Haute Wheels: GrayMatter, Inc. and the AVS
by Stephen Faure
It’s not so much the wheels this time; it’s the driver of the car. Or, more precisely, the driverless-ness of the car—it is driven by robotics. The GrayMatter AVS (autonomous vehicle system) was developed by an unlikely firm, The Gray Insurance Co. They’ve started a separate company, GrayMatter, Inc., to develop the system further and make it commercially viable.
In 2004, Eric Gray, the president of GrayMatter, Inc., who, along with brothers Mike and Walter, is a co-owner of Gray Insurance Co., had read a story in Popular Science magazine about the first attempt at an autonomous vehicle challenge sponsored by the Defense Advanced Research Projects Agency. That event turned out to be a total failure, but Eric could see potential in the concept. “We saw the probability of a business proposition, with the military wanting to have one-third of its vehicles unmanned by 2015,” Eric says. “We asked ourselves, ‘Are we crazy enough to give this a shot?’ And the general consensus was, ‘Yes!’”
“It started as a team-building exercise,” Eric says. “Gray Insurance’s computer programmers finished a project ahead of schedule and under budget.” Having a group of talented programmers with too much time on their hands was what Eric needed to get started. Paul Trepagnier, Gray’s lead programmer, ran with the project. Trepagnier, like the Grays, though, had no experience with robotics or artificial intelligence.
An autonomous vehicle works by first being given a route to follow. The same type of GPS that guides many human drivers tells the vehicle where it is on the route. Sensors mounted on the vehicle act as the eyes, spotting potential obstacles around the vehicle. What’s needed is software to follow the route, interpret information from the sensors and adjust the vehicle’s speed and steering accordingly.
Just as GPS has become readily available for consumer use, the other systems needed for a car to operate autonomously are around—but they can’t be bought at Wal-Mart. Companies such as Electronic Mobility Controls in Baton Rouge manufacture and install “drive-by-wire” systems in cars for physically disabled drivers.
The team-building exercise for the programmers—the Grays also hired students from Tulane’s computer science department—grew into a quest to enter DARPA’s next event, the Grand Challenge. The AVS was their achievement. It connects with the GPS and sensors for navigation, contains the software, processors and memory needed to determine the route and avoid obstacles and connects with the drive-by-wire system for steering, accelerating and braking.
Going up against some of the biggest names in the field of robotics, such as Carnegie Mellon and Stanford universities, Team Gray spent under six months on the project and had a relatively small budget. Carnegie Mellon, on the other hand, had spent three years designing and programming autonomous vehicles.
The 2004 DARPA challenge, which inspired Eric Gray to form Team Gray and GrayMatter, Inc., was an exercise in mediocrity for the agency and participants. A 142-mile course was laid out. Fifteen participants entered, ready to roll. But there was no winner. Carnegie Mellon’s car traveled farthest before getting stuck, its tires spinning so furiously they caught fire, after only 7.3 miles.
Team Gray had higher expectations in entering the 2005 challenge. “We would have been happy to have gone 7.4 miles and then home,” Eric says. Instead, theirs was one of four vehicles of the 23 that started the race to make history. “Up until that day, no autonomous vehicle had gone that distance [132 miles] at that speed without human intervention,” Trepagnier says. “It was a whole new day for robotics.”
It didn’t go off without a hitch, though. But for a bug in their code, Kat-5 (formerly GrayBot; Katrina hit during preparation for the challenge, which inspired a defiant renaming) may have finished first. A wide-open, obstacle-free part of the course should have been the fastest portion of the race. But, “[Kat-5] tried to scan too wide an area,” Trepagnier explains. Safety mechanisms in the software kicked in, slowing the vehicle down so the scanning could catch up—it didn’t know it was trying to scan empty space. “It took about two minutes to fix it after the race.” Kat-5 finished fourth, coming in 36 minutes behind the winner.
GrayMatter’s next test was the 2007 DARPA Urban Challenge, the event featured on the Science Channel’s program, “Robocars.” The car (there were actually two of them, both called “Plan B”) was also a hybrid SUV Ford Escape. There were some big differences between Plan B and Kat-5. The electronics gear needed to run Kat-5 occupied the entire interior of the car—driver’s, passenger’s and both rear seats and the cargo area. Plan B’s AVS box is about the size of a briefcase.
Instead of being in a desert environment, the vehicles had to adhere to the California Driver Handbook. Teams in the Urban Challenge were given a street map, with the speed limits of each street. The car not only had to calculate a route to the destination, but also spot obstacles placed in the way while avoiding other vehicles, both of the autonomous and human-driven varieties. It also had to follow traffic signs and park itself in a designated spot.
The team stretched its design muscles by putting together two different vehicles for Plan B, each having the AVS tapped into an EMC drive-by-wire system.
Unfortunately, Team Gray was not among the eleven teams making it to the finals of the Urban Challenge, but it hasn’t slowed down. The members are still refining the AVS, further reducing the size of the hardware and optimizing the software. Soon, GrayMatter will have a production model that will be one-quarter the size of the one that ran Plan B.
General Motors CEO Rick Wagoner has predicted driverless vehicles will be part of GM’s future. Citing GM’s involvement with Carnegie Mellon in the DARPA challenge, Wagoner addressed the issue in his keynote speech at this year’s Consumer Electronics Show. “[A]utonomous driving means that, some day, you could do your e-mail, eat breakfast, apply your make-up, read the newspaper, watch a video…all while commuting to work. Or,” as he joked, “you could do all the things you do right now while commuting to work…except you could do it safely!”
Other than the military, there are at least two other markets the Grays hope to enter. The automotive industry eventually may have AVS as an option for consumers, and can use it now for running cars on test tracks. Also, “Two mining companies approached us at the Urban Challenge,” Eric says. Both underground and open-pit mines can use the technology, which doesn’t need light to see and makes it unnecessary to put drivers in dangerous environments.
After all, Eric says, “We’re an insurance business, so we know safety.”
