Smart Industry 1/2018

Smart Business Title Story: Self-driving cars
The company has invested heavily in
research involving machine learning,
which Huang says is the “bottom-up
approach to artificial intelligence” –
and probably the most promising
technology today. Machine learning
requires the processing of huge
amounts of data, and as it turns out,
the company’s computative graphics
processing units (GPUs) can do the
job both faster and using less energy
than the traditional central processing
units (CPUs) found at the heart of
most mainframe, desktop, and laptop
computers today.
The computative power of GPUs has
increased as computer images have
become more complex and, in 2007,
Nvidia pioneered a new generation
of GPU/CPU chips that now power
many energy-efficient data centers in
government laboratories, universities,
and enterprises.
It was almost by accident that the
company became a big player in the
nascent autonomous car sector, but it
now plans to release the Nvidia Drive
PX2 platform next year, describing it
as the first "AI brain" capable of full
autonomy.
Nvidia’s approach is so revolutionary
that other chipmakers are scrambling
to catch up. Intel and AMD, two of the
largest manufacturers of computer
chips, have teamed up to pool their
resources in order to head off Nvidia
by developing a GPU/CPU combo of
their own. In addition, Intel made the
How a self-driving car works
Signals from GPS (global
positioning system) satellites
are combined with readings
from tachometers, altimeters
and gyroscopes to provide more
accurate positioning than is
possible with GPS alone.
Radar
sensors
Ultrasonic sensors may be used to
measure the position of objects very
close to the vehicle, such as curbs
and other vehicles when parking.
The information from all of the
sensors is analysed by a central
computer that manipulates
the steering, accelerator and
brakes. Its software must understand
the rules of the road,
both formal and informal.
biggest acquisition of its lifetime
early in 2017 when it paid $15.3bn for
Mobileye, an Israeli-based specialist
in sensors, mapping technology, and
camera-based devices for advanced
driver-assistance systems (ADAS).
This is just one example of many that
hint at the impending disruption
autonomous vehicles will bring to the
automotive and related industries.
As for when self-driving cars will hit
the mainstream, opinions still vary.
But Tesla, Ford, Audi, General Motors,
and Nissan are among those that
believe cars operating without humans
will be on the road within the next
five years.
This visionary or perhaps doom-laden
scenario, depending on your view,
is driving the whole supply chain to
frantically figure out where to position
themselves. Delphi, a UK supplier of
proprietary automobile components
and integrated systems and modules,
acquired NuTonomy, a developer of
autonomous driving (AD) software
solutions for $450m in November
2017. Delphi, which will soon change
its name to Aptiv, has announced it
intends to build a self-driving system
it can sell to all the big auto makers.
ADAS requires immense computing
resources to provide higher levels of
predictability and autonomy.
Typical components are: radar, lidar
(similar to radar but using laser
light), camera, ultrasonic, vehicle-toeverything
(V2X) wireless sensors
Lidar (light dtecting and ranging) sensors
bounce pulses of light off the surroundings.
These are analysed to identify lane markings
and the edges of roads.
Video cameras detect traffic lights, read
road signs, keep track of the position of other
vehicles and look out for pedestrians and
obstacles on the road.
Radar sensors monitor the position of other
vehicles neraby. Such sensors are already
used in adaptive cruise-control systems.
Ranking Autonomy Levels
HUMAN DRIVER
MONITORS DRIVING ENVIRONMENT
0 1 2
No Automation Driver
Assistance
3
Partial Conditional
Automation Automation
The Society of Automotive Engineers
(SAE) definitions of vehicle automation
AUTOMATED DRIVING SYSTEM
MONITORS DRIVING ENVIRONMENT
4High
Automation
5
Full
Automation
■ How autonomous is autonomous?
The Society of Automotive Engineers (SAE) International
Standard J3016 offers some guidance on this important
question with a six-level schema to describe the range of
scenarios that are possible between traditional, unaided,
and fully automated driving.
Level 0 No Automation. The full-time performance of all
aspects of the dynamic driving task, even when enhanced by
warning or intervention systems, is left to the human driver.
Level 1 Drive Assistance. The driving mode-specific
execution of either steering or acceleration/deceleration is
conducted by a driver assistance system using information
about the driving environment, with the expectation that
a human driver will perform all remaining aspects of the
dynamic driving task.
Level 2 Partial Automation. The driving mode-specific
execution by one or more driver assistance systems of
both steering and acceleration/deceleration is conducted
using information about the driving environment, with the
expectation that a human driver performs all remaining
aspects of the dynamic driving task.
Level 3 Conditional Automation. The driving mode-specific
performance by an automated driving system of all aspects
of the dynamic driving task is conducted with the expectation
that a human driver will respond appropriately to a request
to intervene.
Level 4 High Automation. The driving mode-specific
performance by an automated driving system of all aspects
of the dynamic driving task continues even if a human driver
does not respond appropriately to a request to intervene.
Level 5 Full Automation. The full-time performance by an
automated driving system of all aspects of the dynamic driving
task occurs under all roadway and environmental conditions
that could have been managed by a human driver.
photo © SAE International and J3016
14