10.05.2005

Last gasp

After a valiant effort the Indy Robot Racing Team was unsuccessful in reaching the finals of the DARPA Grand Challenge autonomous vehicle event. Late this afternoon, the IRV made what turned out to be its final qualification attempt. And, as so often happens at the Indianapolis Motor Speedway, this 11th-hour attempt to make the field came up short.

After a strong start in today's qualification attempt, the IRV encountered problems with its Global Position Satellite system and stopped for an extended period of time on the course. We were later to learn that the vehicle "thought" that it was a mile away from its actual location, causing a "this does not compute" situation. The irony here is that this latest problem was relatively minor, yet turned out to be the show-stopper unlike far more sophistocated visioning and guidance challenges mastered by the team.

Earlier in the day the team made several practice runs, and undertook minor adjustments to the vehicle. With all the marbles riding on this afternoon's qualifying run, the team was under serious pressure to figure out why the IRV's guidance system was consistently pulling to the left. Said another way, the vehicle was performing like your personal car does when its tires are out of alignment.

Why was the IRV pulling to the left? After much analysis it was determined that a crash on the team's first day of qualifications caused an avalanche of subsequent problems.

During a routine and mandatory DARPA safety inspection, the IRV unexpectedly lurched forward, smashing into a nearby concrete barrier. The front undercarriage and bumper were bent slightly causing a misalignment of the vehicle's various guidance systems. But, the misalignment was just enough to cause the steering problems. Furthermore, as the IRV lurched, DARPA officials failed to push the emergency "kill" button. Had they done so, this might have saved the vehicle from its rude encounter with the concrete.

But why did this happen? Ultimately it was determined that the navigational computer file given to the team by DARPA for its safety inspection contained a fatal flaw. The file name included spaces, not uncommon in the Windows environment. However, a few of the IRV's onboard computers use the Linux operation system which only reads file names WITHOUT spaces. Once the program was loaded, the vehicle's Linux computers reverted back to the last properly formatted code in their memory - in this case, a run command - that contained no spaces in the file name, from a test conducted in the desert several days earlier.

The IRV did what the computers told it to do: accelerate forward rapidly. Only this time the vehicle was just a few feet from a concrete wall, and not in the desert. Crunch! DARPA has since changed its procedures, and won't issue any programs with spaces in file names. But the damage was done to the IRV, and things went downhill rapidly after that.

At tonight's team dinner, Jones took a philosophical approach, saying that while it would have been nice to have made the finals - and nicer still to have won the event - the real accomplishments were the team's tremendous technical advances in recent weeks in preparation for the DARPA Challenge. Jones is convinced that technologies developed for the IRV project will transfer to an emerging robotics industry destined to play a huge role in the creation of tomorrow's automobiles. And, he says Indiana is well-positioned to take advantage of this technology.

So now what? Jones and his team are planning to attend the finals in Primm, Nevada this Saturday. Our plan is to be there too. Stay tuned :-)

[photo courtesy: Steven Wallace/Indiana University]

0 Comments:

Post a Comment

<< Home