News

On Friday, 35 teams of robotics researchers will field their entries in the semi-finals of a major U.S. contest for autonomous vehicles. The cars will have to navigate through a complex, real-world setting - dodging obstacles and following traffic rules - using neither human drivers nor remote controls.

And no less than 10 of the contenders will be relying on state-of-the-art navigational technology from a Canadian company to do the trick.

The technology, developed by Applanix Corp., of Richmond Hill, Ont., not only guides the vehicles physically but also helps them perceive their environment and respond to it.

"We consider Applanix to be the market leader in this field," says Sebastian Thrun, a professor of computer science at Stanford University in Palo Alta., Calif., and head of the school's entry team. "[Its technology] is just very, very precise. It gives us a fantastic estimation of where our vehicle is."

U.S. military contest

Sponsored by the U.S. Defense Advanced Research Projects Agency, the contest is known as the Grand Challenge and flows from the Pentagon mandate that one-third of the military's ground combat vehicles be unmanned by 2015. It's the third of a series of DARPA contests aimed at accelerating research and development and getting the potentially life-saving vehicles into use even sooner.

The challenge this time is to build an autonomous vehicle that can not only move through a mock urban environment - negotiating busy intersections and merging into traffic - but also interpret and obey traffic signals and avoid moving obstacles.

That's where Applanix's technology comes in, and the fact that so many teams are using it is a feather in the company's cap.

"It's very advantageous for us to be accepted by this community because they're incredibly demanding," says Louis Nastro, director of land products for Applanix.

Civilian spinoffs

As with most military research, there are sure to be civilian spinoffs (think nuclear power, computers, lasers, integrated circuits, freeze-dried foods). So how soon can we see self-driving, self-navigating, self-regulating cars on our roads?

It could be as early as 2025, Mr. Nastro says. And it could be even sooner for backup safety systems that could take over if the driver were to make an error, such as trying to take a curve too fast.

"The technology already exists," he says. "The problem is we need to make it cheaper to meet the business needs of the auto industry."

Applanix, launched in 1991 by three Canadians and acquired four years ago by California-based Trimble Navigation Ltd., makes systems that provide highly accurate positioning of moving vehicles - even when the vehicles can't receive signals from global positioning system (GPS) satellites.

Applanix systems use the same basic technology that powers GPS navigation units for cars (but with much more finely tuned antennas) along with "inertial" technology that precisely measures a vehicle's speed, acceleration and orientation, allowing it to accurately position itself in all dimensions.

Applanix's technology, currently used around the world for mapping and surveying needs, was originally developed to get around a problem with GPS: When a receiver doesn't have clear line of sight to receive the satellite-generated radio signals that allow it to calculate its position, accuracy drops off rapidly.

Applanix's inertial technology solves the problem, using gyroscopes and very precise speedometers and odometers to accurately measure the distance and direction the vehicle travels during GPS outages. When the system starts receiving GPS signals again, it can recalculate its position to within centimetres.

Highway and roads departments now use Applanix systems mounted on trucks with video cameras for automated mapping of assets such as signs, stop lights, fire hydrants and so on.

Highway departments also use Applanix technology in conjunction with LIDAR (laser imaging detection and ranging) systems to assess the condition of road surfaces. Laser beams detect and map every pothole and crack, and the system produces detailed 3D-maps of roadways that allow engineers to more easily assess and prioritize resurfacing needs.

The move to automation

In the past, these sorts of jobs had to be done manually. Automating them wasn't possible without Applanix's inertial technology, which measures the exact orientation of the vehicle when video images or LIDAR readings are made - taking into account curves and changes in elevation in the road and the movement of the vehicle's suspension.

Similar Applanix technology is used by vessels doing hydrographic mapping to chart the ocean floor, and in aerial mapping using photography.

Applanix's aerial technology has also been used to map disaster sites, such as the Hurricane Katrina-ravaged Mississippi delta. Not only can officials see what's on the ground in disaster areas, but the inertial technology lets them generate elevation maps that can provide vital information for rescue workers.

In future, Mr. Nastro says, this kind of surveying and mapping could be done by unmanned aircraft - much like the unmanned vehicles being tested in the DARPA Grand Challenge, which carries a top prize of $2-million (U.S.).

Mock urban area

This year, the six-hour contest requires the driverless vehicles to simulate military supply missions in a mock urban area on an abandoned air force base in Victorville, Calif.

In previous contests, vehicles only had to avoid static obstacles and navigate across flat desert where GPS signals are uninterrupted. This time, they have to contend with dynamic obstacles - such as other robots moving around the test site - as well as frequent gaps in GPS coverage.

The inertial data about a car's position, measured along every axis, will be fed 200 times a second to the control system, and LIDAR data about conditions ahead will be vital for the vehicles to navigate the 100-kilometre course safely and figure out in a split second how to avoid obstacles.

In the most recent contest, the team from Carnegie-Mellon University in Pittsburgh used the Applanix technology in its two vehicles, placing second and third; last year's winning team, from Stanford, is using the Applanix system this year.

Costs for systems of the kind Applanix sells, when deployed in commercial applications, range from $40,000 to $300,000. But before we see self-driving civilian cars on the road, that will have to come down to between $300 to $500, Mr. Nastro reckons.