Home » Opinions » DARPA Urban Challenge 2007 » The Technology

The Technology

Although Intel sponsors both the Stanford University team and the Carnegie Mellon University teams, time limitations meant I only had time to talk in depth with the Stanford team but, as you'll see, although the cars themselves look quite different a lot of the fundamentals are the same.

For a start, every car used some form of what's known as a LIDAR or Light Detection and Ranging device to 'see' the world around them. These work much like RADAR by sending out a signal and detecting what is sent back but they use Lasers instead of radio waves. Because of this, they are far more accurate than RADAR and really quite small objects like lamp posts, people and fences, can be pinpointed.

As well as a number of fixed LIDAR devices, the key proponent of the visual detection arsenal of all the cars was a rotating LIDAR unit. This uses an array of LIDAR range finders that spin around at high speed, giving the car a very accurate short-range 360 degree image of the car's surroundings. In the case of the Stanford team, its detector housed 64 LASERs and spun at 10 revolutions per second. When input into some 3D visualisation software, the resultant data creates an image something like this.

Long distance RADAR range finders are also fitted to the front to enable the car to see if the road ahead is clear. This is useful for judging what speed the car can accelerate to; if the road ahead is clear the car can speed up, however if there's something ahead, the car can go more cautiously. This is essential for making sure the car moves as quickly as possible, which after all is the point of a race.

A number of other technologies like street sign and road marking recognition were also in common use and when combined with the LIDAR and RADAR devices they give each car a very complete picture of the road layout ahead and the general surroundings of the car. This was amply demonstrated by Sebastian Thrun, head of the Stanford team, in a simulation that combined all the car's data into a single 3D scene. As you can see, it gives quite a clear picture of what's going on around the car.

comments powered by Disqus