Over the summer of 2019, Transdev, a France-based company that operates
public transportation projects in 18 countries around the world, ran two
autonomous
shuttles connecting Montreal’s Olympic Stadium and Maisonneuve
Market.
Transdev’s 2019 endeavour built on the success of a
project
they had executed the year prior, that drove autonommous shuttles around
the Olympic Park grounds. These kinds of pilot projects for autonomous vehicles
have been
allowed
in Quebec since 2018 and are subject to approval from the Minister of
Transport. The red Transdev shuttles, however, were the first self-driving
vehicles to operate on the streets of Montreal. Equipped with sensors and a
navigation system, the cubic red buses could carry 12 passengers along its
seven-stop, 1.4 kilometre long route and offered free rides every day of the
week from 10 a.m. to 6 p.m. A Transdev operator was always on board, since the
shuttles could not make emergency stops or restart themselves.
More autonomous vehicles like the Transdev shuttles are coming: As technology
develops, we will see more self-driving vehicles on our streets, seamlessly
integrating themselves into society. The advancement is exciting, but the true
impact depends on how the industry, and world around it, will navigate the new
technology.
According to
Quebec’s
Bill 165, unanimously adopted on April 18, 2018, autonomous vehicles
are defined as those that fall into Levels 3, 4, and 5 of SAE International’s
six
levels of autonomy. Cars in Level 0, 1, and 2 of the standard are fully
driven by humans, though they support features such as cruise control and
automatic braking to various degrees. In those cars, the driver must actively
watch the road and monitor the autonomous features, even if they are not
physically pressing the pedals or holding the wheel. Level 3, 4, and 5 cars are
not driven by humans. As the levels increase, the driver has fewer
responsibilities and the cars can operate in a broader range of
conditions.
Companies developing autonomous vehicles require a
variety of environments in which to test them, including private campuses,
public roads, and simulations. A network of lasers, sensors, cameras, radar,
and sonar help the car
develop
a map of its
surroundings. Factors such as
traffic levels, road quality,
and slopes all present different
challenges to self-driving cars.
The car then processes such
information by using complex
algorithms to determine what
actions it should take. To
ensure that all these systems
are working, companies attempt
to train their cars for every
possible situation.
“Cities are just environments
where there’s a whole lot going
on,” Andrew Holliday, a PhD
student in the School of
Computer Science, said in an
interview with The McGill
Tribune. “Things moving
around from all sides, people moving
around, pedestrians, cyclists, so
this certainly makes for a more
challenging environment than a
suburb or a countryside where
there's relatively less to deal
with.”
Holliday is
studying ways to improve public
transit systems using AI. Cities are
useful environments for training
cars, but mistakes in the real world
are costly, as human lives are at
risk. Since 2016, several Teslas in
Autopilot
mode crashed, resulting
in fatalities. Elaine Herzberg
was crossing the street when an
Uber-owned self-driving car struck
and killed her in Tempe,
Arizona in March 2018. However,
testing in self-driving cars in
the situations they will have to
face in the real world is
necessary, so companies rely
heavily on simulations, which
are a safe, affordable, and time
efficient option that come very
close to reality. Waymo—an
autonomous car launched by
Google in 2009 before it spun
off and became its own
company—has driven over 10
billion miles in
simulations. Cruise, a
self-driving car company under
General Motors, runs about 200,000
hours of simulations per
day. Other companies design
specific simulations to improve
their cars’ specific functions
or reactions to situations. For
example, Apple recently used
simulations to train cars to merge
lanes.
However, the value of the
training relies on the quality
of the simulation. Companies
collect data from driving on
physical streets to help make
their simulations more
realistic, from imitating
real-world physics to making
drivers in their simulations act
as the average human drivers
would. Companies can then
analyze how their autonomous
driver reacted to a certain
situation compared to what
actions a human driver took.
Despite all of the advancement,
however, researchers agree that
there is still a lot of work to
do.
“I think the
public perception of how far off
the technology is right now from
being road ready is perhaps
overly optimistic,” Holliday
said. “Technology is still not
quite at the place where it
could be as safe as a human
driver [....] It’s conceivable
that [self-driving technology]
may require theoretical
breakthroughs [...] because the
range of conditions that can
arise out there in the wild on
the street almost inevitably
throws up something that the
engineers won’t have thought
of.”
Quebec’s Bill
165 added provisions regarding
traffic rules, vehicles
standards, and issues insurance
and liability with the intention
of opening the door to pilot
projects in the province.
Additionally, Propulsion
Quebec was founded in
2016 with a goal of “positioning
Quebec as a global leader in
developing and implementing
smart and electric modes of
ground
transportation.”
“I think the
public perception of how far off
the technology is right now from
being road ready is perhaps overly
optimistic,” Holliday said. “Technology
is still not quite at the place where it
could be as safe as a human driver [....]
It’s conceivable that [self-driving
technology] may require theoretical
breakthroughs [...] because the range of
conditions that can arise out there in the wild on
the street almost inevitably throws up something
that the engineers won’t have thought of.”
Another self-driving vehicle pilot project in the province
led by Keolis Canada ran
from Oct. 4, 2018 to
Dec. 31, 2019 in Candiac,
Quebec. The project was financially
supported by the Quebec
Ministry of Transportation,
Sustainable Mobility and
Transport Electrification, and
the Société de l’assurance
automobile du Québec (SAAQ). The
shuttle itself was fully
electric, had heating and air
conditioning systems, and could
carry 15 passengers along its
two-kilometre route at a maximum
speed of 25 km/h. Since it drove
in regular traffic, the shuttle
could adjust its speed based on
its environment, slowing down or
coming to a complete stop when
it encountered obstacles such as
nearby cyclists or pedestrians.
Like the Olympic Park shuttle,
an operator was on board in case
of emergencies and to answer
passengers’ questions. The
Candiac shuttle continued to
drive through the winter but did
not carry passengers as
it was still being tested in
snowy conditions.
The support from the government
for projects like the Transdev
and Keolis shuttles is critical.
According to Léonie Gagné, a
Montreal-based lawyer practicing
insurance as well as product,
civil, and professional
liability law, the city is
already well-positioned to have
an impact in the autonomous
vehicle industry.
“We have a big [artificial
intelligence] hub composed of
varied bodies, from the bigger
international companies, to the
smaller boutique type firms,”
Gagné wrote in an email to the
Tribune. “We have many
universities in a small radius,
the Quartier
de l’innovation and the
Centre
d’essais pour véhicules
automobiles [....] We have
the people and the knowledge to make
projects grow and be
successful.”
Montreal’s
weather also makes it an ideal city
for testing self-driving vehicles.
“Snow is actually a big
challenge for [autonomous
vehicles],” Gagné wrote. “AVs use
many tools to be able to circulate
(such as radar, lidars and sound
sensors). [Autonomous vehicles] also
use cameras, [whose] main focus is
to centre the car in lanes. So we
can imagine that with snow on the
ground, it makes it that much more
difficult for an [autonomous
vehicle] to drive.”
As
the autonomous vehicle industry
moves to tackle such weather
challenges, it seems obvious that
the future of transportation has
incredible potential. In fact, Intel
predicts that it could be worth $7
trillion
by 2050. In a survey conducted from
2005 to 2007, the US Transportation
Department found that 94
per cent of road
collisions were a result of
human error, including
distracted, reckless, and drunk
driving. Self-driving vehicles
could help minimize the number
of lives lost every year from
traffic accidents. Autonomous
cars also make transport by car
accessible to those who cannot
drive, increasing the mobility
of millions of people.
Furthermore, since cars
currently cannot operate without
a person behind the wheel, they
spend a lot of time parked;
self-driving cars would be able
to constantly serve people,
which could help reduce the
number of total cars a
population uses. Self-driving
cars can improve
fuel efficiency and improve
the reach of public
transportation systems.
“Snow is
actually a big challenge for
[autonomous vehicles],” Gagné
wrote. “AVs use many tools to be
able to circulate (such as radar,
lidars and sound sensors).
[Autonomous vehicles] also use cameras,
[whose] main focus is to centre the car
in lanes. So we can imagine that with snow
on the ground, it makes it that much more
difficult for an [autonomous vehicle] to
drive.”
“If we apply these technologies
to automating public transit, having self driving
buses for instance, a public transit system [...]
could actually be much more efficient than present
public transit systems [...] because they’d be a lot
cheaper to operate,” Holliday said.
Using self-driving vehicles would allow public
transit systems to operate smaller buses. In a city
like Montreal, which experiences harsh winters, this
change could be hugely beneficial.
“The
delays that you inevitably get due to the adverse
weather conditions here would have less of an impact
on transit riders if there were more vehicles
running more frequently,” Holliday said. “That would
make people much more willing to consider taking
transit in the winter as opposed to other means
[....] Especially because you have to stand outside
longer and if there are delays, which there almost
always are, you’re standing out even longer, [and]
you’re freezing.”
However, a
self-driving society is not without consequences.
More accessibility could mean more cars which, even
with improved fuel efficiency, could increase the
total carbon emissions from vehicles. Without the
fear of time wasted commuting, people would be more
willing to live greater distances from where they
work, further increasing fuel consumption.
“I think if we see a model whereby, as
much of it is now where people travel via personal
automobiles or ridesharing or some other form of
small cars that carry a few people, it’s going to be
pretty bad,” Holliday said. “We know from history of
studying urban development and transportation in the
20th century is that when you make it easier for
people to travel via automobile, they do it more,
and this induces congestion, actually making travel
times worse in the long run.”
Furthermore, millions of workers in the
transportation and car maintenance industries could
lose their jobs. Cars could be controlled
by malicious hackers,
and the technology’s struggles
with bias could be fatal.
“Autonomous
vehicles will invariably face
the same scrutiny that
AI-enabled systems are subject
to today,” Yuan Stevens,
B.C.L./LL.B (Juris Doctor) ‘17,
a research consultant
specializing in public interest
law, emerging technology, and
computer security, wrote in an
email to the Tribune. “There
is bias in the data collection
and analysis process, which can
result in autonomous vehicles
that don't recognize and can
harm people of colour, for
example.”
With all
of the possible negative effects
in mind, the Union of Concerned
Scientists (UCS) released a set
of seven
principles, intended to
guide companies and lawmakers
alike in their pursuit of
integrating self-driving
vehicles into society. The
principles address a range of
issues, from reducing emissions
of transportation systems to
recognizing the need for
tangible support for workers who
lose their jobs as a result of a
changed industry.
“Autonomous
vehicles will invariably face
the same scrutiny that AI-enabled
systems are subject to today,” Yuan
Stevens, B.C.L./LL.B (Juris Doctor) ‘17,
a research consultant specializing in
public interest law, emerging technology,
and computer security, wrote in an email to
the Tribune. “There is bias in the data
collection and analysis process, which can
result in autonomous vehicles that don't recognize
and can harm people of colour, for example.”
Beyond the technical challenges, there are also numerous
legal and ethical issues that regulators and companies
must address before the widespread use of self-driving
vehicles. Decisions about who to harm and who to save in
the event of a crash is one question that ethics
researchers are still
trying to understand.
Such debates also bring up
concerns over
liability.
“[I]n the
province of Quebec, the law
regulates the same for
[autonomous vehicles] and for
traditional cars,” Gagné wrote.
“[...] In short, the [SAAQ] will
indemnify the victim of a bodily
injury, and the
owner of the car at fault
will be liable for the
material injury, unless he
demonstrates the fault of the
victim, of a third party, an act of
God, or that the AV was actually
under the possession of a mechanic
or ‘a third person for storage,
repair or transportation.’”
Moving forward,
legislators may consider different
approaches. Ignacio Cofone,
assistant professor in the Faculty
of Law, argues that the law is more
concerned with the possible
consequences of the vehicles rather
than the technology itself.
“The most important
applicable legal precedent for self
driving vehicles is not about
self-driving vehicles, but rather
the rules of automobile safety
regulations and product liability,”
Cofone wrote. “Each jurisdiction
will have to decide whether it
considers that these rules are
sufficient for dealing with self
driving vehicles or other, more
tailored rules are warranted for
this specific
technology.”
Policymakers have their work cut out
for them. Even if society is years
away from having fully integrated
self-driving vehicles on the street,
the painfully slow pace of
bureaucracy may fall behind the rate
of technological change.
“The main challenge about regulating
AI is for the legislator to keep up
with the speed at which innovation
changes,” Gagné wrote. “Laws take
time to be drafted and passed, while
technology is ever
changing.”
Widespread
integration of autonomous vehicles
in our cities is coming, regardless
of whether or not lawmakers keep
pace with technological advancements
in the industry. However, Cofone
argues that the progress that
engineers are making is not the most
important evolution to
consider.
“One thing
that we should keep in mind as the
law changes (or as it stays the
same) with technological change is
that we are not regulating
technology, we are regulating how
people use technology,” Cofone
wrote. “Technological change in
itself matters less than how social
relations morph by technological
change, and how this affects
existing relationships of
power.”
The invention of the wheel, the discovery of electricity, and the development of the diesel engine all seemed proposterous once, too. However, today we can see how they irreversibly changed the world - we simply would not live the way we do without such advancements. Self-driving cars are no different: We just have to make sure that we are ready.
