Google surprised the automotive world with the recent announcement that the company has produced its own
fleet of driverless Toyota Priuses. The company says that its cars have
driven themselves for 140,000 miles—with a driver in the seat, just in
case—making Google's fleet the most thoroughly tested autonomous cars
to date. Considering that the first
DARPA competition in 2004 ended without any of the competitors finishing the 150-mile course, the technology is moving fast.
Partially autonomous cars are closer than you think; in fact, they're already here. The
2011 Infiniti M56
uses forward-looking radar to track the car ahead and cameras to detect
lane boundaries. It automatically maintains the gap to the car in
front—with the brakes and throttle—and can also use a single wheel brake
to nudge a drifting car back into the lane's center.
"We are building the blocks for autonomous vehicles," Nady Boules, the
director of GM's Electrical and Control Integration lab and a member of
the team that won the 2007 DARPA competition with a driverless Chevy
Tahoe, recently told PM, "In ten years we'll have the technology for
autonomous vehicles well in hand."
Digital maps and onboard computing power are already up to the task, but
the accuracy and robustness of the sensors needed to determine the
car's position and the obstacles around it require further development.
Current commercial-grade GPS is only accurate to a couple of yards and
it needs to be down to the inch level. The laser and radar sensors and
vision systems require higher resolution and must get smaller in size
before they're suitable for production cars.
Until those inevitable improvements arrive, our cars will gradually take control. Automated steering is on the horizon—
electrically assisted steering
already compensates for road crown and crosswinds—which will enable
computer-controlled highway lane changes and accident avoidance. The
challenge for the automakers that offer these systems will be keeping
the driver engaged. "Until the car is fully autonomous, the driver is
ultimately responsible," Boules says, "Part of the research is figuring
out how to keep the driver from taking a nap."
Vehicular autopilot will be useful even without
vehicle-to-vehicle technology,
which broadcasts a car's speed, position and braking to increase
safety. V2V is only useful if every car on the road has it, and that
will simply take too long. Once V2V is ubiquitous, however, autonomous
cars will be able to shed some of the sensors, reducing the cost.
The safety benefits of the automotive autopilot are obvious, but the autonomous car
can also increase transportation efficiency and reduce congestion. And maybe one day we'll be surfing the Web as our car takes us to work.
0 comments:
Post a Comment