A "Stanley" for the road: DARPA Grand Challenge hints to the future of driving

Westlake village (CA) - Completely autonomous vehicles may be decades away, but Sven Strohband, Volkwagen's Electronics Research Lab lead engineer and project manager for Stanford's Grand Challenge project, says computerized driver-assistance is just around the corner. Still basking in Stanford's win of the Grand Challenge race, Strohband sees computers assisting drivers rather than replacing them in the near future. Some of these "driver-assistance" features are already available in higher-end vehicles such as Lexus, Mercedes-Benz and Volkswagen.

Volkswagen gave Team Stanford three brand new Touareg R5 SUVs for the Grand Challenge. Two vehicles were for testing and the actual race, while the third car was used as a "chase" vehicle. The VW Touareg R5 is already mostly drive-by-wire and the company helped the team with integrating some of the engine's computer data with the vehicle's navigation and position estimator software.

Strohband believes autonomous vehicles have a much better sense of their surrounding environment than humans. The vehicle knows exactly how long it is, how much distance is needed to stop and how much space it needs to turn. "For Stanley, accurate measurements of the distance to obstacles is no problem, actually he gets a map within centimeter accuracy 10 times a second," the engineer explained.

Another advantage of an autonomous vehicle is that it does not get tired. "It does not matter if Stanley has driven for 10 hours straight, it will still perform the same in the next hour. This hardly can be said for a human driver," said Strohband. During the Grand Challenge race, TerraMax, a competing vehicle, was paused overnight in the desert. At 6:40 AM the next morning, it was "unpaused" and finished the course.

Swipe to scroll horizontally
Row 0 - Cell 0
Sven Strohband

Driver assistance features are not just for safety, but can be used for comfort. An automatic or "super" cruise control system is currently in development that will adjust speed to stay a safe distance from the vehicle in front of you. Strohband told us this could reduce driver fatigue in stop and go situations and drivers wouldn't have to engage in repetitive braking and acceleration.

"Convoy driving", a line of cars following each other half a car-length apart, is an obvious extension of super cruise control. Strohnband says this is an easy problem to solve and could be one of the first steps towards a road of fully autonomous driving. The front driver manually drives the car, while the following vehicles turn on autonomous following. Since the cars are so close, fuel economy could be increased by 20%. In addition, more cars are squeezed onto the freeway.

There are some things that autonomous vehicles cannot do well and Strohnband admits that there is room for much improvement, "Although Stanley is quite an astonishing vehicle, humans are still by far the better drivers," says Strohband. The biggest weakness at this point is that computer systems cannot distinguish the severity of an obstacle. At the Grand Challenge race, a bird flew in the path of Stanley's laser sensors and the vehicle swerved to avoid the "obstacle". A human would probably have kept driving, risking a collision and avoiding the chance of a rollover. Strohnband says that the vehicle did exactly what it was programmed to do, adding, "A human can classify obstacles much better and for example distinguish between a tumbleweed, which is safe to drive over, and a rock which is not. This classification based on experience is one of the biggest advantages humans have over autonomous vehicles."

Computer assistance systems are already available in higher-end cars. Lexus offers their Pre-Collision System which pre-tensions the seatbelt and helps with braking, when a collision is imminent. Electronic Parking Assistance, where ultrasonic sensors emit audible and visual signals when you get a few feet of another object, is becoming more and more common in middle-class and higher-end vehicles. Currently, Mercedes-Benz appears to have the most intelligent system with their Distronic Adaptive Cruise Control. Drivers select a safe distance from the front vehicle and a radar unit located in the front grill will help the computer automatically adjust the cruising speed.

Strohband told us that there are many other interesting features coming in the near future. A lane departure warning system will use cameras to sense if the vehicle is straying away from a lane. In early versions, an audible tone will be emitted, but eventually the car might be able to autonomously steer back on course. Another feature in development is a blind-spot eliminator that will emit a tone when the driver is about to turn into an occupied lane.

Related Articles

DARPA Grand Challenge 2005: The road to victory