November 3, 2007 (2330hrs), VICTORVILLE, CA. — The Urban Challenge Final was very anticlimactic. After the intense semifinal evaluation, we were expecting a real intense final. But it turned out, the semifinal was the real grill. The final was mellow, probably designed to prevent a repeat of 2004 GC.The controlled traffic that the finalists encountered was pretty limited. Unlike the NQE test in in Area A, there were no back to back cars at any intersection. Once one or two cars had passed there was ample time for a bot to make a turn. In any case, the traffic appeared to be limited to the loop immediately after the start. There did not appear to be any additional human traffic in the neighborhood.
The gauntlet or the neighborhood scenario, with cars parked on both side of the road, was no where to be seen. There was only one instance where one robot had to travel in a region where cars were stalled/stopped (not parked) on both sides. The tests in Area B were an overkill, with 10-15 cars on both side of the road. A simulated neighborhood with bots traveling from both sides would have raised the bar, and made it a real challenge. In the absence of such a scenario, it sure feels we wasted a lot of time developing a capability that was irrelevant for the final.
There was only one scenario that CajunBot could possibly not handle. It was the drive through off-road terrain with very sparse waypoints. But the capability we developed for Area B took us half-way in that direction, and we had two more days to complete the work. So we would surely have had the capability in place.
The bots did not have to replan their path, a major part of the test in Area C. All they needed to do was follow precedence at intersection.
All in all, the perception of our team and others is the final event was a big let down.
– Arun Lakhotia