Team Case

October 28, 2007

Day 4 - NQE Begins & Many Improvements

It's been a long and productive Day 4 here in Victorville. DEXTER completed the first of three NQE tests today - the "Area A" test of left-hand turns across traffic. DEXTER safely maneuvered through a course with ten stunt drivers driving in a quarter mile loop. He completed ten laps successfully and showed some nice behaviors. For example, he decided at one point that he had made a mistake in entering an intersection, so he re-planned on the fly and took a different route to avoid a vehicle. DEXTER did a good job of staying in his lane, and ran autonomously for about half an hour. Many of the teams had to go out and restart their vehicles, but DEXTER ran without intervention. Tomorrow, we will test in "Area B," which covers parking lot behavior, and in "Area C," which covers intersections and dynamic route re-planning.

We did a lot of electrical and sensors work on the robot today, which has kept me and the rest of the bunch on our toes! We noticed that the robot was throttling unexpectedly before the test this morning due to some sort of problem with the real-time computer. A quick decision by Amaury before the test got us through it - he had a hunch that it was not getting enough power from its 12V power supply, so he disconnect its cooling fan which runs on the same supply and ran the robot with the front box open. This got us through the test.

We needed to address this power issue on DEXTER -- fast! Fortunately, Jeff Munier and Jacinda Clemenzi arrived today from ENSCO and were a super great help in getting DEXTER fixed. First, we are now running the cooling fans off DEXTER's alternator-powered 12V line, instead of the 12 V power supply that's powered by the 48V line. Second, we took advantage of the Compact RIO's redundant 12V inputs and we now drive it from both supply lines. Finally, Jacinda discovered a broken 12V connector that was feeding the Compact RIO, so we took care of that as well. These fixes seem to have done the trick!

We also decided today that based on the performance of other vehicles, we should make sure we can detect obstacles that aren't at LIDAR-level. We had previously planned to do this with our Videre stereo camera, which works great at finding objects up to 50 meters away. Of course, this sensor also follows Murphy's Law, so it decided to stop working yesterday. We decided today to bring up the Point Grey Digiclops stereo camera and also mount a LIDAR at a 90-degree tilt to take care of this.

The Digiclops has been on the robot since the 2005 Grand Challenge, but we never used it much because it has a short range of about 15 meters. But thanks to the hard work by Ashwin Deo to get the driver updated this evening, and by Brad Hughes and Christian Miller to get some bugs worked out and integrate the software with DEXLAUNCH tonight, we may have it totally integrated on the robot by tomorrow afternoon's run. We will log data in the morning at the very least.

As for the 90-degree LIDAR, Arkady Polinkovsky did a great job to get the LIDAR mounted on the front of DEXTER, Jacinda and I wired the LIDAR into the Pelican box, and Scott took advantage of our very modular code base to rapidly bring up a new driver for the unit.

I should also pass along great kudos to all the folks running simulations 24/7 on the RNDF's to find any potential pitfalls and to dig up any nasty bugs, and to those working on the maps to make sure DEXTER has the best data possible about his routes before he hits the roads.

Posted by: Bradley Farnsworth October 28, 2007 01:17 AM | Category:

Posted by: bdf (Bradley Farnsworth) October 28, 2007 01:17 AM | Comments (2)

• Team Case Home Page
• |
• caseurbanchallenge@case.edu

Team Case | 10900 Euclid Ave. | Cleveland, Ohio 44106
© 2007 Case Western Reserve University | Cleveland, Ohio 44106 | 216.368.2000 | legal notice