DARPA Grand Challenge (2005) – Team: GCART@RIT

The DARPA Grand Challenge was a historic race where autonomous vehicles worked their way through the desert course stretching 132 miles.  The 2004 race of the same name resulted in the furthest vehicle completing 7.32 miles of the 150 mile course.

Students at the Rochester Institute of Technology formed a team of undergraduates and called their group the Grand Challenge Autonomous Race Team at RIT (GCART@RIT). They started work on their vehicle late enough where they were not able to compete in the 2004 race, but correctly guessed that no team would successfully complete the course and that another race would be held at a future date.

The team was on a very limited budget of roughly $2,000 in the early days of the project, but slowly grew to $15,000 at the peak of the effort. Other teams reported spending several million dollars developing their vehicles for the 2004 race alone. The GCART vehicle, a 1991 Geo Storm hatchback, was found abandoned on the side of the road with only 3 wheels remaining to hold the rusty frame up off the ground. After some research, the owner was found and the team bought the vehicle for $75. A small amount of equipment was donated to the team and, in the true spirit of the college experience, several parts were obtained through dumpster diving. A work area was provided by RIT, but it could only be used on weekends when classes were not using that space.

Several pictures of the vehicle and the early team members

A friend of mine was on the team and kept asking if I wanted to join the team. I turned him down several times. I was not interested in doing any low-level programming at that time – I was concentrating my studies on artificial intelligence and was heading towards a career in the gaming industry to pursue my passion for AI. I eventually, reluctantly, joined the team. Through my experiences on this project, I found a new way to apply my passion for AI. It was extremely rewarding to see something move in the real world and not just some pixels on the screen. By the time I joined the team, the car was mostly complete in terms of the electro-mechanical solution and was controlled by a video game controller. Several (5, including myself) software people were brought on to give the car its brains. The mechanical and electrical team members continued to upgrade, fix, and add components as needed during this time.

As deadlines approached, the president of RIT allowed us to utilize a barn on his property to allow for more hands on time with the car. It was winter time in upstate New York, which made for some very cold days and nights in the drafty barn. We all had our normal course work to keep up with too, and often the only time we could test was late at night in a remote parking lot on campus.

Of the 195 teams who entered into the competition, 100 were selected to receive a site visit where judges came to your site and had your vehicle perform a series of test, such as navigating a small course while avoiding stationary obstacles. If you strayed too far away from your path or if you hit an obstacle, the run was considered a failure and you moved on to your next test. Teams submitted videos to DARPA, which were evaluated and ultimately determined which teams received a site visit. Below is the video we submitted:

Our hard work paid off by securing one of these site visits, although we knew we did not have all the functionality to pass the site visit just yet. With the car’s ability to successfully navigate a series of GPS waypoints, we now needed to add the ability to avoid obstacles. We decided the fastest, easiest way to do this was through laser sensors. Our budget, however, only allowed us to purchase one of these sensors, which at the time cost close to $10,000; roughly 2/3 of our total budget. We ordered one, and due to some delays in the ordering process, we only received the sensor a few weeks before the site visit. We were not prepared when it came time for the site visit and had no ability to avoid obstacles.

Of the 100 teams who received site visits, 43 teams were invited to the semi-finals. We were never told how well we did against all the other teams, but we were proud of our achievements to make it so far with so little resources.