Austin American-Statesman | By Cristina Peña
Having already designed an SUV that drives itself, a project group at the University of Texas is now working on the technological next step: an autonomous intersection that lets driverless vehicles navigate without stoplights or stop signs.
The group — a partnership between UT and Austin Robot Technology, a team of volunteer engineers dedicated to building the autonomous SUV — might be the only one in the nation researching the possibilities for autonomous intersections, said Peter Stone, the project leader and a UT associate professor of computer science.
The SUV is formatted with a software program, which is layered to have three main functions. The pilot function controls the brake, throttle and steering, while the navigator function chooses paths on the road. Using the SUV's GPS, the navigator targets speed limits to adjust the throttle. These two functions then communicate with a revolving sensor on the top of the SUV that detects the road and objects around the vehicle. The sensor has a high density light radar with 64 lasers arranged to scan the area.
Autonomous vehicles have been a reality for several years, but the research by Stone's team aims to create a system that would interface with those vehicles, coordinating the direction and flow of traffic through an intersection. Then, in theory, multiple driverless vehicles could navigate the streets without crashes.
The research — which is ongoing at UT's main campus and at the J.J. Pickle Research Campus — is using an autonomous SUV built by Austin Robot Technology.
To make the autonomous intersection work, a software program will be installed in the intersection to communicate with the vehicle's software. The two systems use dedicated short-range communication to communicate wirelessly 300 to 400 meters away, which would let the driverless vehicles know when the intersection is clear to pass through.
Stone said the technology could conceivably hit the streets in the next few decades.
"The autonomous intersection, technologically, I believe is already feasible to be widespread," said Stone, who is also director of the Learning Agents Research Group, which operates with the Artificial Intelligence Laboratory in UT's Department of Computer Science. "I wouldn't be against it happening in the next 20 to 25 years."
Because of the expense involved — the sensor array for the car costs about $75,000 — the technology's first practical applications might be for the U.S. military, said Arturo Martin-de-Nicholas, president of Austin Robot Technology.
"The technology is very expensive, so you're not going to be seeing this in passenger vehicles yet," Martin-de-Nicholas said. "The military does have the money, and eventually that technology will trickle down to the consumer."
The military is aiming to have a significant portion of military vehicles be autonomous by 2015, Martin-de-Nicholas said. He said aerial drones are leading the trend, but ground navigation will be more difficult.
"You're starting to see it already with drones in Afghanistan to track things on the ground," Martin-de-Nicholas said. "It's actually harder on the ground because you have people walking and obstacles of varying terrain."
In fact, it was a competition for the U.S. military that launched Martin-de-Nicholas and Austin Robot Technology into researching autonomous vehicles. The federal Defense Advanced Research Projects Agency began hosting the DARPA Grand Challenge in 2004 as a engineering competition for autonomous vehicles. Austin Robot Technology created an autonomous SUV to compete in the challenge in 2005, but it did not make it into the finals.
Austin Robot Technology reached out to UT because of challenges with writing software, Martin-de-Nicholas said.
When Austin Robot Technology decided to take part in the DARPA Urban Challenge, a new competition to find autonomous vehicles that could drive in a city setting without collisions, Stone joined the team and helped prepare the vehicle. The team made it to the semifinals.
The team sought out Stone because it didn't feel they had the robotics expertise to compete in the urban challenge, said Jack O'Quin, a former IBM developer who work on software for the project and with the undergraduate research team at UT.
"We provided the platform, computers and engineering details to actually make the car," O'Quin said. The University of Texas "provided the artificial intelligence, which is really the hard part of the project."
The project — which typically involves about six staffers — is being funded through a Federal Highway Administration grant and by UT's Freshman Research Initiative, which is a program aimed at getting freshmen involved in research programs early in their college career.
The autonomous car and intersection project is an attractive research option for many at UT.
Tsz-Chiu Au, a postdoctoral fellow in the department of computer science at UT, said he hopes the technology will be widely available at some point.
"I don't expect it will happen very soon but I will keep doing research and making it better and better," Au said. "The most motivation for me is to do something that will have some impact on society, even if the impact won't be immediate."
Contact Cristina Peña at 445-3797