ftw!

¶ ORCA 8 - results Posted 20 years, 4 months ago on August 12, 2005

Yesterday we got back from the AUV competition. After 6 whirlwind days and nights, all of which are seamlessly blurred into my fuddled brain, we ended up with 3rd place out of 16 teams, behind University of Florida and ETS.

The actual competition was held out at the Space and Naval Warfare Center out in San Diego, at their Transducer Evaluation Center (TRANSDEC). A transducer is a sonar emitter and receiver, and the Navy built a pond that minimizes multipath, making the acoustic conditions similar to that of the deep ocean and ideal for testing sonar devices in.

The first few days of our trip to San Diego was spent setting up operations at the Vagabond Inn and putting the sub back together. We had to partially disassemble ORCA for the plane ride over, so the thrusters, cameras, and main electronics components got reattached once we reached the hotel. Second task was getting Internet access. Since the Vagabond Inn doesn't currently provide Internet access for its guests, we signed up for a Verizon 3G BroadbandAccess account that we canceled a week later. On Thursday morning, we hauled ourselves out to the transdec and began operating from there during the day. Each team is allocated a tent with some basic utilities like power and tables and chairs.

After we got setup at the TRANSDEC, we began our first practice runs in the actual pond. Team members aren't allowed in the water, so the Navy provided professional divers to operate the sub while we test. We designated Sam as our official dock man, which meant that he was the only person communicating with and giving instructions to the diver. Once everything is ready to go, they hoist the sub into the water and off we go. Unfortunately, our first test run was a complete disaster. Nothing worked the way it was supposed to and after a few minutes we pulled out of the water to do some diagnostics. After a couple hours we figured out and fixed the two problems that killed our first practice run. The first problem was that we put the vertical thrusters on backwards. Yes, backwards. The fore and aft vertical thrusters are a matched pair so they spin differently when given the same commands, and we had the fore thruster where the aft thruster should be, and the aft thruster where the fore thruster should be. So whenever we told the sub to dive, it would try to jump out of the water. The second problem was a wiring problem which we traced down to a bad firewire cable that was bringing down the entire firewire subsystem, which meant our cameras didn't work. Once we figured that out, we swapped a good cable in and resumed testing.

At 7:00 PM, the first day of practice at the TRANSDEC ended and all the teams packed up and hauled out. We grabbed a quick bite to eat and immediately got back to work at the Vagabond Inn. The management there is great to us and lets us use their pool at night, so we pretty much camped out by the side of the pool while we worked. A few of us got tired and crashed for a few hours and the first night shift began. At around 2:00 AM, the sleepers woke up to relieve the first night shift and kept on toiling until early morning.

The next couple days were largely spent testing the sub, fixing bugs, and writing software to handle the actual competition run. By the second day of practice, a long waitlist for practice time developed, so during the times that we couldn't use the actual TRANSDEC, we tested the sub in the small dolphin tank by the side. It wasn't quite the same as the real thing, but worked for the most part.

We even had a miniature scale version of some of the course elements that we used in the dolphin tank to test the machine vision algorithms.

Saturday and Sunday were the competition days. There were two rounds of qualifiers, from which four teams were selected to compete in the finals. It ended up being us, UF, ETS, and Duke. During each competition run, teams were given 5 minutes of dock time to bring up and check their systems, and get the sub in the water. For us, this meant booting the sub (which now runs Debian GNU/Linux), calibrating the sonar equipment and the cameras, and loading the mission scripts. After that, each team has 15 minutes of competition time to accumulate as many points as it can. For our first qualifying run, we made it through the validation gate and completed the most valuable course element - surfacing in the surfacing zone right above the sonar beacon. That was enough to get us into the finals, which was good.

Saturday night was spent refining the machine vision systems and writing mission scripts, to the point where we were able to successfully complete every course element in the hotel pool with 90% reliability. We were going to test this in our second qualifying run, but that ended up being a bust because we entered in some wrong numbers during our dock time and hosed the sonar equipment for that run. We were still reasonably confident, though, and got in a few practice runs before finals. We made our final competition run at 3:30 PM on Sunday, which ended up being a minor disaster. Three minutes into our dock time, we realized that the firewire subsystem had failed again and our cameras weren't working, so we tried a cold reboot. That didn't work, so after deliberating for a few seconds, we tried hotplugging the camera connectors. With normal firewire cables, this is okay because the ground pin is much longer than the power pin, but with our custom-built waterproof connectors, this is not safe because all the pins are the same length and you risk completely frying your electronics. It was a do-or-die situation, however, so we did it and it worked, so we put the sub in the water and started the first mission run with 12 minutes of competition time remaining. The sub made it through the validation gate and to the surfacing zone, but after that it took the wrong heading and went in a direction tangential to the next mission element (we suspect the compass crapped out on us). After watching it go the wrong way for a couple minutes, we stopped the mission and were going to reconfigure it to try again with a different heading, but by that time we only had about 5 minutes left and decided that we didn't have time to risk another run (you lose all your previously acquired points when you try again).

In the end, the top three teams (us, University of Florida, and ETS) all completed the exact same course elements, but MIT was heaviest and UF was lightest so they won by virtue of being a number of kilograms lighter than us. That was pretty disappointing, but we'll be back next year to try again and reclaim our title =P The day after, we took a team trip to the San Diego Zoo to decompress before packing up and heading home.

images link to higher resolution versions.

No comments, be the first!

Comments disabled until the spammers go away. I hope you comment spammers all die horrible deaths and are forced to delete endless streams of comment spam in your days in purgatory.