The first off the mark, 'Little Ben' moved out nicely, turned smoothly onto the test traffic circle, did most of its lap with a chase car tagging along, but then stopped abruptly and needed a human to bring it back.

But the next two bots to launch moved just a few feet and 'paused', as they say here - meaning they stopped themselves, or were stopped dead by a human overseer. Then the pride of Silicon Valley, Stanford's 'Junior' moved out, made a quick left 120 degree turn, and nearly impacted the bot three stalls over. Oops!

Given that all of these systems had already made it through this maneuver over the last few days, having three in a row go Tango Uniform was more than a bit suspicious, and the DARPA coordinators halted the test. I haven't heard any official word on what was going on, but the range of speculation was interesting in itself. The theories I heard in the bleachers and pits included:

• Sitting on the line for so long without launching might cause any system with a memory leak to fail.
• Purely by coincidence, the vendor of the differential GPS systems used on many of the vehicles had staged a new firmware download within the last 24 hours. That shouldn't make a difference, but....
• The rotating laser sensors used on many of the vehicles were interfering with each other.
• Some of the data files passed out to the teams might be corrupted.

Anyone who's ever worked on a distributed, realtime system will be cringing about now. As one team member put it, "There's thousands of ways for this to screw up, and only one way it works right."

DARPA had everyone reload their route files, and then restarted the test with no more three bots on the line at a time, and everything worked fine. Each of the vehicles rolled out, did its appointed lap and recovered at the appointed spot.

Watching from the bleachers, even such a small run shows the range of personalities of the bots. Just as a human driven vehicle has its own sort of 'body language', so each of the Urban Challenge bots gives an impression. Team Annieway's VW feels like a student driver: over-steering at slow speeds, moving tentatively onto and around the circle.

A CMU Team Tartan member commented that their 'Boss' has attitude, and indeed it moved out through the start area at a good clip, and was one of the fastest bots around the circle.

The 25,000 pound Terramax from Oshkosh Trucks has a presence due to its sheer size, starting out with a hiss of air brakes and motoring confidently along.

This qualitative feel is at least loosely related to the systems' characteristics. Are the various physical control systems well synchronized? Can the planner keep up with actual motion? Is the team confident enough to let its bot travel at a good clip? It will be interesting to see if my notes on 'body language' correlate to success or failure in tomorrow's big event.

(Update: Here's another view of the morning's events from a member of Team Cornell. I'm one of the guys who asked to meet him become of 'this little blog segment' of his. He's done a great job of capturing the atmospherics of the place. You've got to see the 'evil army of synchronized Ford Tauruses' in action to understand.)