LAS VEGAS — General Motors Corp. this week provided attendees at the 2008 Consumer Electronics Show (CES) a glimpse of the future, giving them a chance to ride in the robotic driverless vehicle that won the Defense Advanced Research Project Agency’s (DARPA) 60-mile Urban Challenge race in November.
Known as “Boss,” the high-tech Chevrolet Tahoe employs LIDAR, radar, GPS and a computing platform that’s essentially a “supercomputer on wheels” to autonomously brake, steer, accelerate and “see” obstacles in the road. CES attendees were allowed to ride in the passenger seat of the vehicle as it navigated a designated path in a parking lot across the street from the show’s main hall.
GM and a team of engineers from Carnegie Mellon University, Continental Automotive Systems, Caterpillar, Inc. and a host of other companies designed the vehicle and dubbed it “Boss,” in honor of GM’s research and development founder, Charles F. Kettering.
GM engineers at the show said that the vehicle is more than a glitzy technical curiosity. “It’s aligned with our vision,” noted Bakhtiar Litkouhi, manager of vehicle control systems for GM’s Electrical and Controls Integration Laboratory. “It’s very important from a customer and safety point of view.”
Litkouhi added 75% of today’s accidents are caused by driver behavior. Meanwhile, roads are getting more crowded and today’s total of 850 million vehicles is expected to grow to more than a billion as China and India begin to buy more cars.
Boss could help solve the problem by taking driver behavior out of the picture. The prototype vehicle incorporates technology from a wide variety of sensor manufacturers and computer companies. A rotating LIDAR system from Velodyne is mounted atop it. Boss also employs Continental radar systems and planar lasers from Sick Optic, along with three GPS antennae on the roof. Inside, it uses a CompactPCI chassis packed with 10 computing blades, each containing an Intel Core 2 Duo processor and 10 GBytes of memory.
“You have the equivalent of about 20 laptop processors working in there,” said Bob Bittner, Tartan Racing test lead for Carnegie Mellon.
The vehicle uses the huge computing platform to run “sensor fusion” software to enable the vehicle to understand what it is “seeing.”
“You’ve got a lot of different sensors seeing the same thing, and they all see the same thing differently,” Bittner said. “So you have to use the sensor fusion to enable the vehicle to make sense of it all.”
In the CES parking lot, the vehicle used a GPS-based map of the course, then did the accelerating, braking and steering by itself. GM engineers placed garbage cans in the roadway and drove other vehicles across its path at stop signs to show that Boss was capable of reacting to the unexpected.
Engineers at the show said that the technology will begin as an adjunct to the driver, but will eventually become completely autonomous. Litkouhi estimated that complete autonomy could arrive as soon as 2020.
“There’s a lot of computing power on board, but from a production (integration) standpoint, we can eliminate a lot of the computing systems and sensors,” Litkouhi said. “We can clearly see an evolutionary path to full autonomous vehicles.”