DARPA - 2007 Urban Challenge

Introduction: 

The Defense Advanced Research Projects Agency Grand Challenge is a competition that was started in 2004 in order to help push the research of autonomous vehicles. These challenges Tasked participants with designing and creating an autonomous vehicle to specific standards and requirements and have them compete in a certain challenge depending on the year. Specifically the one I will be talking about is the Urban Challenge in 2007. This challenge tasked teams with creating an autonomous everyday car that would obey all traffic regulations and negotiate with other traffic such as cars or obstacles.

Method/Discussion: 

(Figure.1 Map of the loops and Intersections used by Team Jefferson)

Many teams which consisted of college students from their respective schools created vehicles that would be able to drive by themselves through an urban area course while still obeying traffic laws. Since self-driving cars has been a buzzing topic ever since autonomy was first thought of, this challenge was created in order to push that specific field and gain information. Previously however, most research on self-driving cars has been conducted on a highway with little to not interaction with other vehicles. DARPA wanted to expand on this field of study and include other cars as well as every day obstacles that we would face if we had a self-driving car. The teams were given maps of different way points throughout the course so they could program the autonomy to drive to specific locations by itself. Team Jefferson wrote a paper identifying all the components and challenges they went through during the development process. Figure 1 illustrates one of the specific way points which was two loops with two intersections in order to test stopping at intersections, precedence at intersections, as well as blocking of path/rerouting. With these specific routes created, the autonomous vehicles uses lidar, gps, and a camera (Figure.2) in order to recognize its' surroundings. The vehicles were to be programmed to understand stop signs and stop accordingly, recognize obstacles in the way such as debris, as well as other competitors. 

(Figure.2 Diagram of one of the vehicles used and its components for autonomy)

Results:

Of the 11 vehicles entered into the competition, a little over half managed to actually complete the race. Many of the cars hit obstacles or crashed and the autonomous program was not able to recover or understand what to do in that specific situation. Tartan Racing, the team from Carnegie Mellon University finished first in the race and only averaged 14 mph throughout the race.Other vehicles had a similar speed throughout the race as well but no car drove faster. These results might seem frightening at first but when you look at the implications, 6 out of 11 cars were able to complete the course. While they may not have done it at record breaking speeds or as elegantly as we imagine self-driving cars to be, it does show that autonomy is possible. Technology is always advancing and makes way for not innovations. This challenge was conducted in 2007 when autonomy was in its early stages so imagine redoing this challenge today. The results would be much different. Overall, this challenge was not meant to show a high speed race but to push the industry of autonomy to the next level and make way for new innovations