URBAN Challenge -- Oct-Nov 2007
Jim's notes on the big race.   I have been anticipating for this week for almost two years.

    The top three finishers are clear. Stanford passed VaTech, but CMU started 9th and finished third, advancing 6 places making the best time. However there will be bonus and penalty minutes to consider for violations of traffic laws, etc. that will change the final ranking.
THE VEHICLES , in elapsed time sequence.
11 . AnneWay (Germany)-- stalled and was eliminated for excess time.
10 . IntellVehSys -- the pickup truck stopped at an interesection and would not move.
9 . Oshkosh "TerraMax" -- the big truck ran into a post.
8 . U.Central Fla (low budget) -- ran into a house
7 . CarOLO (Germany) - touched MIT, damaged a sensor.
6 . MIT - very hesitant, stopped frequently for ?? shadows, dust, to process data of 40 PCs?
5 . Cornell -- simply drove slowly.
4 . Franklin (Penn/Lehigh) -- drove slowly, but surely
3 . VaTech -- drove competently throughout.   #3
2 . Stanford -- confident with good speed.    #2
1 . Tartan (CMU) -- perfection.   - Official #1

QUALIFIERS : Finish 1 . Tartan, CMU, 'Boss" ( 2)-17   2 . Cornell ( 6)+95 3 . Victor Tango, VaTech "Odin" ( 3) 0 4 . Ben Franklin, "L'tl Ben" ( 4)+60 5 . Stanford, "Junior" ( 1)-10 6 . CarOLO "Caroline" ( 7) x 7 . Intelli.Vehicle Sol. (11) x 8 . MIT ( 5)+100   9 . AnnieWay (10) x 10. Oshkosh Truck ( 9) x 11. U.Central Fla ( 8) x

- - - DID NOT QUALIFY 12. LUX 13. Urbanator 14. Gray 15. Cybernet 16. Case 17. CajunBot 18. OSU-ACT 19. Insight 20. AvantGuardium 21. Austin Robot 22. Golem 23. Axion

- - - 24. Mojovation 25. Jefferson 26. Juggernaut 27. Gator Nation 28. Urbanator 29. CalTech 30. Ody-Era 31. SciAuto/Auburn 32 . Princeton 33 . U. Utah 34 . Berlin 35 . Georgia Tech 36 . UU

LINKS
• DARPA Urban Challenge
• Grand Challenge -- webcast
• Wired News
• TG Daily
  PROMPT BLOGS
• CarOLO [Ger]-- "Caroline", Volkswagen Passat. *6*
• Team Lux [Hamberg, Germany] VW Passat, only three Ibeo laser sensors. -N-
• Tartan (2,3) - Carnegie-Mellon, Red Whittaker, Tahoe, "Boss". *1*
• Gator Nation (18) "Urban Navigator", Toyota Highlander Hybrid . -27-
• Insight (NC)(12) Lotus, "Lone Wolf", smallest entry. -N-
• Sting GaTech, Porsches Cayenne, "Sting 1". -35-
• CalTech (21)- "Alice" , Ford van . -29-
• CajunBot (15) U.Lafayette, LA. Jeep, maps. -N-
• Princeton (19) PAVE, GMC Canyon pickup truck named "Prospect Eleven". -32-
• Jefferson U.Vir, (23) Scion "Tommy Junior" -- good pics of course. -25-
• Cornell U. (20) Chevy Tahoe. *2*
• Oshkosh (5) Big Marine Truck, "TerraMax". *Q*
• Victor Tango (Virginia Tech) (8,9) Hybrid Ford Escape, "Odin". *3*
• Stanford (1) VW Passat, "Junior". *5*
• Test Area A     A
• Test Area B     B
• Test Area C     C

  TEAMS WITHOUT Timely UPDATES as of Wed 31Oct
  AnnieWay [Ger](10) VW Passat. *Q*
  Autonomous Sol. - Utah, Toyota hybrid SUV, "Ted".-N-
  AvantGuardium [Isr.] - "Ronin". -N-
  Axion [4]- Calif.(7th) "Spirit". -23-
  AustinRobot/U.Texas - (24) Isuzu SUV. -N-
  Berlin [Germany]. -34-
  Case Western R. (Cleveland} (6) DuneBuggy, "DEXTER". -N-
  Cybernet, Ann Arbor, Chrysler minivan, "Cyber Van". -N-
  Franklin - Penn/Lehigh, "Little Ben", Prius. *4*
  Golem - (14) 5 guys, Prius, R.Mason. -22-
  Gray -(4) SUV. -N-
  Intelligent Vehicle Systems (17)- Delfi,Ford,Honeywell, F-250, “Desert Tortoise”. *Q*
  MIT - . *Q*
  Mojovation - (13) 10 guys. -24-
  Ody-Era - Kokomo, IN. -30-
  OSU-ACT - (10) Ohio, "Act". -N-
  Juggernaut - (q) DesignJug, SLC,UT,  "Urban Juggernaut". -26-
  SciAutonics / Auburn,CA (16). -31-
  UCF - (U.Central Fla) (q) "Knight Rider". *Q*
  U.Utah, "Red Rover". -33-
  Urbanator - Colorado, Chevy Tahoe, "Rocky". -28-
  

  ---

      The purpose of the DARPA challenge is to save lives on the battlefield. But we will see safety features installed on domestic cars soon (obstical detection and breaking, etc.) to save some of the 40,000 lives lost each year in vehicle accidents. For example, as the Challenge was starting, elsewhere in California, more than 100 cars and trucks crashed on a fog-shrouded freeway, killing at least two people and injuring dozens more. Other applications sought by team sponsers are automatic machines in mining industry (Caterpillar) and for automatic clearing of snow from airports (Oshkosh). What are your ideas?

  ---

  URBAN CHALLENGE. National Qualifying Event . 35 teams competing for 20 spots in the final Urban Challenge next Saturday. The DARPA Urban Challenge, driverless car race, website is:

  http://www.darpa.mil/grandchallenge/index.asp

      They are not doing daily updates.     There will be a web broadcast of the big race next Saturday.
  Expectation was for the 20 finalists to be announced Thursday, but two groups of teams have been told they don't need to continue with the tests. Most have benn invited to other round of trials, so . . .  
  Well, only 11 teams Qualified to enter the final contest on Saturday. All other teams had safety issues that made them a danger to other teams on the challenge course.
      Only 8 of the 35 teams had progress blogs online on Saturday. The first test, last Friday, was safety check and E-stopping. Everybody reporting in Blogs passed this test with much pride and satisfaction. Most team pages contain lots of pictures and videos. Several more teams have added blogs over the weekend.
      These blogs contains confused names of test areas and times allowed. I suspect one problem is they tend to call their first test as "A" whereas it may have been in "Test Area C". I have changed some to reduce confusion, but later gave up. The DARPA handout says:
  Test Area A. is a rectangle with various obstacles. See maps above.
  B . is the largest and most complex layout with four-way stops and road blocks.
  C . is a belt buckle, an oval with 12 event cars creating traffic in both directions and a road across the middle in which the test car makes all left turns to cross the center of the oval to test left turns through traffic, cross the oval, and another left turn to reenter into traffic.
  
      These cars are optimized for the test requirements and would not be comfortable to ride in. They step-out promptly when the computer says it is clear to go, they turn sharply, and they brake quickly when an obstical is detected, real or imagined.
  Some Complaints:
  • The dark blue test cars are difficult to see against an asphalt background. Checkout how "Spirit" got rid of one http://www.tgdaily.com/content/view/34618/113/ by crashing it.
  • DARPA hung some overhead barriers without pillars to indicate an obstacle (higher than a RR crossing) ; cars smashed windshields and wrecked sensors on top of the cars.
  • Concrete barriers are too close to the driving lane -- rules say, and cars are programmed to stay 40 inches from barriers, forcing some into oncoming traffic.
  • Traffic lanes are only 11 feet wide, instead of a normal 14 feet.
  • CajinBot says some qualifiers did not attempt difficult parts of the test, but simply stopped. Noting that they attempted all parts and that DARPA may have dumbed down the results in favor of absolute safety rather than task completion?
  Examples of Things Found Only by Field Testing:
  Safety is such a concern that vehicles have stopped because of:
  A piece of reflextive tape on the road.
  A cloud of dust. In one case this dust moved into a blind spot so the vehicle could not see it dissipitate, so it safely remained stopped.

  ---

  Extracts from BLOGS on Monday p.m.

  ---

     Team-LUX, from Hamburg, Germany with a VW :
  29.10.2007
      Today was a great day for Team-LUX ! Our LUX car gave an almost flawless performance in NQE testing area A [C]- which was the hardest one for many teams so far ! He had to drive as many laps as possible on a road with moving traffic -- and find a gap between moving vehicles coming from ahead or from both sides in order to do a left turn. Not an easy thing to do for an autonomous car! LUX did 14 laps in 30 minutes, only 2 other teams could do more by now! We are very happy and hope it goes on like this!
  28.10.2007
     NQE area C [B] was where Team-LUX had today's testing : the missions were 4-way stops, a U-turn and finding a new route. LUX did a great job in the 2 latter ones. Unfortunately, a software problem troubled the 4 way stop mission - a real shame as this is one of LUX's strengths. Luckily, the team could detect the bug very quickly and LUX will have the chance to present his strong side in the second run in area C on Tuesday!

  ---

      Carnegie-Mellon's (2,3) entry "Boss". Boss’s ability to handle contingencies was tested in expected and unexpected ways this morning as the robot completed the third and final test run in the first NQE series.
      This run, which began at 8:18 a.m. Pacific time, was over a loop course in a former suburban subdivision, [B] though much of the course could not be seen by spectators. The run tested the vehicle’s ability to handle four-way stops – determining which vehicles at the intersection have precedence and then taking its turn. That was no sweat for Boss.
      Things got more interesting about 13 minutes into the run. DARPA officials placed traffic barrels across the road, blocking Boss from reaching a checkpoint. Boss responded appropriately, figuring out an alternative route and then executing a perfect 3-point turn to reverse direction.
      When Boss found its new route also was blocked it began another 3-point turn, pulling up on the far curb and then backing up. It only needed to turn left to complete the turn, but Boss was indecisive, sitting motionless while turning its front wheels left, right, left, right. To members of Tartan Racing, watching from a few hundred feet away, it was if Boss was waving at them.
      After about a minute and a half, a DARPA official came over to ask the team what it wanted to do. They opted to do nothing. Boss was well aware of its lack of progress. After a certain interval without progress, Boss knows it must re-plan its actions. Just moments after the DARPA official asked the question, Boss began moving again. It completed the turn and, in short order, the test run as well.
      But as the team learned later, Boss’s movement had nothing to do with a new plan. What happened is that Boss kicked up a cloud of dust as it backed up. The vehicle’s sensors interpreted the dust as an object. Unfortunately, the cloud of dust also entered into Boss’s blind spot. So, when the dust settled, the sensors no longer showed any obstacle, but the vehicle’s brain still had to assume that there was an obstruction in its blind spot. Perceiving its way blocked, Boss refused to move.
      Luckily, the “waving” of the front wheels must have moved Boss just a tiny bit – enough to reveal part of the blind spot. That’s when Boss realized there was nothing in its blind spot and began to move again. It was a scenario the team had never seen before; changes are being made so Boss doesn’t get flummoxed like this again.
      This afternoon, Boss will begin the second series of NQE test runs, which continue through the end of the NQE on Wednesday. Boss will repeat Test Area B, where the robot must maneuver around obstructions in narrow, winding roads and park in a designated spot in a parking lot. Boss nailed its first run in Area B on Saturday morning, but most other teams have failed to finish their runs there.

  ---

  Insight racing, (12) Cary, N.Carolina, Lotus Elise, named Lone Wolf.
      People always ask “how is your team doing?” The answer is “We don’t really know.” So we will continue to give you our best guess at where things stand.
      Unlike a sporting event, the results of the Urban Challenge runs are not posted anywhere - either now or in the future. I’ll attempt to describe how things work at the National Qualification Event(NQE).
      There are 3 courses here at the Urban Challenge. Two of the courses are largely hidden from the view of teams, media, and spectators. DARPA wants the robots to be robust enough to run in an area with conditions we have no knowledge of beforehand. [So many teams redefine what goes on in Test Area A,B,C that I wonder if DARPA has not redefined them from the spectator's map handed out online?]
      Each team will make 2 runs on each course at a time specified by DARPA and given to us at the NQE. The robot is evaluated by many judges along the route, with judges both in safe areas and also in cars interacting with the robot. In the NQE, each robot runs the course by itself (along with traffic generated by human drivers). At the end of the run, we get neither score nor feedback on how the robot did. The reason for this, I believe, is that there are almost an infinite number of things that can happen with a robot driver, even under closely orchestrated driving conditions. If the robot makes it through the entire course in the timeframe allowed, you will see it come to the finish line. If it does not, part of the launch team goes out and retrieves it along the course. Most robots are not completing the courses, at this time.
      For Course A, you can see your robot the entire time. That is where left turning into traffic is tested. There is a lot of human-driven traffic in this test. If a human driver thinks the robot made an unsafe move, he or she hits the horn and a traffic violation is recorded. DARPA has taken great pains to ensure the human-driven traffic generated is very consistent for each robot. The drivers spent days practicing the spacing and speed for this course.
      Course B tests driving and includes a street containing a lot of parked cars. I’ve heard it called “the gauntlet.” We cannot see it, but know that our car drove the gauntlet successfully by checking our logs.
      Course C tests intersection skills. We can see the intersections well in Course C, but we can not see much more than that. There is also blocked road testing in course C. Although our robots have a digital map, they have to be able to react to many unknown conditions, such as blocked roads. Some very complex intersection precedence scenarios were presented during this test. We could see the intersection work done by the Lone Wolf and it appeared to handle it all correctly, from our vantage point.
      At the end of all the testing, the finalists will be announced. I expect that on Thursday. There will be no standings, no results, and no criteria announced ever. The decision of the DARPA judges is final. You know this going into the race, so you decide whether that is okay with you when you sign up.
      So . . . we will continue to update you with our best guess as to where we stand, but it is just a guess. It is based on our observations and on how other teams and spectators report they’ve done. Keep in mind that this whole event is mind-boggling to think the robots are doing what they are. It won’t be long until talking on your cell phone won’t be such a danger when your car is doing the driving for you!

  ---

     Stanford (1)
  Oct. 29, 2007 — Junior, the robot of the Stanford Racing Team, is now in its fourth day of testing at the National Qualifying Event. So far testing has been going well. With three days of testing to go, the team hopes Junior will continue its strong performance and qualify for the Urban Challenge race at the base Nov. 3.

  ---

     Virginia Tech, (8,9) Porsche Cayenne named "Sting".
  Monday PM: Sting Racing bounces back !   Posted on October 29, 2007
      We’re very happy with our run on Course C today. In this course the vehicle runs loop after loop in a housing area, and other cars (driven by stunt drivers) interact with it at four-way stops. The robot got stuck once, requiring an intervention — our lead programmer had to restart one of our processes. We’re still accessing the cause of the problem. But overall the vehicle behaved very nicely.
      Tomorrow morning we’re back on in Course A. This is the one that is grinding up lots of cars — besides our accident, there have been several others there. We’ve been investing a lot of time tuning our code for that test.

  ---

      CalTech, "Alice" (21)
      * 29 Oct 07, 5 pm: Alice completed its first full run at the NQE! After running the Area C test course this morning (see below), Alice was tested in Area B in the afternoon. The Area B course consists of approximately 2 miles of driving, including a parking lot, a road with cars on each side that have to be avoided, and a region with an opening in a fence. Alice completed all of these tests and came back to the start area in 25 minutes, completing its first full run at the NQE. The logs indicate that there were some small performance issues where we still want to tune Alice's behavior, but overall the run was very good (and a great accomplishment for everyone on the team!).
      * 29 Oct 07, 12 pm: We started the day in Area C, which consists of a number of intersections with human-driven cars, of increasing complexity. Alice passed all of the tests, including an intersection with 4 cars queued up and Alice having to determine when it was its turn to go. At the end of the test Alice had to find its way past a number of blocked roads, which it did successfully until a program failed and was automatically restarted. Unfortunately, when the program was reset it forgot which roads were blocked and so it tried the same paths again.

  ---

     CajunBot (15th) U.of Laf,Louisiana. Jeep named "The Bot".
      Day 3 performance, thumbs down.   October 29, 2007 (1322 hrs), VICTORVILLE, CA
      Our Day 3 performance was not really good, but we knew the issues going in. We still need to take care of a few things, before we'd be ready for passing this test.
      In this test, held in Area B, the bot is expected to get out of a start chute, navigate an open space, go through a narrow winding lane bounded by massive K-rails (concrete barriers), drive around a traffic circle, and then drive into a very convoluted neighborhood. We do not know what all goes in the neighborhood, but the general guess is that it has stalled vehicles parked on the side of the roads. The bot needs to safely travel through that area, avoiding rear ending any vehicle. In this area is a parking lot. The bot has to find a designated parking spot, not run into vehicles parked into adjacent lanes, and then get out of the parking lot. Finally, in the area is an 'obstacle zone.' This is an open space, with barricades forming a gate. The vehicle has to find the gate and go through it.
      CajunBot had trouble right out of the start chute. It was stumped by all the K-rails along the border of the open area, and decided not to budge. After two unsuccessful attempts, Suresh did some quick thinking and made some changes. That made CajunBot, go past the initial hesitation, travel through the winding lane, and onto the traffic circle.
      Just short of completing the traffic circle, CajunBot came to a stop. For a few minutes you could see her attempting to take off, but stopping again. The trouble shooting team, consisting of Mark and Suresh, were rushed to the site. They attempted a few things, but nothing helped. By then we were out of our allotted 40 minutes.
      The good news is DARPA announced a second round of tests in the remaining 2.5 days (Monday PM to Wednesday PM). Earlier they had announced the schedule only through Monday AM. We expected that after the first round of tests in the three areas, some teams will be selected as having passed the tests, and others will be given another go. It turns out all the teams are scheduled to run in the second round of tests. I can infer that to mean that there is clearly no perfect performer.
      In spite of all the long hours, the team is holding up well. Everyone is stretched to the limits, but no one is cracking. There are no outbursts, or name callings, or "I want my mommy" calls.

  ---

      Several teams have new Blogs up.

  ---

      Princeton (19th) .
  Sunday, October 28, 2007 5:53PM PDT
      We just returned from our first run on NQE Test course B. Having overcome the various technical issues that plagued us over the last few days we were able to start up all of our systems and start our run. As soon as the vehicle left the start line, however, we knew something was wrong. Our audible warning signal (which, for anyone who has heard it and was wondering, beeps out “PAVE” in morse code) was going extremely slowly. This told us that the computer was not functioning well. Sure enough, the slow update rate caused our steering controller to oscillate wildly and our vehicle was disabled about 100 feet down the course.
  Saturday, October 27, 2007 11:36PM PDT
      For the past few days, we’ve been busy troubleshooting a variety of problems. Since yesterday, we’ve been wrestling with problems in our State Estimation code and our mechanical E-Brake piston, both of which were preventing us from successfully finishing our E-Stop test. Luckily, we were able to pass the test just in time. Since then, we’ve been working on preparing our navigation code and car detection code for tomorrow’s runs. We’re scheduled to do a qualifying run at 4PM tomorrow, so check back after then for more updates.
  Friday, October 26, 2007 3:14PM PDT
      We had our E-Stop Test this morning, and were unable to pass it. Before we could even get off the start line, one of our computers died with a stop error. Once that issue was rectified, we were unable to get our GPS and State Estimation code to properly display our location & heading. Stuck with these problems, we went past our 30 minute time limit and are going to have to re-schedule our E-Stop test. We currently think that the move from East Coast to West Coast may have affected a few parameters and hope to have this fixed soon.

  ---

  Stanford Racing, (1) VW, "Junior"
      Oct. 29, 2007 — Junior is now in its fourth day of testing at the National Qualifying Event. So far testing has been going well. With three days of testing to go, the team hopes Junior will continue its strong performance and qualify for the Urban Challenge race Nov. 3.

  ---

  Jefferson, U.Virgina "Tommy Jr."
      Today's format was a repeat of yesterday's. : merging into moving traffic, which was today's challenge for Team Jefferson. Area A is cool to watch. A large loop of moving cars that start and stop on command : the tightly choreographed stunt drivers serving as robot fodder. The way this group would move and halt reminded me of the Visa commercial which has the diners coming to a screeching halt when some hapless customer uses cash.
      The robot makes a smaller loop : moving with the traffic, then cutting across it and cutting into it in a series of sequential left turns. Robots had various degrees of success "shooting the gap". Caltech seemed to be overcautious - detection of the barrier wall part-way through the turn would cause a halt. This would expose the vehicle to oncoming traffic, a stop, and an angry honk from the stunt driver.
      We did see a moment of contact. Blue Ford Taurus number 168 was impacted in its rear corner as it tried to scoot out of the way. I believe Axion was the perp. This created a delay to check health of the driver and vehicle.
      Scoring was relatively simple: you were scored for hitting the stop-lines, merging without impeding thru traffic, and completing maximum loops.
      The best robots were handling the merge and timing well. Others were having trouble with the left turn, which was admittingly tight against the K-rail. CMU's Tartan team seemed to have no trouble with the merge itself, but had trouble with the basic challenge: navigating the loop it would detect obstacles on the curves.
      We were called 45 minutes early and had to perform various elements of route planning, detection, and speed control. Also changing parameters like lane width. I probably don't need to say how nerve wracking it is to watch programming and compiling in a moving car 45 minutes before a trial.
      At one point, I casually stuck my elbow out the window, which caused Tommy Jr. to gyrate and swerve left, spewing several pages of log items. "Get your arm back in the car, its seeing it as an obstacle"   Tommy's run was imperfect. When he approached the first merge, he seemed overcautious and took a full minute to advance, nudging himself forward. It may be the case that he felt there was an obstacle in his defined intersection - such as a post or barrier. We are still assessing the data.   He did finally enter traffic, but he hit the curb on the far stretch. A number of robots have been having trouble in this spot, and we suspect there could be an issue with Darpa's mission file. In fact, just as I sit in the pit writing this, our neighbor Golem Group is having its robot towed into the pit. It hit the same curb and flattened its tire.

  ---

     Cornell U., Chevy Tahoe
  Sunday, October 28, 2007
      We've run twice so far -- once each in two out of the three NQE areas. We didn't make it through area B- a basic navigation run, for one legitimate reason and one stupid reason. We've promised not to give details of what happened exactly, since the course is not visible to the public. We drove well and failed safely though, so I'm hoping DARPA will consider that as NQE progresses.
      Area A was a heart attack waiting to happen- a circle of bi-directional traffic into which our car had to merge repeatedly. The traffic was supplied by DARPA- a fleet of specially-outfitted Ford Tauruses driven by professional stunt drivers, no less. The whole thing is very eerie... concrete barriers and vehicle sirens drown out the Taurus' engines, and the stunt drivers are all very synchronized. It gives of the distinct impression of being a creepy carnival ride starting and stopping, so much so that I think it would be complete with a little slow waltz or circus music in some minor key. The situation is very difficult to merge into, with very tight tolerances required to avoid the evil fleet of Tauruses on your left and some menacing concrete barriers on the right. I had a terrible night sleeping last night, thinking, 'frick, the car's gonna crash, the car's gonna crash...' and then waking up and whirling about how this or that under some impossibly unlikely situation would cause our car to destroy its front bumper- a $28000 mistake. To top that all off the entire course is very small and completely visible to the public, so you just have to sit there and sweat bullets while it's all happening. It's the kind of thing that excites morbid curiosity: you don't want to watch, but you just have to do it anyway.
      Well, to cut a long story short, our car made it through. (!) We didn't cause a single accident (the DARPA drivers prevent accidents, but honk if they would've occurred). We checked the data logs- we came within 20 cm of the concrete barrier at one point (make fists with your hands and stack them one on top of another: that's about 20 cm). I'm glad the guys told me about that *after* the run was over, because otherwise I'd have to go change my pants. Still, it's pretty slick (*cough* scary *cough*) that the planner knows its boundaries that well.

  ---

     Oshkosh, marine Truck named "TerraMax"
  Sunday, October 28, 2007 : PRACTICE MAKES IMPROVEMENT
      On a blustery afternoon in Victorville, Team Oshkosh took its second shot at area “C”, which measured TerraMax’s ability to take its turn at a residential intersection, perform ‘y’ turns and re-plan a route when a road suddenly becomes obstructed. Although TerraMax performed well during its first attempt at the course on Saturday, the team was not satisfied. The software was reviewed, and the team went about improving on what was already considered a capable vehicle.
      Imagine turning a 22 foot, 25,000 lb truck around on the street in your neighborhood, Terramax performed the maneuver with ease. In addition, the team was encouraged by the truck’s increased abilities at intersections and avoiding obstacles.
      A noticeable point for improvement was when the truck stopped just short of the finish, believing it had completed its mission. A quick response by the crew ensured the truck finished under the wire and will continue the pursuit to qualify for the Grand Challenge on Saturday.
  October 28, 2007 : NOT YOUR AVERAGE GROCERY GETTER
      As the sun rose on the second day of the National Qualification event Team Oshkosh was busy testing TerraMax in an off-site location. Just after lunch, TerraMax entered test area “B”, a maze of winding residential roads, parked cars and parking lot navigation. Suburbia suited TerraMax well, as the truck completed the majority of the course before time expired. Having rehearsed each of the testing locations once, the team is excited to make some informed adjustments before heading back to the testing areas tomorrow.

  ---

     Victor Tango (Virginia Tech) (8,9) "Odin", Hybrid Ford Escape
  Sunday, October 28, 2007
      Odin is tested on the 2nd and 3rd qualifying events, B and C. The first test of the day, event B, takes place first thing in the morning at 7:30. Quite used to the early mornings at this point (the picture to the right is taken at 5am), the team is in place well before the designated launch time. Test area B includes the most diverse set of challenges. Perched on a balcony, team members look on as their vehicle exits the start chute. Odin is able to finish a good portion of the course before getting stuck at "the gauntlet", a maze of disabled vehicles DARPA scatters in the road. The team is allowed to reset Odin and continue. Odin continues to run the course until time has expired. While the team was hoping to see Odin finish the course, performance-wise Odin compares very well to the competition.
      The second test area of the day is run at Known as "Area C". This test is designed to confirm that the vehicle can follow precedence at 4 way stops, and can detect road blockages and re-plan a new route accordingly. When the start horn is blown, Odin approaches his first intersection and proceeds as expected. The course circles around and brings Odin back to the same 4-way intersection repeatedly ; and repeatedly Odin waits his turn to proceed through the intersection. DARPA throws the first curve ball and blocks a road that Odin expects to travel through. Just as rehearsed, Odin detects the blockage and makes a 3 point U-turn. According to unofficial sources, Odin was either the first, or one of the first, vehicles to complete this test.
  Oct. 27, 2007, Saturday
      Victor Tango is chosen to be the first of the 35 teams to be tested on Qualifying Event A. Up until 5 minutes before the test, the team will have no idea what the test will encompass. After the mission is loaded in the computers, Victor Tango must simply sit back and watch Odin, hoping that they have taught him well. Odin approaches a stop sign at a T-intersection at which he must make a left turn across traffic. Coming to a complete stop and waiting as several vehicles pass, Odin appears to be doing everything perfectly. When a gap is spotted, Odin proceeds to make a safe left turn. The course continues this way for about 30 minutes with only one major hiccup causing the vehicle to stop. At one point, Odin's sensors see a phantom obstacle caused by a strip of reflective tape on the ground. Unable to get around this ghost obstacle, Odin sits patiently until he must be reset. This unlikely occurrence arises only once. Odin is able to pass the same place more than ten times without the phantom object arising.
      The team is happy with the results and feels they have set the bar high. As the day progresses, this proves to be true. Directly after Victor Tango, a team comes within inches of crashing head on into a traffic vehicle. The stunt driver is able to avoid the bot through emergency maneuvers. Later on in the day an autonomous Porsche Cayenne crashes headlong into a K-rail.
  Oct. 26, 2007
      Again the team has a busy morning preparing for a 9:30 practice time. However, everyone's tests today must begin with a mandatory DARPA Safety Inspection and E-Stop test. Just having experienced a failure with the E-Stop DARPA provided us, many team members are uneasy about the test.   Odin passes both the Safety Inspection and E-Stop test. One problem that crops up is collection of dust on one of the rear sensors. Not used to the sandy desert environment, Odin has trouble with the caked on particles taking them as an obstacle.

  ---

  Extracts from BLOGS on Monday a.m.

  ---

     CarOLO, Volkswagen Passat, named Caroline, a German team, writes a good blog for the folks back home --
      " Many competitors were struggling with the testing course, one team even had a head-on-collision with a K-rail. The K-rails separate the testing area and the spectator areas. The K-rails are positioned too close to the right lane. "Caroline" deemed it safer to drive on the left lane -- directly into oncoming traffic. Well, lets say, the first test went "suboptimal". And on the testing areas B and C only two teams finished the whole course autonomously. Many bots had problems leaving the start chutes, and when they did, some decided to stop after a couple of meters. A few teams performed very well, like our fellows from Karlsruhe [Team AnnieWay. There are 4 German teams.], or Stanford and Carnegie Mellon, who passed test area B in 17 minutes.

  ---

      Team-LUX, another German Team:
      Today was the first NQE testing for The LUX car had to find his way through a city within 45 minutes. The mission was a real challenge with many tricky tasks like overtaking, parking and unparking between 2 cars. Although he didn't cross finish line - he was very close though - Team-LUX is very proud of today's performance and looking forward for the next NQE mission tomorrow morning! [was probably test area B]

  ---

      Carnegie-Mellon's (2,3) entry is named Boss. [They came in 2nd and 3rd last time. Only five completed that desert course in 2005. Stanford was first.]
      A dozen human drivers created dense, two-way traffic and "Boss" found its way through it. Boss had to find gaps, choose opportunities to enter and cross traffic, and make its way around a ‘belt-buckle shaped course. A perimeter of jersey barriers encroach on the outer lane, so the course is tight and unforgiving – traffic or no traffic. The lanes are only 11-feet wide (versus the standard 14 feet), so Boss had to thread the gap between oncoming traffic and the rail. Boss made the turns across traffic without incident and drove aggressively so it could complete the 16 laps in the allotted time. The run was not without flaws. Human drivers gave Boss a horn or two. [A "horn", close call, loses points] Several times, it stopped abruptly as it was driving along the jersey barrier and approaching the end of the oval. At one point, it stopped short of the turn point at the middle of oval and, at the request of the team, DARPA officials momentarily paused Boss. At another point, Boss nicked a curb to avert oncoming traffic. Team leader Red Whittaker commented that “The belt-buckle is an unnervingly difficult event. We are thrilled that Boss drove well and survived crash-free. This wasn’t perfection, just best-of-class so far." "Boss" completed 16 circuits of the course in the allotted 25 minutes. That is best-yet, but more is possible, and some great competitors have not yet run this. The next-best performance so far is 11 circuits scored by Virginia Tech. It was the second strong performance for Boss. Two other teams - MIT and Germany’s Team CarOLO – have completed the long, narrow, winding roads with parked cars and other obstructions in Test Area A that Boss aced on Saturday. Tomorrow, Tartan Racing [Carnegie-Mellon] will complete the first round of testing at Test Area B, where vehicles display their ability for tight maneuvering, handling four-way intersections and finding alternate ways to reach blocked goals.
      Georgia Tech/SAIC Sting Racing team's {Actually Virginia Tech] Porsche Cayenne ran head-on into the jersey barrier [K-rails to the Germans] at Test Area A on Saturday - the front equipment bar was cut off, releveled, and welded back on by the Carnegie-Mellon team.
  
      The number in (00) is placement two years ago in Desert Challenge.

  ---

     Virginia Tech, (8,9) Porsche Cayenne named "Sting".
      Saturday afternoon test in Area A "Sting 1" crashed head-on into a concrete barrier. The front sensor mount was bent severely and pushed into the front of the vehicle. After assessing the log data we determined the cause of the accident was a failure in the communication link between our GPS/IMU and the main control computer. About 9 pm last night we were able to get things straightened out and recalibrate in time for this afternoon’s next test. DARPA says we can re-do yesterday’s tests as well, which is great. Many thanks to our friends at CMU for being so supportive and helping us repair the vehicle.

  ---

      Checkout the largest and smallest entries : a marine truck and a sports car. http://www.insightracing.org/images/10-27-2007NQE/pages/DSCF1377_jpg.htm

  ---

     Insight racing, (12) Cary, N.Carolina, Lotus Elise, named Lone Wolf.
     Our first run on the morning of the 27th was a test of turning left into traffic. There were 12 human driven vehicles circling the testing areas moving in both directions. [Test Area C] The first turn was a test of pulling out and making a left turn into moving traffic from a stop sign. The second test was left turn from a lane of moving traffic. Many teams were unable to make the first turn. Some made a few turns. One team made around 12 laps. "Lone Wolf" was in the middle group with about 3.5 laps. One lidar visor cover touched a cement barrier, but no damage resulted. Overnight we added reflectors to make our car more visible to other entries.
      [Day 2, 28th.] We had 2 runs today. On the first run, after a great start, we encountered an unusual condition. Upon restart, the Lone Wolf took off at a good clip and drove around parked cars, cones and barrels [Test Area A]. In our next run, we were tested extensively on interacting with live traffic at 4 way intersections. [Test Area B] The Lone Wolf appeared to perform flawlessly for 25 minutes. We will analyze our video and data from the runs this evening

  ---

     Gator Nation, (18) U. of Florida team's Toyota Highlander Hybrid
     is equipped with 13 dual-core computers, three global positioning systems, five video cameras, eight laser scanners and an inertial navigation system. The laser scanners can sense the road and other vehicles within 600 feet of the vehicle.

  ---

      CalTech, "Alice" (21)
     This involved driving around on a set of roads that had 12 cars circling in the inner and outer lanes. By making a series of left turns, Alice had to repeatedly merge into the traffic and cut across opposing lanes of traffic while it followed its route. The task was made more difficult by the fact that the lanes were very narrow and had K-rails as barriers on one side. Alice completed quite a few laps, but did have a couple of aggressive merges where it came close to other vehicles. Tomorrow we have two tests: an intersection and U-turn test in the morning and a second attempt at the Area B test (navigation and parking).
      [Alice had trouble the previous day getting out of the start-chute because of the tight fit -- notice that everybody is complaining that the lanes are too tight. The collision avoidance sensors stop the vehicle. Testing back home was done in wide parking lots. CalTech loosened their standards that resulted in the "close to other vehicles" noted here. Wonder how the Oshkosh Marine Truck will do?]

  ---

      CajunBot (15th) U.of Laf,Louisiana.
      Day 1. We could not finish Test Area B. The area is a ring with two 4-way stops (intersections) on opposite end. The bot is expected to go around the ring, and at each 4-way stop she is met with other traffic vehicles. She has to maintain the rules of a 'stop' sign, give precedence to other vehicles that come before it. CajunBot did well for the first two interactions. In the third interaction, there were three vehicles already at the intersection. She let two vehicles go, but then took off before the third vehicle moved. This was a complete surprise to us. You kind of get a feeling she was irritated to be given such stupid test. As the loop progressed she was to encounter a blockage in the road and expected to turn around. This is also something she has done many times. But this time around, she just went stopped in front of the blockage and decided not to move.
      What caused this misbehavior? Our analysis indicates two things: first, it looks like after two loops "the bot" began experiencing a GPS drift, she was shifting the middle of the road about two feet to the right that made her ride the curb. Second, we have never tested in a situation like this before. All our tests are in wide open spaces. This is the first time we are testing in a real neighborhood. The main difference now is that our sensors pick up a lot, lot more data in this environment. In the process something has tripped.
      Day 2. The test in Area C is plain and simple scary. The track consists of a two lane (one each way) rectangular road, about 120m long and 40m wide. Across the middle of the long segment, top to bottom, runs a one lane road. The test setup consists of 12 cars making laps around the course, six in each direction. The mission is to loop around as well, but not to travel the full rectangle, instead to cut across the rectangle, along the one lane road. (Belt buckle]
     The vehicles in the outer loop have the right of way. The test requirement is that when CajunBot turns left from the outerloop to the one lane segment, she has to yield to the traffic in the oncoming lane. When she makes the next left turn from the one lane segment to the outer loop, she has to yield to the traffic coming from the left and the right.
      The traffic flows around 8 to 12 miles/hour. The traffic vehicles are manned by race car drivers. They are driving specially built vehicles with amazing amount of reinforcement. The drivers too are in full race-car regalia, with helmets and suites designed for race car driving.
      The evaluation is based on number of loops completed in the 40 minutes allotted, the number of loops made without any safety error. There is penalty for making the other drivers honk. There is also a penalty for taking too long to merge into traffic.
      I didn't keep track of the count. The experience was too gut wrenching to worry about counting loops. The grapevine has it that our performance is amongst the top quartile. The team is very upbeat.
      Day 3. If you do not believe in angels, this should make you a convert. After looking at the data from Area A test we are completely stumped why CajunBot did not ram into the traffic vehicles. She came quite close to running over some of the cars. But she always stopped, just short. Our conclusion is the vehicle was running WITHOUT its primary sensors, the Ibeo and SICK lidars. The physical sensors themselves were working, but our program that analyzes the sensor data to identify objects in the world had crashed. And by crashed it means, the program was not running. The sensors were generating data, but there was no program using the data. For most practical purposes, CajunBot was running blind. CajunBot was running on her secondary sensors, the radars, two of which were acquired and mounted after we came here. Those saved the bot from total destruction. There must be an angel watching over CajunBot.
      We have diagnosed the reason for the crash. When we run the bot in Cajun Field, a wide open parking lot, CajunBot sees a very few (between 50 and 100) objects. Our obstacle detection and tracking system was configured to track 100 obstacles. But in the test area, an abandoned air base, there is a lot of clutter, The area has buildings, trees, bushes, and other traffic. As a result the sensors see about 500 obstacles. This exceeds the max limit of 100 programmed into the system, and our obstacle detection program simply crashes. The fix for this problem is rather simple, and has now been taken care of. QUALIFIERS" 1 . Tartan, CMU, Boss 2 . Cornell 3 , Victor Tango, VirTech, 4 . Ben Franklin, L'l Ben 5 . Stanford, "Junior" Scratched: 37 . Georgia Tech 36 . Berlin 35 . U. Utah 34 . Princeton 33 . SciAuto/Auburn 32 . Ody-Era The word was not so good for six other teams. Georgia Tech/SAIC Sting Racing, Team Berlin, University of Utah, Princeton, SciAutonics/Auburn Engineering and an Indiana team called Ody-Era were all scratched today by DARPA. Tartan Racing was the first team to qualify for the Urban Challenge final event. DARPA officials, who earlier excused Boss from any further qualifying runs, told the team early this afternoon that Boss is in the race. By 3 p.m., four other teams – Cornell, Virginia Tech’s Victor Tango team, the University of Pennsylvania’s Ben Franklin Racing Team and Stanford – also had made the race.

  ---

  Jim
  This started as an email to Carolyn, became too long, so I converted the email and my menu into a web page so she and others can read it wherever.
  Peggy -- there is no word yet about the Ben Franklin Prius (Philadelphia), nor the other Prius. But Pittsburgh, CMU, and other interesting teams are doing well.
  URL: ==   www.manorweb.com/grand.html