Failure begets more interest
This year's Grand Challenge race
may have 200 entries
By Allison Bruce, abruce@VenturaCountyStar.com
February 24, 2005
The failure of any vehicle to
finish last year's Grand Challenge, a race across the
Mojave Desert for unmanned vehicles, did not damper
enthusiasm for this year's race.
The competition has grown from more than 100
applicants last year to just shy of 200 for the 2005
Grand Challenge. The prize has doubled this year to $2
million.
Two area
teams that were finalists last year are signed up for
another try.
The SciAutonics/Auburn Engineering team's RASCAL
vehicle stopped just short of a mile from the starting
line, mostly from too much jostling of the hard drive on
the rough course. Many of the team members are based in
Thousand Oaks.
Axion Racing, based out of Westlake Village, also is
returning for another chance.
The challenge was introduced in 2004 as a way to
advance the technology of autonomous vehicles, which
could lead to development of unmanned vehicles for
military use. The Grand Challenge is run by the U.S.
Department of Defense's Defense Advanced Research
Project Agency (DARPA).
Congress has demanded that one-third of the Army's
ground vehicles be autonomous by 2015, said Tom Goodwin
of the DARPA communications staff.
Last year's race was supposed to stretch from Barstow
to Primm, Nev. Nobody made it that far, leaving the $1
million prize unclaimed.
When asked why the SciAutonics/Auburn Engineering
team is taking another stab at the competition,
SciAutonics President Reinhold Behringer said, "We are
asking that ourselves sometimes."
But the team learned something the first time, has
the basic technologies in place, knows the time and
planning that will go into the vehicle and is ready to
try again, he said.
Teams not involved last year have little chance of
winning this time, said Bill Kehaly, team leader of
Axion Racing. The Axion Racing team entered "Spirit of
Kosrae" in last year's race.
"Autonomous vehicle racing is not for the weak of
heart," Kehaly said. "It's brutal just physically having
to go out to Barstow and drive around for weeks. ... I
know everybody from Barstow to Primm."
There are things teams learned last year that could
really pay off for this year's race.
For instance, the California Speedway in Fontana --
where the qualifying trials are held -- is built on the
site of an old steel plant. "If you have a
dead-reckoning system based on a magnetic compass, you
are in a world of hurt," Kehaly said.
His team is working on getting its vehicle to do the
course without using GPS, even though there are four GPS
units on the car. Kehaly predicts that the car that
makes the quickest steering decisions based on its
sensor data will be the winner.
Seizing a challenge
A majority of the applicants are newcomers.
That includes the DTI Rebel, a modified Jeep from the
team led by Ventura resident Ralph Benham, an
independent consulting engineer, and John Ward, owner of
Diamond Tech in Rocklin, Calif. Benham said he was
attracted more by the challenge than the $2 million
prize.
The duo didn't even pursue sponsors.
"I would sure like to find an angel to finance this
... thing," Benham said. "But, on the other hand, we did
not seriously look for anyone to help us financially.
Our bad guess is each of us will have about $20K in
this."
Unlike the teams that competed last year, Benham's
team isn't tweaking an existing vehicle but assembling
one from scratch. It all has to be done in time to send
an application video in to DARPA by March 11.
Teams will provide DARPA with more detailed
information about their entries, including a vehicle
specification sheet and video demonstration of the
vehicle. Autonomous systems means there are no remote
controls.
Kehaly is not above a little marketing to get his
team to stand out from the competition. His selection
video includes the "Axion Racing Twins," attractive
sisters who got the team mentioned in Playboy in
Germany. They also are featured prominently in shots of
the vehicle and team.
Selected teams will have on-site visits in May. This
year, 40 vehicles will be invited to qualify at the
California Speedway in Fontana between the end of
September and beginning of October, and 20 will
participate in the race on Oct. 8.
DARPA has authority to run Grand Challenge-type
events through 2007, but whether there will be another
challenge after this one might depend on what comes out
of this year's competition, Goodwin said.
"If there were to be a winner on Oct. 8, DARPA may
very well decide to have another challenge in another
area of technology, or it might decide to take some of
the technologies gained running the challenge for two
years and apply them to DARPA advanced research
projects," he said.
For the teams, building their autonomous vehicles
often means putting in hours of their own time and money
to make it work. Benham doesn't expect to have much free
time from now to March 11 as he and his team work on
building the DTI Rebel.
As general manager of Axion LLC, Kehaly said he is
often searching for the lowest cost on sensors,
including searches for the right price on eBay.
SciAutonics LLC has no income, so its members work on a
volunteer basis to compete.
Learning curve
Last year's SciAutonics team consisted mostly of
employees from Rockwell Scientific in Thousand Oaks, as
well as some other engineers and scientists. This year
adds a contingent of engineers from nearby Amgen Inc.
"Although that's not their primary product interest
there, it's seen as a team activity," said Reinhold
Behringer, president of SciAutonics and a Rockwell
Scientific employee.
The team includes a partnership with Auburn
University, which has the vehicle and is working on its
steering, while those in Thousand Oaks write software
for the vehicle.
The team plans to replace the hard drive with a
solid-state hard drive this year and make sure the
computers are mounted where they won't deal with the
shock of too much vibration.
"The hardware is something that needs to be robust
before someone can even think of software that is
intelligent," Behringer said.
The next thing the team is working on is the
vehicle's obstacle recognition -- figuring out there's
something in its path and getting around it. Better
software should improve that as well, he said.
Sharing ideas
After last year's competition, there have been some
discussions among teams about what worked and what
didn't.
The competition is fierce between teams, but the
Grand Challenge also tried to encourage innovation by
making the technical papers for the finalist teams
available on its Web site. That means that any
competitor could visit the site and see what went into
the other vehicles.
Benham admitted he's looked at the specifications for
several of last year's finalists.
"There's no way in any product you're going to
develop you don't look at the competition," he said.
DARPA has created a forum for its Grand Challenge
participants to talk with one another online. So far,
the discussion forum, which started getting posts not
long after the 2004 competition, has more than 3,100
posts from close to 860 users on topics such as obstacle
detection, navigation and rules.
In the interest of science, there is discussion,
Behringer said. But, of course, each team is trying to
keep its competitive advantage, he said. "The openness
stops at a point."
Last year's challenge stirred several promising
ideas, Goodwin added.
Those included a shock isolation system for
autonomous sensors to protect the sensors from the
bouncing and jostling of the rough terrain; an algorithm
to estimate the terrain ahead, analyze the route and
quickly plan a path in real time; and a trailing arm
suspension for absorbing the impact of running over
large obstacles and changing the center of gravity of
the vehicle.
Even as they plan for the race, competitors are
thinking about commercial applications.
Kehaly said he has several possibilities in the works
with different organizations, though he couldn't discuss
them yet. He said some applications included assisting
the disabled.
Behringer envisions systems that help drivers avoid
collisions by decelerating when the car senses an
obstacle, such as a pedestrian or another car.
|
| 
Home
)
)
.gif){kind=spacer}
