Вы находитесь на странице: 1из 30

Sensor Systems in Autonomous Vehicles

How the design of the sensor systems interact with existing


infrastructure and conditions

By
Aleah Young
Brent Gruenke
Jay Jiewen Luo
Kate Cowley

March 14, 2017


Contents

Research Report
Introduction 1
History and General Info 2
Sensors, Road Quality & Vehicle Behavior 4
Communications 7
Legality 11
Conclusion 14

Evaluative Statement 15

Annotated References 18

Interview
Introduction & Bio 23
Transcript 23
Research Paper

Introduction
Autonomous vehicles (AVs) are estimated to reduce the number of traffic accidents by 90%,
reduce congestion, and increase mobility. Partially autonomy in vehicles can already been seen
with common features such as parking assistance, cruise control, and brake-assist. Companies
such as Uber, Google, and Tesla are already experimenting with prototypes of fully autonomous
vehicles.
With autonomous vehicles quickly becoming a reality, how this technology will be integrated
into the current infrastructure is still in question because the current road, signage, and legal
systems are all developed for human drivers. Many are confident that this technology will make
the roads safer, provide transportation to underrepresented populations, and influence the
economy positively. However, because the promise of this technology has not yet been proven,
there are still many concerns over the development and implementation of fully autonomous
vehicles. This paper investigates autonomous vehicle technology, with a specific emphasis on
the sensors used to perceive the outside world, to see how it integrates into current
infrastructural systems.
This paper discusses a brief history of autonomous vehicles and their navigational technologies,
followed by a description of several common types of autonomous vehicles sensors and their
abilities to sense their environments. Along with this, autonomous vehicle communication and
environmental sensibility is examined. Finally, the current legal infrastructure is developing in
anticipation of fully autonomous vehicles, and this paper considers possible changes to the
legal infrastructure.

1
History of AV and General AV Info
In order to understand the current state of autonomous vehicle technology, it is important to
understand the past. Autonomous cars are path dependent; in the social sciences path
dependency means that history matters. Without decades of research on sensors, computing,
and AI, relatively safe autonomous vehicles would never have become a reality.
In the 1860s an English engineer, Robert Whitehead, invented the first naval torpedo that
could maintain a consistent depth and direction, and had a built-in gyroscope to accomplish this
(Britannica, 1998). Whiteheads torpedo was a big step towards the automation of navigation-
based technology. He proved that a vehicle carrying a payload could be intelligent enough to
consistently self guide itself and hit its target. Sailboats were most likely the first self-steering
vehicle capable of carrying passengers. It worked with a wind vane controlling the tiller on the
sailboat, keeping it on course. Soon after sailboats gained the auto-tiller, airplane autopilot was
invented (Weber, 2014). The first autopilot worked by using a heading indicator and attitude
indicator to control the elevators and rudder on the plane. This allowed the plane to maintain a
consistent heading and reduce the cognitive load on the pilot (H.W, 2012).
Bringing the same level of autonomy of sailboats and airplanes to cars has proved challenging,
because cars require sensors and artificial intelligence (AI) to perceive and communicate with
their environment. In other words, cars need more than just a heading to be autonomous.
There are more physical obstacles and entities to interact with on the road, compared to the
open water or sky, which allowed autopilot in sail boats and planes to be invented earlier on. In
order for cars to become autonomous they had to successfully see and sense their
environments. The advent of computer vision in the 1960s at MIT allowed for this to happen.
Professor Huang discusses the current state of computer vision in his 1996 paper. Computer
vision aims to build autonomous systems which could perform some of the tasks which the
human visual system can perform (and even surpass in many cases) (Huang, 1996). Giving
computers the ability to see is what allowed for autonomous vehicles to become a reality.
Some of the first research on autonomous vehicles was done in the Carnegie Mellon Robotics
Institutes Navlab. The goal of Navlab was to apply computer vision, sensors, and high speed
processors to create vehicles that drive themselves (Carnegie Mellon University, 2014). In
1985, the Department of defense sponsored research into laser range finding and vision and
navigation systems (Kweon, Herbert & Kanade). Laser range finding and vision sensors allowed
the cars to sense their environment and react appropriately. Carnegie Mellon has worked on
autonomous cars for 33 years now, starting out autonomously navigating a few centimeters per
second to having a self driving Cadillac SRX that avoids pedestrians, can navigate the current

2
road infrastructure, and obeys traffic laws. It does all of this while looking just like a traditional
Cadillac.
One of the first successful examples of autonomous driving was in 1986 when Ernst Dickmann
from Bundeswehr University Munich outfitted a Mercedes-Benz van with cameras and other
sensors. The van was able to navigate on the highway at high speeds without issue. This set a
precedent for AV technology and many other engineers iterated upon Dickmanns work. Along
with the research Carnegie Mellon was doing, this helped pave the way for autonomous vehicle
technology (Weber, 2014). While this might not necessarily be the first self driving car ever
credited with achieving autonomy, it is seen as a major milestone.
Outside of academia, auto-pilot and partial autonomy have become commonplace in the
automotive industry. The Santa Clara Law Review defines three levels of system integration:
warning and information that helps the driver to maneuver the vehicle in certain situations,
assisted driving such as adaptive cruise control, and finally fully autonomous driving which is
defined as "all driving tasks automated for all use cases" (Beiker, 2012). Current players in
autonomous vehicles include Uber, Google, Tesla, Mercedes-Benz, BMW, Audi, and Delphi (de
Looper, 2015). This is just an example of the current market as almost every current car
company is trying to prepare for the future. Each of these new autonomous cars use slightly
different navigation systems but have the same goal of making driving easier.
Autonomous cars are path dependent; in the social sciences path dependency means that
history matters. Without decades of research on sensors, computing, and AI, stable
autonomous vehicles would never become a reality. Furthermore, the advent of computer
vision allowed development to be done on autonomous cars. The threat of war influenced
Robert Whitehead and the researchers at Carnegie Mellon to develop autonomous
technologies as well as dream that cars would one day be self driving. Other factors that have
steered the path of autonomous vehicles have been physical infrastructure, the development
of history, social contexts, and legal restrictions. Autonomous vehicles are an increasingly
popular topic. This can be attributed to large companies developing autonomous features.
Mainstream automotive and technology companies are testing their version of the
autonomous vehicle.
Since the 1980s research and development has been happening on autonomous vehicles. The
recent boom of mainstream companies investing in AV technology has brought attention to the
topic. This brief insight into the history of autonomous vehicles hopes to shed light onto the
possible future path that autonomous vehicles will navigate down.

3
Sensors, Road Quality & Vehicle Behavior
According to a public perception report summarized by Daniel, et al., the public expects
autonomous vehicles to be able to operate safely. Only based on solid and reliable machine
vision, good road quality, and preset behavior systems can autonomous vehicles understand
the environment correctly and ensure human safety.

Machine Vision
One of the most important requirements for autonomous vehicles is that they should be able to
sense the surrounding environment and respond accurately during their operation on the road.
According to Jain, et al., machine vision is described as the technology and methods used to
provide imaging-based automatic inspection and analysis for such applications as automatic
inspection, process control, and robot guidance, usually in industry. There are many different
kinds of sensors used for developing machine vision for autonomous vehicles such as laser
sensors, radar sensors, ultrasonic sensors, and vision based sensors. They all have their own
characteristics and role in navigation.

Laser Sensors
Laser Sensors, also commonly known as LIDAR (Light Detection and Ranging), utilize the
Doppler effect in order to detect the distance of the object. According to Doppler, the Doppler
effect happens when an increase (or decrease) in the frequency of waves as the source and
observer move toward (or away from) each other. The distance can be calculated by
subtracting the time that the reflected signal was received from the time the signal was
emitted. Daniel et al. claims that laser sensors are one of the sensors that can assist
autonomous vehicles in actively sensing the environment, however, they usually have a low
spatial resolution, slow scanning speed, and little effectiveness during rain and fog, delaying of
the scanning output.

Radar Based Sensors


Like laser sensors, radar-based sensors also use the Doppler effect, yet they can continuously
scan the environment with waves of higher frequency. This makes it much more resistant
towards weather like rain and fog. However, using the same physical principle as laser sensors
causes radar-based sensors to share the low spatial resolution and scanning speed.
Furthermore, radar-based sensors are costly compared to laser sensors. According to an article
written by Alex David, the cost of a radar sensor system is approximately $125-$150 for long
ranges and $50-$100 for short ranges, compared to around $8000 for a reliable LIDAR sensor.

4
Ultrasonic Based Sensors
Ultrasonic sensors are similar to radar and sonar in their capabilities of navigating obstacles.
Their goal is to detect the distance of objects. In comparison to radar, sonar, and LIDAR
sensors, ultrasonic sensors are relatively cheap. Ultrasonic or sonar sensors, utilize the echo
principal. They work by transmitting electrical signals to ultrasound and detecting them as they
bounce off objects, and is generally done with transmitters, receivers, and transceivers.
Despite its simplicity and low-cost, the sensor allows resolutions of better than 1 mm to be
obtained in quiet conditions. The sensor output is updated every 20ms and an additional digital
output allows an easy implementation of smoothing techniques of the car computing system
(Carullo, Parvis).

Vision Based Sensors


Vision based sensors are passive sensors that are responsible for important functionalities of a
self driving car, such as traffic sign recognition and obstacle identification by scanning the image
it captured within its range. They can provide information in a non invasive way and do not
require any modification to the environment, in contrast to other sensors. However, compared
to radar based sensors, they are not effective enough during fog, rain, or darkness.

Requirements for road quality


It has been concluded that some autonomous vehicles seem to have a higher requirement for
road quality than regular cars operated by a human driver do, while other autonomous vehicles
are more adaptive and robust.
Inability to recognize the road and signs is a significant safety concern. As previously discussed,
autonomous vehicles mainly rely on vision based sensors to recognize traffic signs and line
markings by scanning the images captured by the sensor. Therefore, corrupted or unclear
markings, special weather conditions such as rain or fog, or low light conditions will yield the
sensors unable to recognize these signs and markings properly, resulting in inaccurate readings
of the environment.
Kuan, et al. performed a test on an autonomous robotic vehicle on a road, and the vehicle was
operating at the speed of 19km/hr. During the test, they found that during certain
circumstances, including curved roads that only have one side of the road visible to the vehicle,
larger shadows casted onto the road led to incorrect readings, and can usually be detected and
corrected by the vehicle. However, there were exceptions where the vehicle was confused
and stopped advancing until manually corrected by a human operator.
Sometimes the vehicle cannot recognize the road because of poorly maintained infrastructure.
For example, Alexandria Sage mentioned that Volvo's North American CEO, Lex Kerssemakers
claimed that their semi-autonomous vehicle could not recognize the lane markings because

5
they are poorly painted. He called out for a better maintained road system so that there will be
a better foundation for autonomous vehicle technologies to operate on.
Autonomous vehicles sensors do have a high requirement towards the road quality, and they
can be negatively influenced by extreme weather conditions or complex environments.
However, pre-programmed vehicle behaviors can make the vehicle less affected by sensorial
and environmental limitations. For example, Reynolds, et al. designed and discussed an
autonomous characteristic for vehicles with a set of preset behaviors such as obstacle
avoidance, path following, steering. Thus, instead of operating based on the scanning of the
environment, the autonomous vehicle can use a combination of environmental scan data and
preset behaviors to operate more efficiently.

Autonomous vehicles rely on sensors and behavior systems in order to operate. And their
performance can be optimized with well constructed and maintained infrastructure such as
road systems. Different kinds of sensors work together on autonomous vehicles so that they
can provide machine vision. However, even with the progression of the sensor technologies,
the sensors effectiveness is still limited by real world elements such as weather with poor
visibility or snow and rain. In order to surpass the limitations of the sensor system, pre-
programmed behavior systems can help the vehicle pre-adjust their own behavior based on the
limited data provided by the sensors. However, it is also important to notice that autonomous
vehicle requires a good infrastructure, road maintenance this case, in order for it to operate on.
That is because the vehicle might not be able to recognize the road or the signs if they are not
well painted and maintained.
From a technical perspective, autonomous vehicle still have a long way to go but they are much
closer to being a reality than thirty years ago. Sensors such as LiDAR, radar, ultrasonic, and
vision based sensors allow for autonomous vehicles to have machine vision. There are still some
concerns about vehicles sensors adapting to real life traffic conditions and situations. There is
some concern about road maintenance so autonomous vehicles should be designed to not only
account for traffic but for poor weather conditions in the future.

6
Communications
A large component of autonomous vehicle environmental perception is through communicating
and sharing data between other vehicles and infrastructure. There are several directions of
information exchange involved in vehicular travel, both with autonomous and traditional
human-driven technologies. In autonomous vehicle systems, navigation and other critical safety
functions are enabled through dedicated short-range communications (DSRC).

Directions of Information Exchange


The two primary directions of information exchange involved in vehicular travel are vehicle-to-
vehicle and vehicle-to-infrastructure. Vehicle-to-vehicle includes all communication between
the vehicles, drivers, and even pedestrians or bicyclists.

Vehicle-to-Vehicle Communication
For autonomous vehicles, vehicle-to-vehicle communication is one of the dominant methods of
sensing external data such as the location, speed, direction and braking status of other
vehicles. It includes any information exchanged between the technical or human components
of a vehicle. In addition to sharing status data, vehicles can also share information collected
from their individual camera and sensor systems, to help other autonomous vehicles have a
more complete vision of the external environment and provide vision in any blind spots. This
type of communication also presents interesting challenges in replicating certain forms of
driver-based communications.

Fully autonomous vehicles are "capable of sensing their environment and navigating without
human input through embedded technology and sensors (Kgler, Majdalani & Viereckl).
However, there is a level of driver-based communication that so far has not been replicated in
way compatible with autonomous systems. According to Dr. Don MacKenzie, University of
Washington Assistant Professor of Transportation Engineering, human factors remain one of
the lesser understood components of autonomous vehicle design. Driver-based, vehicle-to-
vehicle communication refers to the subtle gestures and eye contact that are frequently used
to communicate critical information such as right of way. Autonomous vehicles lack the ability
to read and respond to eye contact and hand signals to make appropriate judgements (Heaps,
2017). These subtle gestures are difficult to reproduce and transmit through existing vehicle-to-
vehicle wireless protocols. (D. MacKenzie, personal communication, February 2, 2017).
Recently, at the 2017 Consumer Electronics Show, Nissan has released a new technology,
Seamless Autonomous Mobility (SAM) that introduces a human element into autonomous
technology (Heaps, 2017). SAM relies on the vehicles cameras to assess the current situation,

7
and uses artificial intelligence to decide whether control of the vehicle needs to be escalated to
a human Mobility Manager to proceed. Nissan claims that as the artificial intelligence system
matures, the reliance on human intervention should decrease (Heaps, 2017).
Vehicle-to-vehicle communication enables rapid exchange of safety-critical information and
most likely represents the reality of how autonomous vehicles will communicate in the future.
With this type of communication, costs are diffused across new vehicle buyers rather than
forcing local and federal governments to make costly changes to infrastructure to enable
vehicle-to-infrastructure communication discussed in the next section (D. MacKenzie, personal
communication, February 2, 2017).

Vehicle-to-Infrastructure Communication
Vehicle-to-Infrastructure communications represent the bulk of traditional, driver-based
communications. It includes any road signs, lane markings, traffic lights, or other external
sources of traffic direction. Currently, all vehicle-to-infrastructure communication elements
have been installed under the traditional arrangement of drivers interpreting the information
(Redefining Mobility: Autonomous & Connected Vehicles, 2015). There is an embedded
understanding that each driver will assess the situation, including signage, and process that
information accordingly (Kgler, Majdalani & Viereckl). Additional infrastructure elements have
been proposed that do not require a human drivers interpretation, though not all are in
agreement that this is necessary or wise.
Several developers claim that specialized infrastructure will be required in order to roll out
autonomous vehicles. They believe that infrastructure must be altered in order to
"accommodate autonomous cars' weaknesses, like their inability to interpret a construction
worker's hand gestures (Ng & Lin 2017). Some suggestions for alterations include wireless
beacons or wired roadways.
These alterations will have a significant cost. For that reason, some believe that any
autonomous vehicle technology that relies on dedicated infrastructure to support
communications is doomed to fail. In addition to a large fiscal cost, specialized infrastructure or
lanes have an opportunity cost- If you take away a general purpose lane today, you are
imposing a huge cost on everyone who uses those lanes today because they are going to be
forced into fewer lanes and experience greater congestion (D. MacKenzie, personal
communication, February 2, 2017).

Dedicated Short-Range Communications


DSRC was specifically developed to support safety applications and communication between
automated vehicles (vehicle-to-vehicle) and external infrastructures (vehicle-to-infrastructure)
for navigation and collision reduction ("Intelligent Transportation Systems - DSRC: The Future of

8
Safer Driving Fact Sheet", 2017). It is a specific protocol for high-speed, high security wireless
communication.
The Intelligent Transportation System (ITS) is a joint program within the US Department of
Transportation that researches and develops smart transportation technologies. The ITS has
been allocated by the Federal Communications Commission (FCC) a 75 MHz of spectrum in the
5.9 GHz band for use with vehicle safety and mobility applications, i.e. vehicle-to-vehicle
communication ("Intelligent Transportation Systems - DSRC: The Future of Safer Driving Fact
Sheet", 2017).
As an isolated technology, DSRC is widely considered an appropriate choice for communicating
and sensing information between autonomous vehicles. It is compatible with existing
infrastructure, low latency, highly reliable, and interoperable ("Intelligent Transportation
Systems - DSRC: The Future of Safer Driving Fact Sheet", 2017).
However, not everybody believes that DSRC use is safe for consumers. Some proponents doubt
that manufacturers will limit DSRC use to safety and navigational purposes only, and instead
use this technology for mobile payments and other for-profit uses (Alleven, 2016). Others are
unconvinced that data shared over the network will be safe from hackers. Ultimately, the use of
DSRC to sense and share relevant data between autonomous vehicles continues to be tested.

Privacy and Security


Vehicle-to-vehicle communication primarily involves sharing data elements such as position,
speed, and road conditions in order to create a situated awareness for each automated vehicle
(Redefining Mobility: Autonomous & Connected Vehicles, 2015). Determining what data to
share, to whom, and how to protect it remains an ongoing issue.
The US Department of Transportation insists that no personal information will be collected, and
no identifying information will be stored (Connected Vehicles and Your Privacy, 2017). Though
certain commercial applications of autonomous vehicles, where vehicles require special
licenses, may in fact need personalized identifying information in order to be compliant with
other laws.
DSRC channels will have built in authentication and privacy features, ("Intelligent
Transportation Systems - DSRC: The Future of Safer Driving Fact Sheet", 2017) yet there are
concerns in addition to being shared between vehicles. This data is potentially at risk of being
exposed to other agencies and communicating across DSRC channels will create new
opportunities for cyber hackers and identity thieves (Alleven, 2016).
Despite insistence that no personally identifying information will be shared, stored, or
communicated across DSRC networks, some believe that autonomous vehicle technology will
enable governments to automatically regulate weight and speed limits (Kgler, Majdalani &
Viereckl). They claim that allowing government agencies to automatically and conditionally

9
respond to data collected in autonomous vehicles overrides the statement by the US
Department of Transportation that vehicle-to-vehicle safety messages will not be used by law
enforcement to cite or enforce laws (Connected Vehicles and Your Privacy, 2017).
The US Department of Transportation states that consumer privacy will be preserved through
laws and regulations, as well as the Fair Information Practice Principles, though the system may
still be vulnerable to malicious hackers (Connected Vehicles and Your Privacy, 2017). The US
Department of Transportation (DOT) has not fully satisfied those with security and privacy
concerns. Many believe that in addition to the US DOT guidance, robust security protocols will
be needed as autonomous vehicles would be vulnerable to different types of security attack
and... everything that applies to the Internet in terms of security concerns would apply to in-
vehicle communications" (Anderson, Kalra, Stanley, Sorensen & Oluwatola, 2014).

Autonomous vehicles engage in multiple directions of communication: vehicle to vehicle and


vehicle to infrastructure. Vehicle to vehicle communication is currently the primary means by
which autonomous vehicles sense data about their surroundings. Information in this exchange
is transferred mostly over dedicated short-range communications channels. DSRC has been
specifically developed for this form of communication, though concerns remain over the
privacy and security of data in this exchange. The US DOT currently cites Fair Information
Practice Principles as a means of preserving consumer privacy, and further laws and regulations
are currently being developed to address privacy and security concerns.

10
Legality
While AV technology is developing rapidly, it is very important to make sure that the
technology is being incorporated into the current legal infrastructure, and that the future
infrastructure is designed to incorporate AVs. There are many different aspects to look at,
including laws, government spending, policies and regulation, and safety and liability. These
aspects have been discussed, and since AV is such a new technology, many of them are still up
in the air. When discussing AV and legal issues, it is also important to keep in mind the level of
automation that is being discussed. As with many of the other types of vehicles previously
discussed, most of todays vehicles on the roads are already partially automated. For example,
traditional vehicles that implement cruise control and self-parking, or fully autonomous
vehicles such as the Cadillac SRX developed by CMUs Navlab is fully autonomous. It is
important to examine what is being done in the legal sense currently, and what changes are
needed in the future to accommodate for complete automation. The past and present societal
factors that went into designing the legal infrastructure are taken into account in this analysis
of integrating AVs into the current legal infrastructure. This is a very complicated issue and
integrating AVs into the current legal infrastructure will be a gradual and uneven process due
to the changes that will be needed to solve problems that arise during the transition to AVs
(Tech Policy Lab, University of Washington). Since the transition to full automation will be such
a lengthy process, this section focuses on creating a system that supports both traditional and
automated vehicles for the foreseeable future.

Transportation Laws and Safety Regulations


This section focuses on the state of US traffic laws and the future of transportation and legal
infrastructure, while discussing possible changes that will need to be made for the adoption of
AVs. The first topic to take into account is the current state of transportation laws. These laws
are based on the fact that vehicles will be managed by a competent driver. Some of these laws
will become obsolete once fully automated vehicles are on the road. For example, driver laws
like the fact that drivers cannot be distracted or under the influence. In an AV, there will not be
a driver, therefore these laws do not apply to this situation. Although, the sensors being
developed will mimic a competent driver by providing the same ability to perceive the
environment and make smart decisions based off of this data. Therefore, perhaps these laws
can be reworded instead of completely removed. To combat liability confusion that will arise
when something in an AV goes wrong, there may need to be more laws on the upkeep of AVs,
to attempt to keeping malfunctions to a minimum. When looking at the transition in Seattle,
Washington, there are various ways to go about it. The municipality could promote AVs on the
streets, give permission to AVs on the streets, or apply cautious and limiting regulations to AVs

11
on the streets (Tech Policy Lab, University of Washington). Whichever strategy is chosen will
dictate the laws and regulations applied to traffic. Though preparing for AVs before they are
integrated into society is smart, in the interview with Don Mackenzie, he explained that he
believes most policy changes would be reactive. Laws are not the only thing that will be
changing to accommodate for AVs, traffic safety policies and regulations will have to change
too, along with insurance plans and actions.
Along with advances in society and infrastructure, modifications will also need to be made to in
traffic safety regulations. There are already some policies being made by US departments
regarding AVs. For example, the Department of Transportation (DOT) and the National Highway
Traffic Safety Administration (NHTSA) released guidelines for states to follow in reference to
AVs that said: new Federal Motor Vehicle Safety Standards (FMVSS) for AVs must be set and
complied to, any AV defects must be investigated and managed, and the public must be made
aware of AV technology. The Federal Trade Commission and the White House also released
data privacy guidelines in the Consumer Privacy Bill of Rights (Tech Policy Lab, University of
Washington). Municipalities like Seattle have a lot of power that they can use to make systems
more AV friendly, and increase the efficiency of their traffic by doing things like editing traffic
laws and minimizing empty parked AVs (Tech Policy Lab, University of Washington). All in all,
certain cities in the US are currently experimenting with integrating AVs into their current
transportation systems, and the US government is issuing guidelines for the municipalities to
follow.

Insurance and Liability


Along with these governmental regulations, there will be many changes to liability. Civil liability
is the responsibility of payment for damages as assigned by the court. This raises many
questions in the case of damage being done when an AV is part of the equation because more
parties are involved. There are many questions currently proposed, for example: If a collision
involves only AVs, who is at fault? Is it the current owner of the AV? Is it the company who
designed and engineered the AV? Or if a collision occurs between a traditional vehicle and an
AV, did the driver make a mistake or was the AV not engineered to drive with traditional
vehicles? A mix of AVs and traditional vehicles creates a very complex environment that will
need new systems to determine liability. System failures can occur due to the many different
interactions between vehicles and infrastructure, and because the technology will be so deep, it
may become impossible to trace back to the technical failure (SAFESPOT), which may call for a
completely new way to manage liability.
With these issues in hand, then comes along the issues of insurance. Since there is not any
statistical data on the risks involved with AVs at this point, insurers will have to cover unknown
risks (Margriet van Schijndel-de Nooij, et al., 2011). Some other problems insurers are going to
have to face are: competition among other insurance companys sensitive relationships with
clients, investment and administrative costs, and regulation (Margriet van Schijndel-de Nooij, et

12
al., 2011). It has also been recommended that insurance companies will have an agreement
with AV production that if they AV malfunctions, liability will be moved from the insurance
company and to the manufacturer (Margriet van Schijndel-de Nooij, et al., 2011). In conclusion,
insurance and liability are very complex issues when it comes to the case of AVs, and to be able
to prepare for these types of issues, one must understand AV technology. It will be a group
effort between governments, insurance companies, and AV manufacturers to create a system
that works for everyone.

Emergency Services
When incorporating AVs into our current infrastructure, there are a lot of government costs
involved. Though AVs may reduce government costs in the long run by decreasing the number
of collisions and reducing the need for police services, the government first has to invest in new
system. One of the proposed solutions for this is developing multiple alternative revenue
sources; for example, using miles traveled or AV lane tolls and taxes on AVs (Tech Policy Lab,
University of Washington). On the subject of government involvement, there is also the
question of how emergency vehicles and police activity will change. AVs will provide a lot of
benefits to police like decreasing the amount of traffic stops making it cheaper to do stake outs,
or even building in the ability to let police halt or move AVs (Tech Policy Lab, University of
Washington). If police can have control over other cars, that can bring up the ethical issues of
control of personal property, data and privacy, and abuse of the power. After police cars have
been automated, that could make way for a greater amount of police automation. Some police
work is already automated in our society, by using signs that display speed and speed bumps.
With more automation it is possible that even more tasks like these could be automated. These
benefits can also be applied to other types of emergency vehicles like fire trucks or ambulances,
and greatly increase their speed and efficiency, saving many lives. Overall it seems that most
emergency vehicles can benefit from automation, but there are drawbacks and a large cost of
implementing this technology.

In conclusion, there are many different aspects to the legal and governmental changes that will
need to be made to make way for AVs. There are the issues of laws and policies, traffic safety
regulations, insurance and liability, government costs, and changes to government sectors. A lot
of these issues are connected to other areas of infrastructure like the physical layouts of cities
and their transportation systems, along with the way that people interact in society. A
multidimensional perspective will likely be needed to connect this technology to society and
solve the problems that arise during this process.

13
Conclusion
This paper provides a brief overview of the history and current state of autonomous vehicles,
defines various modern communication and sensor systems, evaluates many different types of
communication pertaining to autonomous vehicles, and discusses the current legal
infrastructure and how it does not support fully autonomous vehicles on the road. Each of
these sections examine challenges on improving AV technology and incorporating them into the
current infrastructure. There are many different types of sensors to choose from in making AVs
better at perceiving their environment, and often times these sensors work together to provide
a complex computer vision. Certain sensors have advantages over others in certain situations.
Following this, when looking at physical infrastructure and AV communication, there is
uncertainty in changing the current infrastructure and changing the way that AVs function. Each
option has its drawbacks and costs, and there is still much uncertainty in implementation.
Future development of specialized infrastructure is highly dependent on a large number of real
world variables, and has the possibility to greatly influence how AVs communicate. There are
also many privacy and security concerns that must be taken into account when designing this
technology and proposing the integration of AV technology in the current infrastructure.
Lastly, the current legal system is very outdated when it comes to AVs, and many changes will
need to be made to laws, traffic and safety regulations, insurance and liability, and
government-controlled transportation practices. This is a multifaceted subject, and the process
of integration will most likely move more slowly than the technology develops. When
understanding issues surrounding AV sensors and infrastructure, its important to consider
aspects of history, technology, and society in order to gain a complete understanding of the
technology.

14
Evaluative Statement

After investigating the content and context surrounding autonomous vehicle (AV) sensors and
infrastructure, it is abundantly clear that their is a lot of uncertainty on the topic. AV technology
will have a widespread effect not only on the people purchasing the vehicles but on people who
dont purchase them. It seems inevitable that autonomous vehicles will become commonplace
in at least some specific industries in the future. It is unclear how soon fully autonomous
vehicles will be fully integrated into Americas transportation systems. Some believe it will be
sooner than expected and others think it may be a long time. It could take many years for
public perception to endorse AVs or it could happen fairly rapidly depending on public
perception. Since there's so much uncertainty surrounding autonomous vehicles it is
challenging to think about them without considering both the drawbacks and advantages of
them.
Some scholars believe that the application of autonomous vehicle will reduce traffic accidents
since most decisions are made by cars sensor, behavior systems, and artificial intelligence. We
believe that the built-in algorithms will minimize human factors as well as errors. However,
these algorithms are not that entirely reliable enough yet, at least for now. They are getting
closer every year. Most of the behavior designs and AI systems are still going through the
design and testing phase. And there are a lot of issues have been discovered during the field
tests, such as inability to operate on terribly maintained roads. As a team, we agree that
autonomous vehicles are safer than regular cars but it also depends on many factors in each
scenario. AVs have the potential to save thousands of lives if the technology is created
correctly. Is the AV on a bad road in the middle of a storm? Perhaps an experienced driver
would be better at operating the vehicle in this situation rather than an artificial intelligence. If
the onboard sensors or computer systems fail, the car will not be as safe as a car under the full
control of a human.
The pre-designed algorithms also led to another question, which is one of the hottest topics
about autonomous vehicles, about whether they are able and reliable to make ethical
decisions. This raises the question, can we trust autonomous vehicles? With complex machines
such as cars there are a lot of things that could go wrong and AV manufacturers will have to get
it right in order for people to trust them. Human trust of AI, computer vision technology, and
designers will have be paramount for this technology succeeding initially. Someone has already
died while using Teslas autopilot feature. The average consumer will not want to think about
these things when they get into their self driving car. This could affect how the algorithms are
created and tested. Furthermore, Google is one of the leading researchers into AV technology.
Are they creating an AV just so they can collect more data on their customers? Who is gets
access to your car's data? These are questions that may influence peoples trust in the

15
technology. Our team is conflicted in whether or not to trust AV technology mostly due to the
uncertainty in the industry.
Will the rise of autonomous vehicle free peoples hands from the steering wheel? Our answer is
yes. The impact will be more than just liberating drivers hands. While some commuters will
have free time to catch up on reading or to watch a movie, other commuters will likely feel
negative impacts during their commute. There will be more people willing to commute because
they wont actually need to be driving during commuting. We can also assume self driving
cars give underrepresented populations more freedom. For example underrepresented
populations
such as elderly people, handicapped, or injured people who otherwise wouldnt feel
comfortable driving could have elevated mobility freedom. This could mean more congested
traffic in cities since more people are given access to the roads. It could also mean that taxes on
gasoline and car registration are raised to accommodate all the new autonomous cars on the
roadways.
It is important to note that while autonomous vehicles will give some groups more freedom,
they also exhibit some barrier to entry. Poor people still may not be able to benefit from this
technology if self driving features are only rolled out in luxury vehicles. Poor people may only
see the negative aspects of self driving cars. It would be nice to see public transportation or ride
sharing services incorporate autonomous vehicles so that a larger population could benefit
from the population. Perhaps taxi services will use autonomous vehicles in the future.
Government will have a large role in controlling how autonomous vehicles are taxed, perceived,
and controlled.
If autonomous vehicles become popular, will there be more vehicles and what will be their
impact to the environment? Since driving will be easier, more people may choose to have a car.
This will likely increase the greenhouse gases in the atmosphere contributing to climate change.
When autonomous vehicles first gained popularity, one of the selling factors was energy
efficiency. A computer system could more accurately control the fuel usage in the car, making
for a more fuel efficient experience. Additionally, many people associate electric cars or hybrids
with self driving cars. It is uncertain if the widespread adoption of autonomous vehicles would
be more harmful or less harmful to the environment. If the sharing economy adapt the
autonomous vehicles well, autonomous vehicles might reduce traffic congestions and emission
by cars. However, all these assumptions are dependent on how industries such as trucking,
sharing economy, and logistics will adapt to the autonomous vehicles in the near future.
To conclude, autonomous vehicles are an up and coming technology that has been being
researched and developed for over the past thirty years. They promise a safer driving
experience and have the potential to use computer vision to save thousands of lives. While they
can save lives, they also can take lives. This raises ethical implications of designing a computer
that will be programmed to make life or death decisions. Undoubtedly, owners of AVs will have

16
the potential for more enjoyable commutes travel more, but there are also negative impacts of
having more drivers on the road. Lastly, it is unclear how low-income families and the
environment will be impacted by the widespread adoption of AVs. Autonomous vehicles have
many speed bumps to overcome in the future, but perhaps one day they will be a reality. AV
technology undoubtedly has a large role in the future, and the technology behind them is
impressive but not without controversy. Perhaps one day everyone will be cruising around in
KITT from Knight Rider, until then team ScrewDrivers will be waiting.

17
Annotated References
Alleven, M. (2016). FCC seeks comment on spectrum sharing techniques in the 5.9 GHz band.
Fiercewireless.com. Retrieved 26 February 2017, from http://www.fiercewireless.com/tech/fcc-
seeks-comment-spectrum-Sharing-Techniques-5-9-ghz-band
This article discusses the FCC process of adopting Dedicated Short-Range Communications
(DSRC) in the 5.9 GHz band. It is published in FierceWireless- a publication centered around
issues in wireless communication across multiple industries.

Anderson, J., Kalra, N., Stanley, K., Sorensen, P., & Oluwatola, O. Autonomous vehicle
technology (1st ed.). Washington, DC: RAND. 2014
Published by the RAND Corporation, a nonprofit institution that provides research and
analysis to inform policymaking in the united states. This goal of this publication was to
identify the most salient policy issues, and provide tentative guidance to policymakers. It
contains information about the technology, laws and regulations, enforcement, and other
issues related to autonomous vehicle technology.

Beiker, Sven A. and Ryan Calo (2012). "Legal Aspects Of Autonomous Driving". Santa Clara Law
Review 52 n. 1145. Web.
Dr. Sven A. Beiker works at the Center for Automotive Research at Stanford University. M.
Ryan Calo works at the Center for Internet and Society at Stanford University. This article
briefly discusses various levels of autonomous driving technology and the legal barriers and
other non-technical challenges that face autonomous vehicle technology. The authors
suggest several courses of action to explore and develop a legal framework that supports
autonomous vehicles.

Britannica, E. (Ed.). (1998, July 20). Robert Whitehead. In Encyclopdia Britannica. Retrieved March
05, 2017, from https://www.britannica.com/biography/Robert-Whitehead-British-engineer
Encyclopedia article about Robert Whitehead the British engineer who created the first self
guided torpedo. Provides a brief history of Whitehead and his accomplishments.

Carullo A.; Parvis M. (2001). An Ultrasonic Sensor for Distance Measurement in Automotive
Applications. In: IEEE SENSORS JOURNAL, vol. 1 n. 2, pp. 143-147. - ISSN 1530-437X
Research paper discussing new ultrasonic sensor technology and testing in a motor vehicle.
Discussion of optimization and cost of ultrasonic sensors.

Connected Vehicles and Your Privacy. (2017) (1st ed.). Washington, D.C. Retrieved from
http://www.its.dot.gov/factsheets/pdf/Privacy_factsheet.pdf
The US Department of Transportation has published this fact sheet on privacy concerns with
using the Dedicated Short-Range Communications (DSRC) network for autonomous vehicle

18
communication. It takes a positive stance, and highlights all of the mechanisms by which
DSRC preserves and protects consumer privacy.

Davies, A. (2015). Turns out the hardware in self-driving cars is pretty cheap. Wired Blog, 4(22), 2015.
An article discussing the pricing of the hardwares of autonomous vehicles. The discussion
focused on the development and manufacture cost of sensors, processors, and actuators.

De Looper, C. (2015, October 08). The Complete List of Self-Driving Cars in Development. Retrieved
March 6, 2017, from http://www.techradar.com/news/car-tech/the-complete-list-of-self-driving-
cars-in-development-1306382
A brief overview of 2015's self driving car manufacturers. Images and summaries of the cars
provided in the clickbait style article.

Doppler, C. (1842). Ueber das farbige Licht der Doppelsterne und einiger anderer Gestirne des
Himmels: Versuch einer das Bradley'sche Aberrations-Theorem als integrirenden Theil in sich
schliessenden allgemeineren Theorie. In Commission bei Borrosch & Andr.
This paper mainly summarized Dopplers principle that the observed frequency changes if
either the source or the observer is moving, which later has been coined the Doppler effect.

From 0-70 in 30. (2014). Retrieved March 5, 2017, from


http://www.cmu.edu/homepage/environment/2014/fall/from-0-70-in-30.shtml
A history of Carnegie Mellon University's research and development of autonomous
vehicles. Discusses how the Navlab started research in 1984 and how has developed a full
autonomous Cadillac SRX that has strong enough AI to avoid pedestrians, merge, react to
traffic lights, and take highway ramps. Highlights major achievements accomplished by CGU
in the area of autonomous vehicles.

Heaps, R. (2017). CES 2017: Nissan Overcomes Tricky Autonomous Vehicle-to-Human Communication
Issue - Autotrader. Autotrader. Retrieved 6 March 2017, from http://www.autotrader.com/auto-
show/CES2017
At CES 2017, Nissan presented a new autonomous vehicle technology specifically designed
to work with mixed traffic (human and autonomous drivers). This article discusses how that
technology recreates and responds to traditional human-led communication.

H.W. (2012, August 30). George the AutoPilot. Historic Wings. Retrieved March 5, 2017, from
http://fly.historicwings.com/2012/08/george-the-autopilot/
History on Lawrence Sperry, the inventor of the first autopilot in airplanes. Explanation of
how Sperry developed a gyroscopic automatic stabilizing device. A discussion of the history
of autopilot in airplanes.

Huang, T. S. (1996). Computer Vision: Evolution and Promise. CERN EUROPEAN ORGANIZATION FOR
NUCLEAR RESEARCH-REPORTS-CERN, 21-26.

19
T.S. Huang discusses and analyzes the history of computer vision and unpacks the concept
in an academic paper. HUang is from the University of Illinois at Urbana-Champaign. A
definition and history of computer vision is presented as well as some applications for
usage. Huang claims computer vision is difficult to achieve but the future promises many
large opportunities for advances in computer vision.

Intelligent Transportation Systems - DSRC: The Future of Safer Driving Fact Sheet. (2017). Its.dot.gov.
Retrieved 25 February 2017, from http://www.its.dot.gov/factsheets/dsrc_factsheet.htm
The US Department of Transportation has published this fact sheet on dedicated short
range communications (DSRC) the type of communication that is to be used in autonomous
vehicles in vehicle-to-vehicle and vehicle-to-infrastructure communications. The eligible
developers, usage, and applications of DSRC in autonomous vehicles are briefly discussed.

Jain, R., Kasturi, R., & Schunck, B. G. (1995). Machine vision (Vol. 5). New York: McGraw-Hill.
An introductory book to machine vision. Basic concepts are introduced with only essential
mathematical elements. The details to allow implementation and use of vision algorithm in
practical application are provided, and engineering aspects of techniques are emphasized.

Kgler, Ulrich, Fadi Majdalani, and Richard Viereckl. Trucking To The Future: How GCC Governments
Can Open The Road For Autonomous Trucks. 1st ed. strategy& (Formerly Booz & Company), 2015.
Web. 12 Feb. 2017.
This article discusses the growing freight trucking industry, the potential benefits of
autonomous fleets of vehicles, and suggests ways countries can support and prepare for
autonomous vehicles. The authors are Dr. Ulrich Kgler, Fadi Majdalani, and Dr. Richard
Viereckl, who are all partners with Strategy& a global strategy consulting firm. Dr. Kgler is
positioned in Dubai and specializes in technological trends in the Middle East. Mr. Majdalani
is positioned in Beirut, and leads the firms engineered products and services division. Dr.
Viereckl is positioned in Frankfurt, and has a background in the automotive industry where
he has worked with a number of different automotive companies.

Kuan, D., Phipps, G., & Hsueh, A. C. (1988). Autonomous robotic vehicle road following. IEEE
Transactions on Pattern Analysis and Machine Intelligence, 10(5), 648-658.
The field test summary of a two-ton autonomous vehicle. The vehicle in this paper mainly
rely on camera sensors to identify the surroundings

Kweon, I. S., Hebert, M., & Kanade, T. (1985). Sensor Fusion of Range and Reflectance data for
Outdoor Scene Analysis (Publication). Pittsburgh, PA: The Robotics Institute.
A report detailing the results of using sensors to obtain range and reflectance data. Usage of
sensors to detect outdoor scenes such as trees and roads. Experiments and research
sponsored by the Department of Defense carried out by researchers at Carnegie Mellon
University.

20
Ng, Andrew and Yuanqing Lin. "Self-Driving Cars Won't Work Until We Change Our Roads - And
Attitudes". Wired. N.p., 2017. Web. 18 Feb. 2017.
The authors, Andrew Ng and Yuanqing Lin work for Baidu, Inc (a Chinese web services
company) in Baidus Institute of Deep Learning. They are involved in a variety of research
and development projects and are proponents of automated vehicle technology. This article
discusses the authors believe that in order to adopt autonomous vehicles, the government
and community will need to make changes to existing infrastructure to accommodate the
new technology. It draws a historical analogy to the adaptation of railways.

Redefining Mobility: Autonomous & Connected Vehicles. (2015).


This video is a short interview between Casey Dinges, the Sr. Managing Director of Public
Affairs, Membership, Leadership & Programs at the American Society of Civil Engineers and
Randy Iwasaki, the Contra Costa Transportation Authority Executive Director. In it, they
discuss autonomous vehicles as a means to redefine mobility and touch on how the
technology will interact with existing road systems.

Reynolds, C. W. (1999, March). Steering behaviors for autonomous characters. In Game developers
conference (Vol. 1999, pp. 763-782).
This paper presents solutions for one requirement of autonomous characters in animation
and games: the ability to navigate around their world in a life-like and improvisational
manner. These steering behaviors are largely independent of the particulars of the
characters means of locomotion. Combinations of steering behaviors can be used to
achieve higher level goals. This paper divides motion behavior into three levels. It will focus
on the middle level of steering behaviors, briefly describe the lower level of locomotion,
and touch lightly on the higher level of goal setting and strategy.

SAFESPOT Integrated Project. (2010). Cooperative Vehicles and Road Infrastructure for Road Safety.
Retrieved from http://www.safespot-eu.org/
SAFESPOT creates dynamic cooperative networks where the vehicles and the road
infrastructure communicate to share information gathered on board and at the roadside to
enhance the drivers' perception of the vehicle surroundings.

Sage, A. (2016). Where's the lane? Self-driving cars confused by shabby U.S. roadways. Technology
news, Reuters, 2016
An article discusses how the poor road maintenance in the U.S will affect the future
development and operation of autonomous vehicles.

Van Schijndel-de Nooij, M., Krosse, B., Van Den Broek, T., Maas, S., Van Nunen, E., Zwijnenberg, H.,
Schieben, A., Ford, N., McDonald, M., Jeffery, D., Piao, J., Sanchez, J. (2011, June 29). Definition of
necessary vehicle and infrastructure systems for Automated Driving. Retrieved from http://vra-
net.eu/wp-content/uploads/2014/12/SMART_2010-0064-study-report-final_V1-2.pdf

21
European study report on the necessary infrastructural changes to make in order to
incorporate automated vehicles. Has an automated vehicle introduction, state-of-the-art,
Vienna Convention, liability, reliability, control units, x-by-wire, sensor systems, positioning,
applications, key players, cross border driving/standards, and human driver interaction.

Weber, M. (2014, May 08). Where to? A history of Autonomous Vehicles. Retrieved March 5, 2017,
from http://www.computerhistory.org/atchm/where-to-a-history-of-autonomous-
vehicles/#disqus_thread
An in depth history from the Computer History Museum. Discussion of Robert Whitehead's
torpedo, sail boats, airplanes, and W2 missiles. Details about "The Dream" of self driving
cars and how the idea is not new. Making smart cars is difficult and computing had to
improve before they could be designed well. Sensing, processing, and reacting are the main
functions of self driving cars. Early pioneers of self driving cars and current manufacturers
are discussed briefly.

22
Expert Interview
Introduction & Biography
Don MacKenzie is an Assistant Professor of transportation engineering, and leads
the Sustainable Transportation Lab at the University of Washington Department
of Civil and Environmental Engineering. He holds a Ph.D. in Engineering Systems
from the Massachusetts Institute of Technology and has authored a number of
publications about transportation issues and clean energy. His background
includes over a decade of experience with clean energy and vehicle research &
development, as well as technical and policy analysis for the Union of Concerned
Scientists in Washington, DC.

Currently, his research includes explorations into clean energy, vehicle automation, and consumer
behavior.

Transcript
This informal interview was conducted on Friday, February 24th with the team (Brent Guenke, Aleah
Young, Jay Jiewen Luo, and Kate Cowley) and Dr. Don MacKenzie. It took place in a conference room in
More Hall, the Civil Engineering Building at the University of Washington Seattle campus and lasted
approximately 30 minutes.

This transcript has been paraphrased at points for conciseness.

Brent Gruenke (BG): Could you tell us a little bit about your yourself and your current projects?

Don MacKenzie (DM): OK. So in the area of connected and automated vehicles. My interests are
mostly about how do they influences peoples behavior. particularly do they stimulate additional
travel? So does it make travel safer, cheaper, more convenient. We think the short answer is yes
but it is what we are trying to dig into.

Aleah Young (AY): That would be hard test

DM: Yeah its hard to test but we have some fun ideas. But it's highly uncertain. Weve looked
more broadly at the energy impacts which include the demand response as well as operational
efficiency and the changes in vehicle design that are enabled in vehicle design.

BG: Fantastic, could you tell us more about energy efficiency.

DM: Three, four, five years ago when people started talking about vehicle automation it was very
utopian. A lot of people were focusing on energy efficiency. Energy was kinda like not really a
driving force here it was more about safety, improving mobility for underserved groups, it's about

23
productivity, and convenience. Its really not being driven by energy. That being said one of the
many benefits people were talking about was this will make our transportation systems so much
more efficient. Because number one we can optimize the energy our vehicles use especially
internal combustion engines use a lot more energy moving at low speeds and in stop and go traffic.
People were like wow we will eliminate congestion, get rid of traffic lights, and use a lot less
energy. And this will make like Uber like more attractive than owning your own car. Everyone will
use just Uber and this will open up all these other options. It was very utopian. But a lot of that was
focused on the energy efficiency piece. We dug into that and said yes theres a lot of opportunity in
energy efficiency but theres a lot of risk for increased energy demand. We dont know how those
balance out and we dont know to the extent people will replace private vehicle ownership with
autonomous mobility services.

Kate Cowley (KC): I had a follow up questions with that. You mentioned there was profound
implications for energy. One of the things we were curious about was- maybe I made this
assumption in my own mind- maybe that autonomous fleets would be electric.

DM: Yeah, everyone makes that leap. Im not sure why. There are potentially some synergies but
its not obvious that one follows from the other. Reasons electric might be more attractive when
you have automation one is - OK when you have automation and connectivity - like a smart car
with cloud connectivity can anticipate where charging is available it can predict things like weather
and traffic it can alleviate range anxiety- that is pushed off on the vehicle the driver doesnt have to
worry about it. So range anxiety goes away and this is seen as a big obstacle to AV adoption.
Another reason is the infrastructure costs. Maybe when you get into self driving -a vehicle can drive
off and recharge itself. Finally, when you get to on-demand fleets mobility of the service
fleets- that does a couple things. The biggest thing is you probably accumulate miles faster. If this
vehicle is in service- I dont know its probably not in service 24 hour a day. But if its running 10
hours a day at 200 miles a day, 70 thousand miles a year the vehicle will have a 2 to 3 year life. So
when you look at the savings from operating an AV. Those are fuel cost savings. Typically those are
spread out so that is spread out over 15 year life looking out only 15 years isn't much. So
compressing all that useful life in a couple years there are more savings then there is a better
economic case for electrification. But again it doesnt necessarily hold. This is something that
should be looked at as an opportunity. This is what folks are calling
ACES or FAVS. Autonomous connected electric and shared and fleets of autonomous vehicles
electrics and shared. So this idea of having sharing and electrification along with automation is seen
as a desirable outcome and an opportunity but we run a risk of assuming what could happen and
what wed like to happen and then we mistake that for what is going to happen - which happens a
lot.

KC: So they are compatible and perhaps they compliment but they are independent.

DM: Yeah there is not necessary link between the two.

BG: Going off of that, do you think there needs to be a large change in infrastructure and power
grids for AV to happen?

DM: Do you mean like the transportation infrastructure?

24
BG: Yeah, all inclusive infrastructure for example the highway system thinking about Autonomous
trucking.

DM: Any autonomous vehicle technology that relies on dedicated infrastructure is doomed to fail.
We have tried to promote alternative technologies that have infrastructure requirements,
particularly with alternative fuels. Its just too expensive to build this. In particular, if you need
special lanes of some kind, that will make sense at some point in the future. But there is a huge
opportunity cost. If you take away a general purpose lane today, you are imposing a huge cost on
everyone who uses those lanes today because they are going to be forced into fewer lanes and
experience greater congestion. Other than test fleets, there are no autonomous vehicles on the
road. There will be massive public pushback on any effort to take away right of way and reallocate
that to autonomous vehicles. I dont see that being viable. So you need a technology thats
compatible with the existing infrastructure. At some point in the future, when they are a
substantial portion of the fleet, then it might make sense. I dont see that as being viable in the
near time.

AY: When you were saying that them depending on the infrastructure would be bad, did you mean
they should be adaptable?

DM: Theres some people that think we need to have V-to-I, vehicle to infrastructure, connectivity
for these to work. Theres others that think they need a dedicated lane if youre going to allow
these things to run autonomously. When it comes to dedicated V-to-I connectivity, weve done that
in some places and its incredibly expensive. Youre basically putting out Wi-Fi routers every 500
feet or so along the road.
A smarter approach is to get some vehicle to vehicle (V to V) connectivity so the costs are diffused
across new vehicle buyers. So you put out a mandate that you have V to V capabilities but you
don't have large expense wiring up the roadways. The other thing is if you use commercial
networks-makes sense for anything not safety critical. anything that doesn't need millisecond level
and incredibly high reliability that makes a lot more sense to send to commercial networks. That
makes more sense to the send through v to v DSRC dedicated short range communication if youve
heard about that. The idea is you want something that is compatible with existing infrastructure
this is (as you may know) is a generally accepted principle for architecting complex systems. As you
want the system as you build it out, to be stable along the way you want it to be stable along the
way. You want incentive for adoption every step along the way.

BG: Which areas of industry outside of automotive that you think will be most impacted by AV
technology?
DM: Retail is going to have a huge impact. It already is just with online purchasing. Thats adjacent
to transportation.

BG: How do you anticipate public policies will change in response to or anticipation of autonomous
vehicles?

DM: Probably mostly in response. Some states are now putting in laws that allow for testing, and to
regulate testing. In general, I think a lot of the policy response will be reactive because thats the
way these things go. I think what were going to see is that when you have self-driving, even
partially, vehicles, the risk you run is that you make people want to travel more. You make people

25
more willing to take long commutes. The reason why is the cost of the drivers time is the single
largest component of the total cost of operating a vehicle. So you take that away, and youve
basically taken away the biggest portion of the cost so people will travel more. So how do you or do
you even want to stop that? We find that when people travel a lot they impose external costs on
other travelers. Theres congestion from using roadway capacity, which in the short term is finite.
By traveling more you are depriving others of the quality of roadway thats available. Basically,
increasing travel times for others. Its an externality and its unpriced. I think the most important
policy response is probably going to be that we need to price our roads or price travel in some
other way. Either have congestion pricing, tolling, or a crude instrument like VMT (Vehicle miles
traveled) fees. But thats pretty crude because these external costs are very sensitive to time and
space. I5 is heavily congested, but not at midnight. If somebody is accumulating all their miles in off
peak hours then its not that big of a deal. But if you are increasing peak hour demand, then you
are increasing these external costs and were going to need to start pricing that. Will policymakers
do it? I dont know. The writing is going to have to be on the wall. I think things are going to need
to get pretty bad before jurisdictions move on it.
An opportunity for policymakers is to act in the near term and to basically impose these prices on
self-driving cars. Thats not without its risks. This risk being, the why are you doing it? To manage
the congestion, and travel demand, and energy consumption. But what that ends up doing is acting
as a deterrent to the development of these products in general. You are taxing this product that
offers huge safety benefits. That could be an undesirable outcome. But this road pricing regime can
be put in place if you start from the premise that all vehicles are going to be automated in the
future, then the tax policy can be rolled out today before there is a large base to oppose it. That
would be the strategy that I think a smart policymaker would undertake. But they dont want it to
be so onerous that they deter the technology altogether, you just want people to use it more
smartly.

BG: How do you anticipate existing road structures will support AVs? We talked about this a little
bit already.

DM: Its a little bit circular logic. Successful AVs will be the ones that can use the existing road
infrastructure. Theyll support them by working for them. Theres not much there. Theyll support
them, otherwise vehicles wont be there. Some of the state DOTs are looking at nuts and bolts
stuff, like what does this mean for lighting and striping and maintenance and repair on the roads.
Theres some issues there. Yeah, we can build a vehicle that doesnt need wired roads or DSRC
units every quarter mile, but that car still needs to orient itself to where it is on the roadway and
the way it does that is by looking at the paint and if the paint isnt there then it doesnt know
where its going. Basically you need to have good maintenance because human drivers can adapt to
poor conditions but automated vehicles sensors may not allow the to do so.

KC: This may be not the question to ask here but, we wanted to know These automated vehicles
arent being in a road built for them, theyre being put in a world built for us. And theres great V:V
communication, but do you know if currently these test vehicles are put into place to deal in the
ideal world or in reality.

DM: This is why Google brought their car up to Kirkland to do some testing here because they
realized they were doing all their testing in Silicon Valley in ideal weather conditions. Its frankly
very suburban there. Theyre not testing in downtown San Francisco. The idea is its a pretty

26
controlled, highly engineered environment thats very car oriented. The developers realize they
need to get out and test these things in a multitude of environments. They know, but everything
starts in the lab and gets rolled out gradually.

The other thing thats interesting about how they interact with human drivers, is that we have
already have vehicle to vehicle communication and thats the drivers. So vehicle-to-vehicle and
vehicle-to-pedestrian communication and its all through the drivers eye contact and subtle
gestures. A lot of that is really hard to build. Its easy to build a wireless protocol to send data, but
how do you replicate a nod and a hand gesture. Theres also people who are working on that. At
the Automated Vehicle Symposium last summer, some of the developer and some of the suppliers
were presenting work theyve done on how vehicles communicate to, say, a cyclist, Yes, I see you
there. So how you get that confirmation to a pedestrian or a cyclist. Theyre working on that, and
my sense is the human factors are one of the less understood pieces of this overall. I assumed that
was going to be all you wanted to talk about being from HCDE

BG: Thats definitely interesting. Do you guys have any more questions

KC: I was hoping if you would mind expanding on your studies about online shopping and freight
and how that will influence or be influenced.

DM: I havent studied it, I just have opinions on it. We looked at freight a little bit in the work that
weve done on this but it was primarily in long-haul. So looking at how intercity freight, trucking,
competes with rail and air. If you reduce the cost of the driver, how does that affect demand. In the
urban setting. That will impact the supply chain. But in the urban setting, theres this question of
what will this do. So we reduce the cost of delivery, essentially, and so once we reduce the cost of
delivery by reducing the cost of the driver, how much more demand does that stimulate. And thats
what we really dont understand. We know that on a trip for trip basis, its generally much more
efficient to have trucks doing urban deliveries, going out and serving 10,50, 100 homes on a route
because you can optimize that. You can have a truck visit every one and optimize that route and
that hugely cuts down on VMT, substantially cuts down on energy requirements compared to all of
those people making an individual trip to the store.
Whats not know is how many more trips are generated. When I go to the grocery store, I make
one big trip. When I shop on Amazon, I have prime so I dont have to consolidate my shipping so,
Im like ship it, ship it, ship it. And in the end I still end up going to the grocery store but its just to
pick up one bag of things that I forgot to order or they didnt have. Im still making that trip, Im
buying less stuff, but Ive generated all those individual deliveries as well. What we dont
understand and is an open area of research is basically that shopping behavior and how they
interact and the rate of substitution of personal trips to the store. We know directionally, but we
dont really have a good handle.

BG: Youve been really helpful, thank you for doing this. Do you have any specific resources you
recommend us looking at?

DM: You mentioned up front your goal is to what? Articulate again what you said at the start

BG: Were trying to unpack the content and context. Content is how the thing works, the history of
it, and what it does. Context is how it fits into society and the different controversies.

27
DM: Your final question is what other resources?

BG: Yes

DM: Have you talked to Ryan Calo, hes in the Law School and you should definitely talk to him.
Hes a professor in the law school who has been doing work in this area. Hes sort of a leading
figure in robotics law. Definitely talk to him.
You could also talk to a couple guys who have a small non profit, Steve Marshall and John Niles.
They might have something to say. Their agenda is to get these things integrated in the local
transportation system in a way that conserves energy and preserves or enhances mobility. I dont
know who else. You could talk to the DOTs. If you wanted more on what the DOTs are thinking they
need to do, they are some people but they may not be the most useful. Theyre mostly thinking
what they have to do to keep the system running.

Ryan would be the number one guy to talk to.

All: Thanks

28

Вам также может понравиться