Sharing the experiences of Team Case as they compete in DARPA's 2007 Urban Challenge.
November 03, 2007
Well, today's the big day. It's a huge production! UC is webcasting the whole event. They have aerial photography and cameras placed all around the course. The cast of Mythbusters are on-hand for commentary and the traffic vehicles have already been dispatched. We'll be watching anxiously here in Victorville, and we all are pulling for a safe, successful race.
8:36AM All 11 vehicles have been launched. CMU and TerraMax already had a very near miss on a head-on collision as CMU was riding the center line in the road. The camera quickly switched to another view, but both vehicles are still running fine, so the hit was avoided.
8:38AM We are being told that Team Ben Franklin has broken down and is being moved off the road. Team Ben Franklin has not been officially retired yet.
8:45AM Ben Franklin is on the move again.
8:58AM CarOLO is off the road and stuck in the dirt in an off-road section of the course.
9:09AM CarOLO is being dug out of the pile of dirt he is stuck in.
9:11AM UCF and IVS came up to an intersection and both stopped. Cornell came up behind UCF and decided to pass UCF's chase vehicle and so it was left of center. IVS made its turn and was coming along the other lane and deftly avoided Cornell. Cornell moved completely into the left lane and is waiting for IVS to clear the intersection. Ben Franklin also made the same turn passed Cornell (who is still in the left lane) using in-lane obstacle avoidance, by moving to the left! We have a huge robot traffic jam. UCF is still stuck at the stop sign and Cornell has been paused.
9:21AM UCF has finally cleared the intersection. Cornell has recognized this and moved back into his lane behind UCF and cleared the intersection. This event was really costly for Cornell as they were waiting on an effectively disabled vehicle for about 10 minutes.
9:26AM IVS is stuck at an intersection. Little Ben, two chase vehicles, and Virginia Tech are all queued up behind IVS. It's a one-way street so the vehicles won't replan. We are in all-pause now.
9:31AM TerraMax has been passed before taking out a structural pillar in a building on the side of the road near Area "A." So much for the "safest vehicle in the competition!" We'll have to see if they let TerraMax restart.
9:35AM Oshkosh, Annie Way and IVS are being removed from the course. Team Case is all sitting here watching the action and we can't help but think that DEXTER could be competing well against these machines.
9:42AM It looks like Little Ben is trying to pass into Knight Rider. They are within about 5 inches of a collision. Maybe Little Ben tried to pass a chase vehicle. The chase cars are following far enough behind the robots that some cars might try to pass them and pass in ahead of the chase cars. We are in all pause. Oshkosh, Annie Way, and IVS are out.
10:00AM We are back under way.
10:10AM MIT is having a lot of trouble on the off-road section. They are inching forward and backward and passing phantom obstacles.
10:14AM VT is the first team to complete mission 1 of 3.
10:16AM UCF is in a house!! Looks like they are done. Stanford is finishing mission 1 right now.
10:21AM VT is back on the course for mission 2. MIT and CarOLO are stopped nose-to-nose. No contact is apparent, but both vehicles have stopped. It looks like CarOLO cut off MIT in a left turn in an intersection.
10:24AM Stanford is back under way for mission 2.
10:38AM UCF's Knight Rider is out of the race. UCF has a '96 Subaru Outback and only 10 team members, so congrats UCF on their performance to date!
10:48AM Little Ben is the 3rd machine to finish mission 1.
10:54AM CMU's Boss is the 4th to finish mission 1.
10:57AM MIT and CarOLO are getting into trouble again. They are stuck head-to-head in a traffic circle.
11:07AM CMU and Ben Franklin are on mission 2. Cornell has returned from mission 1. CarOLO is now the 5th car out of the race. MIT is the only team that's still in that hasn't finished mission 1 yet.
11:10AM MIT has completed mission 1 and Cornell has relaunched for mission 2. All 6 remaining teams have now completed mission 1.
11:49AM VT is the first team to complete mission 2. VT is looking smooth and confident.
12:00PM VT is on the move again for mission 3!
12:15PM Stanford and CMU are on mission 3. VT looks like it may be having some problems (loss of real-time, perhaps?), but after a brief pause, it's back up and running smoothly.
12:31PM MIT passed Cornell's chase car, and then went to pass Cornell's car. Cornell, who had been stopped for a while, began to move again and MIT then cut off Cornell, smashing into the left front corner of Cornell's machine. MIT and Cornell are currently paused and tangled up. Stanford has been paused because he was entering the zone with the collision. It looks like MIT will be out, and it is unclear if Cornell will be out too.
12:43PM MIT and Cornell are both back up and running. Neither team has been disqualified for the collision.
12:49PM Little Ben has finished mission 2. All but MIT have finished mission 2 now.
1:40PM It looks like the winner is going to be a toss up between Stanford and VT. Both are just about to finish. Likely it will come down to point totals!
1:43PM Stanford has finished the race. They started third overall.
1:44PM Carnegie Mellon has finished the race. They started second overall.
1:50PM VT has finished the race. They started first overall.
November 01, 2007
After Tony Tether's announcement about 11 teams being accepted to the final competition today, it was also announced that the finalists would be allowed to run their vehicles through the final test course for two hours to log data before the competition.
Up until now, the course has been guarded 24 hours a day by military personnel to prevent anyone from obtaining prior information about the track. The teams aren't allowed to log GPS, but they can log vision. Some teams rely heavily on vision, so this is a huge boost to some and erodes the benefit of having robust vision algorithms for others. I believe that this is an arbitrary last-minute decision by DARPA that is against the spirit of the competition. I realize that it is important for DARPA to show that the teams they have sponsored can perform well, but I think it's also important to keep the competition transparent and rigorous to not degrade its prestige and credibility.
1) Victor Tango
3) Ben Franklin
6) Tartan Racing
9) Annie Way
So, that's the end of the road for Team Case in 2007. It's been a great ride, though! I'll follow up with any additional details later. Thanks all for following us on the blog!
The Urban Challenge finalists will be announced at 11am PST Thursday. As far as I know, the field is already down to 18 or 19 contenders including 6 who are definitely in. I'll post more as soon as I get the word... until then, sleep well! --Farns
October 31, 2007GPS antenna cables was loose, but fortunately Amaury was quick to think of it. Once it was tightened, our IMU resumed streaming physical state information and DEXTER was off to the races! Track "C" consists of two 4-way stop intersections where DEXTER must observe intersection precedence, as well as a blocked street where DEXTER must re-plan his route and perform a U-turn to find another way to the checkpoint. DEXTER was safe, smooth and sexy, and we considered it a successful run. We also requested a re-run for all three courses this afternoon. DARPA denied our request and replied that they "already have all the information they need" to make their decision about DEXTER.
Two more teams have been scrubbed today – Axion Racing and Golem Group. Although we're sad to see the Axion Racing Twins go, DEXTER has a better shot at making the cut as other teams are scrubbed. Of note, Golem and CALTECH were both Track A teams that did not pass NQE. Track A teams received up to $1 million up front from DARPA. Team Case is a Track B team with no DARPA funding. We are one of 21 teams (including the 6 that are definitely in) vying for at most 20 spots in the final competition.
October 30, 2007
Howdy friends of DEXTER! The good news is - we're still in the running! DARPA has pared the field down to 23 teams (including Team Case). We'll will compete tomorrow for to land one of the 20 spots in the final competition.
It was touch-and-go today. We uncovered a bug that never manifested itself in simulation during our official run of track B today which cause us to graze a Jersey barrier (or "k-barrier" in California) on the way out of the launch shoot. We were on the verge of elimination after our second failed attempt at track B - Dr. Tether, the DARPA director, gave Dr. Newman "the call." However, we argued back to tell them with confidence that we understood the problem, we've fixed it, and we want to demonstrate that DEXTER is really up to the challenge. Dr. Tether and DARPA decided to give us another shot at "B."
On our second run of "B," DEXTER ran like a champ!! He unofficially had the fastest time to the parking lot at 5 minutes, 14 seconds. He had some trouble in the parking lot, so DARPA ended our run early, but overall, DEXTER ran smoothly and looked like he was operated by a seasoned driver. We have some excellent first-person (first-robot?) video of DEXTER tackling the track, which I'll try to get posted if I can.
So, it's track "C" tomorrow morning at 7:30 AM PST. Wish us luck!!
Hello, friends of DEXTER! I apologize for not updating yesterday, but I assure you our time has not been wasted! The field has narrowed by six teams as of this writing, but DEXTER and Team Case are still in the race!
On Monday, we replaced DEXTER's batteries. This has greatly improved DEXTER's idling and electrical performance. We also re-ran course A on Monday. We had a problem where the left LIDAR (laser scanner) was tilted too far down and was seeing the road to the front and left of the vehicle as an obstacle. This caused DEXTER to err on the side of caution and not take his turn at the intersections at track A. We've since fixed this issue, and hopefully we'll get another shot at "A."
Another issue that had been giving us trouble is that our vision computer was running on the same 300W 24V DC/DC converter as the LIDAR units, so that power bus was right on the edge of its budget. This situation was exacerbated by our old batteries. So, we decided to mount an additional Vicor 500W 24V DC/DC converter to the vehicle's 48V bus. Unfortunately, during the upgrade, the 24V line came into contact with hot 48V and we lost three LIDAR units - the side two and the rear one. I learned of this around 1AM, so I woke back up and got to work on repairing the system through the night. This involved re-wiring the power connector blocks (which were charred), moving some of our existing and spare LIDARs (the front-left and front-vertical units were moved to left and right, and a spare was mounted at the rear), and updating our sensor placement configuration files to take all this into account.
This morning, the Georgia Tech Sting Racing Team, who has already been eliminated from the competition, graciously lent us two of their LIDAR units. Using these, we were able to re-populate all of our LIDAR mounting locations. CALTECH also let us borrow one of their spare LIDARs. We've ordered new units, but they won't be here until after the NQE is over. So, we are extremely grateful to GT and CAL, who are a class act for helping us out!
As I'm typing, DEXTER is getting ready for his second run of Course B, which is the most similar track to the final competition. I'll post an update later today with DEXTER's results! If you haven't seen it yet, you can also watch DEXTER's first track A run online. Go Team Case!