The race is on.
Or soon will be. Next month, a group of tricked-out, high-tech cars
will compete for a $2 million prize in a contest sponsored by the Defense
Advanced Research Project Agency (DARPA).
There will be no controversy over female drivers in this race because these cars have no drivers. The main goal is to use high tech-assisted guidance to complete the course.
And that’ll be easier said than done for the 43 contestants. In last year’s race, Carnegie Mellon University’s modified Humvee went the farthest (almost 8 miles) before getting stuck close to a cliff’s edge. To win this year’s prize, entrants need to complete a difficult 175-mile course through the Mojave desert in less than 10 hours using no human intervention.
A first time entry team from Stanford University has packed a Volkswagen
Touareg R5, affectionately named Stanley, with enough high tech components
to make the Batmobile envious. The car includes a turbocharged powerplant
and chassis, a cluster of six onboard Intel Pentium M computers that process
information for five laser range finders, a radar system, monocular vision,
GPS and an inertial measurement system.
To avoid the kind of software defects that have thrown previous entrants
off course the Stanford team enlisted a lot of high tech support including
Coverity, a San Francisco-based maker of software quality and test
solutions. Founded by Stanford University scientists in 2002 Coverity works
with such well known companies as Juniper Networks, Veritas, McAfee,
Synopsys, Sun Microsystems, Wind River as well as NASA, on software
quality.
Stanley in motion. Source: Stanfordracing.org |
“A project like this isn’t that different than what we normally do, we
look for the same kind of defects in the software, bugs are bugs,” Coverity
CEO Seth Hallem told internetnews.com. “But there is a different
complexity here with all the onboard sensors and computers.
Coverity is already working with telematics companies on car navigation
systems and Hallem expects to be doing more of that work in the future. “The
onboard computer industry an important industry when you look at what’s
being added in the way of communications and entertainment. The CEO of Ford
said he expects there to be 100 million lines of code in the average
consumer car by 2010.”
Coverity’s Prevent and Extend source code analysis products are being
used to find and eliminate software defects in six custom software modules
designed by the Stanford team: Computer vision; Inertial Navigation;
LIDAR — Light Detection And Ranger (similar to radar but with lasers);
Planning and optimization; Control; and Reliability.
Engineers at the Stanford Artificial Intelligence Lab (SAIL) tested and
analyzed Stanley’s software which might someday lead to driver assistance
systems designed to enhance automotive safety. SAIL found 68 defects split
about evenly between custom software it had developed and software libraries
that other engineers created.
“We were able to identify and subsequently
address software issues using Coverity’s technology that we were unable to
find through manual testing,” Dr. Alex Aiken of the Stanford Computer
Science Research Group, said in a statement.
Another sponsor TYZX, based in Menlo Park, CA, is one of several
companies helping with Stanley’s guidance system. TYZX uses stereo vision,
which helps a computer see like humans do through passive light, as opposed
to sending out a signal, such as radar, and waiting for a return. The TYZX
stereo vision also helps a computer accurately determine depth – much like a
humans.
The company’s name is a play on its approach. In computing a digital
image, T represents time, X and Y represent height and width, while Z
represents depth. In the Tyzx 3DAWARE system, depth perception is measured
by comparing two images and calculating the precise shift in a particular
pixel – or picture element-in each image.
More than 30 Stanford engineering faculty and students from computer
science, mechanical engineering, aeronautics and astronautics, and
management science and engineering departments are involved in getting
Stanley prepped to compete as well as engineers from the Volkswagen
Electronics Research Lab. Other supporters include staff from early stage
venture capital firm MDV-Mohr, Davidow Ventures, and sponsors Honeywell,
Intel, Red Bull and Android.