autonomous cars - recent development
DESCRIPTION
1) Definition and Classification2) Potential + Advantages of Autonomous Cars3) Challenges and Barriers to introducing Autonomous Cars4) Previous visions regarding Autonomous Cars5) Current developments: Technological status quo6) Future prognosisTRANSCRIPT
ŠKODA AUTO UNIVERSITY
Seminar Work
Topic: Autonomous cars – recent
development and the future
New Trends in Automotive Industry
January 2015
Lucie Tajovská
1
1 Definition and Classification
An autonomous car, also called a driverless car, a robot car or self-driving car is in
general a „computer controllled car that drives itself“ [1]. Acording to one of the
definitions, the term autonomous is described as „having the power for self-
government“ [2]. In the context of autonomous cars it implies that the vehicle can
complete a transport operation without any interference from humans, except
initiating the operation. There are several levels of the vehicle’s autonomity. An
NTHTSA, the National Highway Traffic Safety Adminitraton, has created an official
classification system, scoring vehicles on thier automation level [3]:
No-Automation (Level 0): The driver completely controls the vehicle.
Function-specific Automation (Level 1): certain vehicle functions are
automated, such as system of electronic stability control or automatic
braking.
Combined function Automation (Level 2): At least two controls can be
automated and work together in key situations
Limited Self-Driving Automation (Level 3): An example of this category is
the well-known Google car. The car is capable of performing of all driving
functions, assuming the driver is available for occasional control.
Full Self-Driving Automation (Level 4): The vehicle is fully automated and
performs all safety-critical functions, while the driver, or rather the
passenger, is not expected to control the vehicle. As this vehicle would
control all functions from start to stop, including all parking functions, it
could include unoccupied cars [4].
2 Potential + Advantages of Autonomous Cars
Reducing accident rate is the main advantage of autonomus cars, as more than
90% of all accidents are caused by the operators' (drivers) incapacitation,
distraction or misjudgment [5]. Replacing the human operator with an automated
system not prone to the same deficiencies can therefore potentially reduce the
number of road accidents to almost none; in turn saving society large resources
currently absorbed by health care, insurance and premature deaths.
2
Allowing groups of people unable to operate a car the same mobility as drivers.
These groups could be very old people, children/youngsters, disabled and people
who are on medication preventing them from driving.
There is also a potential for a self-parking car, as parallel parking represents a
major undertaking for many people. As of January 2015, BMW is about to present
a car, that can drop-off passengers and then park on its own [6]. Mercedes in its
Luxury in Motion Concept has taken it even further – the car can actually find
parking on its own and return when summoned. Pedestrians are of course taken
into consideration [7]. The major challenge lies of course in cars equiped with
intelligence adequate to deal with all the complexity of real-life environment and
unusual situations.
Driverless vehicles transporting cargo between manufacturers and from
manufacturers to retailers or direct delivery to consumers. This has the potential of
significantly reducing the cost of cargo transport as a human operator is not
needed for most of the operation; only loading and offloading of the vehicle would
require human intervention - and even these operations could prospectively be
automated.
3 Challenges and Barriers to introducing Autonomous Cars
The main challenge to operating autonomous cars and cargo vehicles is the
tremendous complexity of the environment wherein the vehicles will operate. The
environment include other vehicles of course, bikers, pedestrians, animals and
virtually any imaginable object blocking the transport corridor. And all of these can
potentially move in any direction at any speed. Add to this that weather conditions
can restrict visibility and change the way the vehicle reacts to physics... e.g. how
quickly a vehicle can change direction on a wet surface while still being under full
control of the operator.
All these complexities place huge demands on the flexibility of sensors and the
processing capability of the intelligence ensuring the safe operation of the
vehicles. The human eye and brain in combination have far more processing
capacity (in most circumstances) than any automation system based on artificial
intelligence developed to date. But unfortunately the human processing of data
3
doesn't always lead to the same result; which result in the before mentioned driver
attributed accidents.
Artificial intelligence offers more consistent and uniform data processing, but
processing capacity is lower and slower; just as the capability of sensors like
cameras, radar, laser-rangers and similar can distort data processing if they are
not in perfect working order or conditions reduce their performance.
There's also a legal challenge to introduce autonomous vehicles; namely the
question of responsibility for accidents, damages or even death of humans.... all of
which relate to the issue of insurance. With a human operator the responsibility
question is easy to answer.... the current human operator has the responsibility.
When a computer or artificial intelligence is the operator then it becomes close to
impossible to identify who are responsible. Is it the system designer, the
manufacturer, the retailer, the system maintainer, the user or the owner? The
insurance companies and car manufacturing industries in a dialogue with
governments are still trying to find a solution to this problem...
4 Previous visions regarding Autonomous Cars
Previous decades ideas generally concentrated on reducing the complexity of the
environment wherein the autonomous vehicles would operate. This strategy was
caused by the available technology only a few decades ago, which did not allow
for the complex and fast processing of data that most people take for granted
today.
Dedicated lanes or entirely segregated roads - often called guideways - would
allow the safe operation of autonomous cars with relatively simple sensors and
modest processing capacity. These ideas very often included supply of energy
(electricity) to the autonomous vehicles via a power rail along the before
mentioned guideways. One sub-set of these ideas are dual-mode vehicles;
meaning vehicles under autonomous control when driving on segregated
guideways while human operated when driving on conventional roads.All these
systems generally only exists as concepts; in a few cases with basic prototypes
built. They have names like RUF, BiModal Glideway and HiLoMag [8].
4
5 Current developments: Technological status quo
The present state of research and development is somewhere between Level 2
and Level 3 on the classification system described earlier. Many production cars
are already equipped with 2 or more automation system easing the driving task for
the operator, e.g. automatic lane-keeping and automatic braking if an object (other
car) is detected within a safety zone in front of the car. Level 2 automation is
rapidly approaching product maturity as more and more series-produced cars are
equipped with these systems, which also benefits from the practical experience
gained from having many systems in operation.
Level 3 automation, i.e. cars that are fully automated under certain situations,
exists in advanced prototypes and is in increasing numbers being demonstrated to
journalists and other members of the public. These systems cannot cope with the
full complexity of the environment of for example an inner city, but are intended to
be used primarily in environments where the complexity is significantly reduced;
the most obvious possibility being normal highways where the only type of object
expected are other moving cars and vehicles, all moving in the same direction and
at roughly the same speed. Level 3 automation relies on vehicle mounted sensors
working in conjunction with onboard computers and electronic maps of high
precision. All these things are almost ready for production, but one element of the
system is not completely ready and will take some time to field. This element is
vehicle-to-vehicle communication (V2V), which is necessary to warn vehicles - and
their drivers - about conditions that would require the driver to switch off the
automation and revert to manual control.
Level 3 automation does not have legal implications as the driver is responsible for
deciding when and where to drive in automatic mode automation and therefore it is
still the driver who are fully responsible if the use of automation causes an
accident. Since there is no legal barriers and the technology is at an advanced
prototype stage we can expect to see the gradual introduction of Level 3
automation within a few years...
5
6 Future prognosis
The final stage of automation, Level 4 in the above mentioned classification
system, and what most people would probably consider a truly self-driving car is
still not ready to be introduced to the public. The main problem is still the
complexity of the driving environment and the way the artificial intelligence
interprets the input from sensors. It is still some way from being able to interpret
very varied and unpredictable inputs in the same way as humans do. Humans also
take social context into consideration, for example a uniformed person moving his
or her arms in a certain way can be a signal to the driver to stop the vehicle.
Artificial intelligence would not understand such a context and there is significant
disagreement when - if at all - it will be capable of such advanced data processing.
When 500 experts in automation technology were surveyed about the prospects of
completely autonomous cars more than half said they would be a reality in 2030 at
the very earliest; 20% expected it to be 2040 while a minority of 10% did not
expect it to be possible ever... [9]
6
Resources
[1] http://www.pcmag.com/encyclopedia/term/57132/autonomous-vehicle
[2]
https://books.google.hu/books?id=59z6AwAAQBAJ&printsec=frontcover&dq=inaut
hor:%22Stetz,+Thomas%22&hl=sk&sa=X&ei=nHy6VPaiHYXNygOy2YL4CA&ved
=0CCAQ6AEwAA#v=onepage&q&f=false
[3] http://www.thecarconnection.com/news/1084651_nhtsa-lays-out-groundrules-
for-autonomous-vehicles
[4] http://en.wikipedia.org/wiki/Autonomous_car
[5] http://www.alertdriving.com/home/fleet-alert-magazine/international/human-
error-accounts-90-road-accidents
[6] http://www.techtimes.com/articles/22848/20141224/bmw-previews-wearable-
tech-self-parking-valet-mashup.htm
[7] http://www.autotrader.com/research/article/car-news/233807/mercedes-benz-f-
015-luxury-in-motion-concept-detroit-auto-show.jsp
[8] http://faculty.washington.edu/jbs/itrans/dualmode.htm
[9] http://www.technologyreview.com/news/529466/urban-jungle-a-tough-
challenge-for-googles-autonomous-cars/