Telematics and the autonomous car

Autonomous cars—vehicles that park themselves, stay on the road, find preprogrammed destinations, and wake you when you start to fall asleep at the wheel—have made a splash in the last few months.
In October, a driverless orange Piaggio van pulled up at the 2010 World Expo in Shanghai, four months and more than 9,320 miles after departing Parma in June.
The van, part of the VisLab Intercontinental Autonomous Challenge, was equipped with GPS, of course, and a variety of sensing devices.
Three laser scanners mounted on the front bumper detected pedestrians, obstacles, and other vehicles.
A panoramic vision system stitched together input from three synchronized cameras to detect and track the leader car, a semi-autonomous vehicle in which a human driver could make the split-second decisions required in city traffic.
The autonomous vehicle followed the first car using either visual cues or GPS information broadcast by the lead car.
Three color cameras mounted on the rooftop solar panel array provided a frontal stereo sensing system for locating obstacles, determining the terrain slope, and finding lane markings.
Two additional color cameras in the back watched for obstacles in the rear.
Cars similarly equipped might run on freeways, as well as downtown streets, according to Alberto Broggi, professor of computer engineering at the University of Parma and director of the VisLab-Artificial Vision and Intelligent Systems Lab.
The software, developed by VisLab over the last 15 years, is quite complex.
“We’re now working to prepare it on open standards, so that we will be able to share it with different players in the vehicular field,” Broggi says.
“Agricultural, mining, road works, and the military are the first fields that will benefit from this technology.”
Look Ma, no driver!
VisLab claims its intercontinental trek is the longest autonomous drive to date, but Google is hot on its tail.
The search advertising company unveiled its autonomous car program in October, too.
The company refused an interview request, but in a blog post lead engineer Sebastian Thrun, who is also head of Stanford University’s Artificial Intelligence Laboratory, said automated cars (with trained drivers in them to take over, if necessary) have driven down San Francisco’s crookedest street, navigated the winding Pacific Coast Highway, and crossed the Golden Gate Bridge, logging more than 140,000 miles.
In early September, an Audi TTS—sans driver—was tested on the part-gravel road to the summit of Pikes Peak in Colorado, a grueling and twisting route taken by numerous racing rallies.
The Autonomous Audi TTS Pikes Peak was developed with the Stanford University Dynamic Design Lab and the Audi Electronic Research Laboratory (ERL).
One feature is a differential GPS system that can keep it within a couple centimeters of the road’s center line by correcting for minor errors inherent in satellite transmissions.
The lane-keeping system was developed with the aim of bringing it to everyday driving, although it won’t appear in consumer cars for many generations, according to Burkhard Huhnke, executive director of Audi’s ERL.
Improving safety
One promise of autonomous driving is reducing accidents.
The World Health Organization estimates that there are more than 1.2 million traffic fatalities worldwide each year.
However, robotic drivers are still a lot worse than human ones, according to Jeffrey Rupp and Anthony King, Ford Motor engineers who received an award at the SAE Convergence 2010 conference for their paper “Autonomous Driving—A Practical Roadmap.”
They point out that in the most recent DARPA Grand Challenge, the mean mileage between errors for autonomous cars was a mere 100 miles; human motorists drive an average of 50,000 miles between crashes.
But technology outperforms people in other driving tasks.
A recent study by Ford and the Massachusetts Institute of Technology New England University Transportation Center found that Ford’s Cross-Traffic Alert technology allowed all drivers in a test to stop and yield to oncoming traffic, while only 71 percent of drivers without the alert did so.
The researchers also found that Ford’s Active Park Assist technology, which aids in parallel parking, significantly lowered drivers’ heart rates.
An elevated heart rate is a sign of stress, and chronic stress can lead to heart disease, diabetes, and other ailments.
In the US, parallel parking is one of the more difficult and potentially stressful driving situations; 31 percent of US drivers actively avoiding parallel parking whenever possible.
The drivers using Active Park Assist also exhibited less stress when they were preparing to park, compared to those having to maneuver on their own.
So, autonomous cars could even keep us healthier.
The telematics opportunity
Government mandates and insurance company incentives have already made driver-assist technologies, such as electronic stability control and warnings for lane departure and collision, attractive.
On the road to the autonomous vehicle, there will be opportunities for developing new kinds of hardware and software systems to do more advanced calculations using input from multiple sources.
At the same time, makers of sensors and cameras may see sales quadruple or more as cars begin to ship tricked out like VisLab’s Piaggio.
Vehicle-to-vehicle and vehicle-to-cloud communications may require a new Internet-like system that could lead to a kind of dot-car boom.
Jeffrey Rupp, a safety manager for Ford, says that OEMs will need to broaden their partnerships to include a variety of players outside the traditional automotive space, including academia, skunk works, startups, and even hobbyists.
At Ford, he says, “We look for other players out there in this or other industries who are starting to innovate, form some cooperative relationships, and see where it goes.”
Susan Kuchinskas is a regular contributor to TU.