Research scientist Dean Pomerleau and robotics doctoral student Todd Jochem won't touch the steering wheel but they will operate the throttle and brakes of a 1990 Pontiac Trans Sport, known as Navlab 5. During their trip, which they've dubbed "No Hands Across America," they will prove the roadworthiness of on-road autonomous lane keeping and lateral roadway departure warning and support systems they and their colleagues at the Robotics Institute have been developing for the past 10 years. Pomerleau says they expect to be over 98 percent autonomous on this trip.
The brains of Navlab 5 is a new, vision-based software system that was developed by a Carnegie Mellon spinoff company. As the vehicle moves along, a video camera mounted just below the rearview mirror reads the roadway, imaging information including lane markings, oil spots, curbs and even ruts made in snow by car wheels. The camera sends the image to a portable computer between the car's front seats that processes the data and instructs an electric motor on the steering wheel to turn right or left.
The driving system runs on the PANS (Portable Advanced Navigation Support) hardware platform. The platform provides a computing base and input/output functions for the system, as well as position estimation, steering wheel control and safety monitoring. It's powered from the vehicle's cigarette lighter and is completely portable.
Over the past six months the PANS platform has supported over 3,000 miles of autonomous lane keeping, including 30 miles on a closed test track on which Navlab 5 reached a top speed of 90 miles per hour.
Pomerleau believes there could be spin offs from this technology in terms of warning systems, where turning in the proper direction would be made easier than turning in the wrong direction. He predicts these kinds of technologies should appear in consumer vehicles "long before we see a vehicle driving fully automatically."
Navlab 5 will be crossing the country on major interstate highways, first to Indianapolis, St. Louis and Kansas City, then to Denver and Las Vegas and finally to Los Angeles and San Diego.
In May, Pomerleau and Jochem did a preview of the California journey with a 305-mile trip from Pittsburgh to Washington, D.C. On that trip, the driving system autonomously steered the Navlab 96 percent of the way at an average speed of 57 miles per hour. Faster and cheaper computers have enabled researchers to make tremendous advances in autonomous vehicle technology. In 1986, the first Navlab, a blue Chevrolet panel van, contained five racks of computing equipment, including a supercomputer. Its top speed was two miles per hour. In 1990, a camouflaged Army ambulance known as Navlab II contained three high-end computers plus two others for low-level control jobs. It could move over rough terrain at six miles per hour and reached on-road speeds of up to 70 miles per hour.
"This trip will validate our software in the real world," says Jochem. "It will also be the longest, continuous, autonomous interaction for a robot in a real world environment."
At Carnegie Mellon, research into autonomous vehicles began in 1985 and has received continuous support from the Advanced Research Projects Agency (ARPA). More recently, additional funding has come from the U.S. Department of Transportation which hopes to use smart car technologies to make American highways safer and more productive for the nation's drivers.
The primary sponsors of the Navlab 5 journey across America include Delco Electronics, AssistWare Technology and Carnegie Mellon's Robotics Institute. Andrew Corp. and RDI Computer Corp. are providing additional support.