In the fall 2004, Cris and a team of 11 undergraduate students set out to enter the DARPA Grand Challenge, a race for a completely autonomous vehicle that would run a course of 132 miles in previously unknown and rough terrain, with no one in the car and with no remote intervention nor refueling nor servicing. The route of the race was kept a secret. Only a while before the race the point of origin was announced and the course to be followed was made available by DARPA just before the race, and specified as GPS waypoints and tolerances away from the GPS waypoint coordinates. The vehicles were to rely on their own sensory information and AI to avoid obstacles while staying in the specified corridor path by adjusting speed, steering, and braking.
The original entry was made under the name Green Wave. The team had a lot of enthusiasm, a lot of talent, excellent spirit, but absolutely no resources, and was going up against big league teams that had participated in the previous Grand Challenge event and thus had advanced apparatus and lots of experience. Our team was working on designs with only a hope on a go-cart type of vehicle that was welded together by students. The hope was that this round would be some kind of replay of the previous race where no vehicle went farther than a few miles, thus we were hoping to be in the next competition and our sight was to develop the team for that next round. That was until we came in touch with the Gray Insurance company who had found in the Grand Challenge an opportunity for immense fun. Well funded and with serious hardware a couple of Gray’s IT people and a handful of the students left over from the original team started working on a Hybrid Ford Escape SUV equipped with Drive-by-wire, SIC ladars and a few computers. Working through long nights and weekends and around exams and classes, the goal was the satisfaction to make it in the 40. On May3rd 2005, the vehicle managed to perform the required maneuvers during the DARPA site visit. Now we had a pressing desire to do what best we could, we were still in the race. We thought that no matter how good we make our techniques and programs, the big dogs out there will probably do better, but, hey, we could put up a decent performance that at least was not out of reach. By the end of summer it looked like our stuff was performing really well and we were looking forward to the race. The car left New Orleans for the Mojave desert just before Katrina hit. And it had to avoid Rita as well. At the NQE pre-finals it was evident that the car, re-named Kat-5 for obvious reasons, had a real chance. In one of the NQE runs DARPA placed deliberately the center line of the waypoints right along a wall that was on a turn but the Kat-5 algorithms were intelligent enough to manage that and rise up to more such challenges.
After the NQE we were in the 23 finalists and had come farther than we thought we would a few months before and we were now making a run to actually win. It was an exhilarating experience to be in that final run and nerve wrecking the more distance Kat-5 was covering coming closer to the finish line. Kat-5 finished 4th right behind Stanford’s team and the two Carnegie Melon teams. It covered the 132 miles in 7 hours and 30 minutes, just 15 minutes behind CMU. Needless to describe scenes of hysteric celebration of course. We had started with nothing and in our first participation we were the 4th of the five teams who actually managed to make it to the finish line.
This was one of the best and most fun ideas the DARPA people have had after the development of the ARPA net. The rumors that there will be a next round have actually come true. DARPA has announced the Urban Challenge next. This is a far more demanding challenge in a structured environment with other moving obstacles (vehicles) and the need to recognize and obey traffic rules. The original team is now all employed by GrayMatter Inc, a new company to market the technology. Cris is now at Southeastern and new students from Southeastern LA University have been recruited to join the Gray Team for the Urban Challenge.
Some papers on the techniques used on Kat-5 can be found here.