mark Fields confirmed
a timeline of five years to have these
vehicles on the road and that the car would
not feature any “handover” function where
a human would be able to take back
controls.[9]
Along these lines was an earlier report that Ford Motors received a patent for an
“autonomous vehicle entertainment system” which combines a ceiling-mounted projector with a
roll down projector screen right in front of the windshield.[10]
Although such a vehicle has not been
tested on real roads, it is important to note that this is the ultimate goal for a self-driving car.
Figure 2: Ford’s drawing for a patent for a projector
screen setup.

8. Conclusion
Comparing Planes to Cars
Although the rules and skills for flying are vastly different from driving, there is still a lot
of important technical similarities to consider. Compared to cars, no plane has ever implemented
Level 4 automation, nor do they ever expect to, according to reports. Although automation is used
for majority of all flights, there are key areas where a pilot needs to be able to have control. For
Ford to completely remove an individual’s ability to take over controls in extreme situations is
dangerous, especially considering the harmful effects automation has already had in aviation.
When self-driving cars are to be implemented, new laws and regulations will have to be devised
to maximize safety.[14]
As with aircrafts, Tesla also requires its drivers to monitor the vehicle at all
times. However, this is not an automobile law and will be difficult to implement. Additionally,
many kids will grow up with the notion that automation in vehicles is normal, and will never learn
proper safe-driving practices or skills. If an accident does occur, the ignorance of proper safety
procedures is likely to result in more detrimental injuries. The lessons learned from flights 447 and
214 should show that deskilling and mode-error pose a huge threat to self-driving cars.
Preparing for a Future with Autonomous Vehicles
There is no doubt that autonomous vehicles will be introduced into society, drastically
changing the day-to-day on-goings of people everywhere. Thousands of unnecessary deaths could
be prevented. This technology could provide the mobility freedom for otherwise disabled
individuals who are unable to drive. Self-driving cars could even replace Uber and other delivery
services, creating a society-wide ride-sharing phenomenon that would decrease the number of cars
on the road, improving the environment. The amount of free time people would then have to focus

9. on other things could lead to even greater feats of innovation. However, precautions should still
be taken before self-driving cars are fully implemented.
Looking ahead, Level 4 Automation is something that should be met with trepidation. Even
if the technology were perfected, there are still thousands of cars that fall into Level 0 to Level 2
categories. This mixture of semi-automatic and manual drivers will hinder the ability for a
completely self-driving car to be properly used to its potential. Currently, Level 3 Automation still
has a lot of progress to make before it can be carried out in the masses. As seen from the aircraft
accidents of flights 447 and 214, perhaps we shouldn’t be rushing to install that projector just yet.

10. Bibliography
[1] National Transportation Safety Board, "Crash of Asiana Flight 214 Accident Report," 2014.
[2] L. C. W. Sheck, "Lawrence Sperry: Autopilot Inventor and Aviation Innovator," Aviation History,
November 2004.
[3] Federal Aviation Administration, "Automated Flight Control," in Advanced Avionics Handbook,
2009.
[4] W. Langewiesche, "The Human Factor," Vanity Fair, September 2014.
[5] NAVYA, "Navya Tech," [Online]. Available: http://navya.tech/?lang=en.
[6] National Highway Traffic Safety Administration, "Statement of Policy Concerning Automated
Vehicles," 2013.
[7] Lowry, "Experts Caution Self-Driving Cars Aren't Ready for the Roads," The Press Democrat,
2016.
[8] Bradley, "10 Breakthrough Technologies," Technology Review.
[9] M. Murphy, "Coming in 2021: A self-driving Ford car with no steering wheels or pedals," August
2016. [Online].
[10] Davies, "Ford Reckons Your Self-Driving Car Needs a Movie Projector," 2016. [Online].
[11] Vanderbilt, "Autonomous Cars Through the Ages".
[12] Korosec, "Elon Musk Says Tesla Vehicles Will Drive Themselves in Two Years".Fortune
Magazine.
[13] Google, "Self-Driving Car Monthly Report," 2016.
[14] Greenblatt, "Self-Driving Cars Will be Ready Before Our Laws Are," IEEE Spectrum, 2016.
[15] Tesla Motors, "Autopilot," 2016.
[16] R. Nasr, "Autopilot: What the system can and can't do," March 2015. [Online].
[17] W. Harris, "How Autopilot Works," October 2007. [Online].