skip to main content

Virginia Tech engineers demonstrate high-speed mapping for self-driving cars

November 26, 2013

image of RoboCar, self driving vehicle

Blacksburg, VA

Virginia Tech College of Engineering


The ability for self-driving cars to master precision parking, navigate urban traffic, or identify to the centimeter, is critical technology needed before self-driving automobiles and trucks can be accepted on roadways around the world.

An international team of researchers that includes 10 engineers each from Virginia Tech, the University of Technology, Sydney, Australia (UTS) and ZMP, Inc., a Japanese robotic company, is one of the first groups worldwide to solve this problem. Their technology for navigating and mapping at high speeds is being demonstrated this week at the Tokyo Motor Show. Running from November 22-December 1, 2013, the Tokyo Motor Show is one of the three largest international venues for the automotive industry.

“Most of the self-driving cars navigate with GPS, which is typically accurate to about a meter and not very accurate in bad weather,” says Tomonari Furukawa, professor of mechanical engineering, who leads the Virginia Tech team. Differential GPS is more accurate, but requires an infrastructure of ground stations, he notes.

In contrast, the Virginia Tech/UTS/ZMP technology is accurate to the centimeter. “Accuracy at high speeds is difficult,” he says. “Typically you get high speed or high accuracy, but it’s a tradeoff.”

His team uses simultaneous localization and mapping (SLAM) technology to map and navigate areas, using a laser rangefinder. Initially, Furukawa’s team developed software for indoor robotic mapping using SLAM. “We developed techniques that were extremely fast and realized this speed and accuracy could apply to autonomous cars,” he says.

He credits Xianqiao Tong (Ph.D., ’12) and Kunjin Ryu (Ph.D., ’12) for much of the initial work on indoor mapping software and Kuya Takami for his work applying the techniques to self-driving cars.

According to Furukawa, a major advantage of SLAM technology is that it functions everywhere, with no additional infrastructure. It does not need a GPS signal or any kind of pre-loaded map. "If there is no GPS available, for example in a mountain range, urban area, or underground parking garage, conventional autonomous vehicles can't do anything," Furukawa explains. "In such cases, our SLAM-based techniques can still create the map and drive vehicles."

The team first demonstrated its solution on a RoboCar developed by ZMP at October’s Intelligent Transportation Society (ITS) World Congress in Tokyo. With no information on the environment, or maps, the car successfully navigated its path and parked successfully.

“We demonstrated our technology at 20 kilometers per hour (about 13 miles per hour), but estimate that it will be effective at 65 miles per hour,” says Furukawa.

The Virginia Tech researchers are collaborating with researchers led by Gamini Dissanayake, professor of mechanical and mechatronic engineering at UTS. Dissanayake is one of the early pioneers of SLAM technology. The project is supported with funding from Microsoft and NVIDIA.

Eileen Baumann