220 Just ordinAry robots
data to determine where to station trac cops—not in areas with slow
trac, but in locales where people were speeding. And even though
the company makes the data anonymous and aggregates it for the
purposes of providing trac information, the controversy led to a
public outcry and a government investigation in which the company
was cleared of violating Dutch data protection laws.*
Data can be decrypted by the suppliers of navigation systems at
the demand of investigators when a car is found at a crime scene. But
what should be done with the information when a driver ees after
a trac accident, for example? Should the police have the ability to
trace this car? After all, it goes one step further than just being able to
follow a motorist on a map. Or does privacy prevail and so the system
guarantees anonymity, such as the TomTom navigation system (see
Section 5.4)?
In cooperative systems, another problem arises because dierent
parties have to exchange data with each other. e questions are who
should own or control the data, how should time data and location
data relating to the driver be stored, to what ends will the data be used
and how can it be ensured that the privacy of the motorist is guar-
anteed. Now that cooperative systems have proven their worthwhile
application in real life, it is high time that politicians and information
lawyers started discussing these questions.
As the use of autonomous cars expands, maintaining individual
privacy within the transportation system may become even more
arduous (cf. Glancy, 2012). A lot of data would be gathered by
autonomous cars in order that cars could be driven safely, but these
data can be misused. Numerous stakeholders have commented on
the high value of data that would be gathered by in-vehicle com-
munications platforms about the vehicle itself and its driver. For
example, insurance companies would be interested in individual
driving habits and retailers in attracting motorists to their locations
(Anderson etal., 2014). As we have already mentioned, law enforce-
ment agencies have a considerable interest in using these data.
In California, the Consumer Watchdog’s Privacy Project direc-
tor John Simpson states that the autonomous vehicles legislation
*
http://data-informed.com/protable-lessons-tomtoms-brush-data-privacy-controversy/.
221who drives the CAr?
should “provide that driverless cars gather only the data necessary
to operate the vehicle and retain that data only as long as neces-
sary for their operation. It should not be used for any additional
purpose, such as marketing or advertising, without the consumer’s
explicit opt-in consent.*
5.6.3 Security and Safety
As we have already mentioned, advocates state that autonomous cars
will eliminate 90% of crashes, but autonomous cars will introduce
new risks with regard to security and safety, such as reliability, nega-
tive behavioral adaptation, and cyberattacks.
5.6.3.1 Reliability System failures, when a car is operating in an
autonomous mode, could be fatal to vehicle passengers and other
trac participants. All critical components will need to meet high
manufacturing installation, repair, testing, and maintenance stan-
dards, similar to aircraft components. is will probably be relatively
expensive (Litman, 2014). Furthermore, the special equipment,
including sensors and computers, needed for an autonomous car is
already very expensive.
Because of the fear of liability, producers will only put these sys-
tems on the market when they are found to be perfectly safe. ISO
standards are important to reduce liability, since then car manufac-
turers can argue that they are operating at the state-of-the-art level
of the industry and have observed mechanisms for functional safety
(Anderson etal., 2014). For most driving assistance systems, ISO
standards are developed, such as the ISO standard 15622:2010 for
adaptive cruise control systems.
Only after such a norm is created
and much research will producers dare to take this plunge. A lot of
the research in recent years has focused on increasing the safety of
driver assistance systems, and therefore, these benets are presented
as unique selling points. We can see a huge increase in these systems
in all car models. is trend is also expected for cooperative systems,
*
http://www.consumerwatchdog.org/blog/dmv%E2%80%99s-autonomous-vehicle-
regulations-must-protect-users%E2%80%99-privacy.
http://www.iso.org/iso/catalogue_detail.htm?csnumber=50024.
222 Just ordinAry robots
which are relatively new. For that purpose, large-scale studies and
ISO standards are needed.
5.6.3.2 Negative Behavioral Adaptation Cooperative driver assistance
systems can also lead to negative behavioral adaptations, such as a
reduced level of attention and overestimating the systems advan-
tages, with the result that some of the positive eects are nullied.
It appears that the ACC system not only has a major positive eect
on road safety but also leads to higher speeds and shorter head-
ways (Dragutinovic, Brookhuis, Hagenzieker, & Marchau, 2005).
Another possible unintended negative eect—with respect to mixed
trac—may be that road users without such a system will anticipate
an assumed behavior of cars with such a system or that by guess-
work or imitation they behave as if they do have such a system,
for example, drivers of nonautonomous cars may be tempted to join
autonomous car platoons. In addition, these systems can lead to de-
skilling, so that driving ability may deteriorate. is can lead to
dangerous situations at times when the (semiautonomous) car does
not respond autonomously and control should be taken over by a
driver who has become less road savvy.
Research conducted by Virginia Tech Transportation Institute
and General Motors in cooperation with the U.S. Department of
Transportation Federal Highway Administration in 2011 to under-
stand the factors that impact the eectiveness of alternative con-
cepts of operation for cars with limited ability autonomous driving
features (specically, adaptive cruise control capable of maintain-
ing a set speed and headway, cooperative driving and lane center-
ing capable of autonomously following a single lane on a highway)
has addressed concerns that drivers (1) have become overreliant
upon the systems; (2) operate the systems outside of design param-
eters; or (3) are not aware when the systems are not operating as
intended. Evidence also suggests that drivers tend to spend less
time looking at the roadway in front of them, have longer o-road
glances when operating the semiautonomous car compared to the
nonautonomous car, and do not adequately anticipate when the car
needs to operate in a nonautonomous mode (Llaneras, Salinger,
& Green, 2013). According to a study by Jamson, Merat, Carsten,
223who drives the CAr?
and Lai (2013), drivers demonstrate increasing symptoms of
fatigue with a car in an autonomous mode and become more heav-
ily involved with the in-vehicle entertainment. Drivers, however,
do demonstrate additional attention to the road as trac condi-
tions become more congested, implying that these responsibilities
are taken more seriously as the supervisory demand of a car in an
autonomous mode increases.
5.6.3.3 Cyber Security Cooperative systems (and autonomous cars)
have to deal with the security of the information and communication
network. Cooperative driving, for example, necessitates both com-
munications hardware and a link to the engine management system
so that the vehicle can control its own speed. A disadvantage of this
is that the system is fragile and the car could become the victim of
hacking attempts. U.S. researchers at CAESS (Center for Automotive
Embedded Systems Security) have shown that it is possible to hijack
and take over full control of the car (Checkoway etal., 2011). In the-
ory, malicious people could take over a highway node, causing colli-
sions and trac disruptions. e European research project Preserve
(Preparing Secure Vehicle-to-X Communication Systems), started in
2011, deals with the development and testing of a security system.*
Securing data is complicated because cryptology increases the infor-
mation ow, and the available bandwidth is restricted. In the United
States, the National Institute of Standards and Technology is cur-
rently developing a framework to improve the critical infrastructure
of cyber security, and recommendations that stem from this frame-
work may be incorporated into automated- and connected-vehicle
technologies (Fagnant & Kockelman, 2013).
5.6.4 Better Drivers
Autonomous cars developed by, for example, Google and the Free
University of Berlin make the driver redundant. Many researchers
see the autonomous car as a method of preventing trac accidents,
for conscious or unconscious human error is involved in almost all
*
www.preserve-project.eu/.
224 Just ordinAry robots
trac accidents. Several studies show that in more than 90% of cases,
accidents occur due to human error and that only 5%10% are the
result of deciencies in the vehicle or the driving environment (see,
e.g., Broggi, Zelinsky, Parent, & orpe, 2008; Dewar & Olson,
2007; National Highway Trac Safety Administration, 2008).
Autonomous vehicles have continuous complete attention and focus,
keep within the speed limit, do not get drunk, abstain from aggres-
sive behavior, and so on. In addition, Peter Sweatman, director of the
University of Michigan Transportation Research Institute, states that
“[h]umans are not suited to monitoring tasks like driving [because]
human attention is easily diverted.* In fact, the researchers at the
institute state that the solution to preventing trac accidents is not
more speed limits, airbags or distracted driving policies, and so on,
but getting humans out of the loop. But before the human factor can
be switched o in trac, the autonomous vehicle must be thoroughly
tested in the actual dynamic trac before safely functioning on the
road. Levy and Murmane state that “[a]s the driver makes his left turn
against trac, he confronts a wall of images and sounds generated by
oncoming cars, trac lights, storefronts, billboards, trees, and a traf-
c policeman. Using his knowledge, he must estimate the size and
position of each of these objects and the likelihood that they pose a
hazard. … Articulating this knowledge and embedding it in software
for all but highly structured situations are at present enormously dif-
cult tasks” (Levy & Murmane, 2004, p. 28). Or, if an autonomous
car is programmed to faithfully follow the trac regulations, then it
might refuse to drive in automatic mode if, for example, a headlight is
broken, or it might come to a complete stop when a small tree branch
pokes out onto a highway because crossing a continuous line is pro-
hibited whereas humans would simply drift a little into the opposite
lane and drive around it (see also Lin, 2013).
According to Smith (2012), autonomous cars cannot yet be claimed
to be signicantly safer than those with human drivers. In his analy-
sis, which uses a Poisson distribution and assumes the accuracy of the
crash and mileage estimates, he concludes that for autonomous cars
to be declared safer, they would need to drive 1.17 million kilometers
*
http://www.engin.umich.edu/college/about/news/stories/2013/november/
driverless-connected-cars.
..................Content has been hidden....................

You can't read the all page of ebook, please click here login for view all page.
Reset
18.218.135.227