64 SERVO 12.2015
mounted Sick AG LIDAR units on a
rack that fed data to the car’s
computer running their Mapper
program. The computer interpreted
the 3D LIDAR map and compressed it
into a 2D version. Though astonishing
many people — as Stanford did not
compete in the previous competition
— their mastery of AI and
autonomous vehicle control is really
no surprise. MIT and Carnegie Mellon
may take top honors in robotics, but
Stanford leads everybody in AI with
their world-renown SAIL lab. It is easy
to see why two Stanford PhD students
have made such a success with their
company.
Quanergy is another new startup
in the Silicon Valley area that is on
track to produce the LIDAR unit
shown in Figure 15 at about $250.
They have collaborated with Mercedes
(as seen in Figure 16) to develop an
inexpensive sensor strictly for the self-driving automobile market. DARPA has
also developed a unique LIDAR for
autonomous mobile vehicles that does
not utilize a mechanical rotating turret
to scan the laser beams. Their interest
is certainly steered towards
military applications, but
the consumer applications
abound.
Unlike the less complex
SRS RoboMagellan robots
that need to traverse an
open area of ground to
locate visible waypoints and
a final target area,
autonomous self-driving
cars carry human beings in
an unknown environment
over many miles of roads. We all
realize that the RoboMagellan robots
utilize GPS to determine their position
in a contest area, and use odometry to
augment the GPS data and cameras to
locate waypoints and targets. If this
type of robot gets stuck in a bush or
against a curb, the builder can take his
creation back, change a few lines of
code, and give the machine a second
try. If the autonomous car steers itself
across several lanes of traffic into
oncoming vehicles, well, no amount of
new code can correct the tragedy that
may occur.
One thing all manufacturers hate
to think about is product liability. It
has been said that one of the greatest
costs of a ladder is for the probable
liability lawsuits that come from folks
who don’t take the proper safety
precautions (like the guy who places
his aluminum ladder against a power
line and his angry wife sues the
company because he was
electrocuted). It’s bad enough when a
drunk driver kills someone on the
highway, but who is responsible when
a self-driving car drives into another
vehicle in a scenario like the one I
mentioned? An autonomous car must
be absolutely ‘fool proof,’ or as close
as possible because the driver is no
longer a human, but a product of a
company. As of this time, no Google
car has caused an accident, though
one was struck from behind.
The Google Self-Driving Car
Technology
Whereas NASA would require
only one of the Space Station mobile
remote manipulator system robots at
a cost of almost a billion dollars (back
then), Google had to look at a
consumer product that would sell in
the hundreds of thousands or even
millions, so costs were critical. The
The computer can take
GPS information to
Figure 15. Quanergy LIDAR.
Figure 16. Dual Quanergy LIDAR sensors mounted on a Mercedes.
Figure 14. Stanford's 2005 DARPA Grand Challenge winner on a VW Touareg chassis.
