On the 22nd of October, CUAUV along with a research group from SUNY Cortland, and members of the Ithaca High School Robotics Team boarded the Floating Classroom Project and set course for the mouth of Cayuga Lake. The purpose of our visit was to observe the effectiveness of an herbicide in deterring the spread of Hydrilla, an invasive plant species, to the mouth of Cayuga Lake. The vehicle was deployed in two locations and sent to a maximum depth of 12 meters in order to observe the lake bottom. No Hydrilla was seen from our dives; additionally, it appeared that other plant life remained unaffected by the herbicide. Our vehicle's voyage concluded by inspecting the propeller of the Floating Classroom Project.

Here is a video containing footage taken by the vehicle:

http://www.youtube.com/user/CornellAUV#p/a/u/0/GBoRBMqZgic

For more information on the Floating Classroom Project, click on the link below:

http://floatingclassroom.cayugawatershed.org/

CUAUV collaborated with Diversity Programs in Engineering and hosted a balloon-powered boat event for students outside of the Cornell community. With the help of CUAUV team members, students were able to build a boat propelled by balloons and learned the concepts behind the propulsion. They worked in CUAUV’s lab workspace and had an opportunity to learn about the team and last year’s competition vehicle.

As we are currently working on the design of our vehicle this year, we have finalized the date for the Open Pool Test to be November 20th from 5:30 – 7:00 pm. The Open Pool Test gives the Cornell and Ithaca community an opportunity to see last year’s vehicle maneuver through an obstacle course similar to the one at the competition.

Thank you to everyone who applied and congratulations to our 12 new members. Team profile information can be found on the Subgroups Page.

This year, CUAUV had the privilege to be a part of AUVSI’s Unmanned Systems North America 2011 Symposium. The event is a premier forum for reviewing, assessing, and discussing the current state of the unmanned systems industry. More than 6,500 attendees from over 40 countries participated in the conference, which took place August 16th to August 19th in Washington, DC.

The team had its own exhibit at the convention, which afforded us the opportunity to connect with our sponsors such as Seacon, Lockheed and Martin, Solidworks, Microstrain, PNI, Pelican Storm Case, Maxon, and Teledyne. CUAUV also mingled with fellow teams such as SONIA AUV from Quebec, Robotics @ Maryland from the University of Maryland, and URI AUV from the University of Rhode Island. More pictures available in the photo gallery!

Are you interested in building a robotic submarine? Do you want to get involved in an internationally-recognized, hands-on engineering project? Then you should consider joining CUAUV! The team will be hosting recruitment info sessions during the second week of classes. Applications will be given out during our info sessions and will be due Sept. 9, 2010. Looking forward to meeting you!

Information sessions:

Wednesday, Aug. 31, 5:00pm - 7:00pm, Kimball B11;
Thursday, Sept. 1, 5:00pm - 7:00pm, Kimball B11;
and Saturday, Sept. 3, 2:00pm - 4:00pm, RPCC 218.

Recruitment Schedule:

Sunday, Aug. 28: Clubfest
Monday, Aug. 29 - Friday, Sept. 2nd: Ho Plaza Tabling
Friday, Sept. 9: Applications due by 6:00pm to Upson 108
Friday, Sept. 9 - Saturday, Sept. 10: You'll hear back from the team
Saturday, Sept. 10 - Tuesday, Sept. 13: Interviews (In the ELL, basement of Upson)
Saturday, Sept. 17 - Sunday, Sept. 18: Training

It was mentioned in the last post that the vehicle suffered from a race condition in the mission code and that caused it to not perform as well as it did in its qualifying runs. This post will go more into depth as to exactly what went wrong and how the problems will be fixed in the future, for those who are interested. As a preface to this analysis, it should be made clear that the members at the dock during the run made all the correct decisions given the information that they had, and without their quick thinking the team would not have been able to salvage a very competitive run from this bug.

The first attempt at the full mission was where the race condition came into play. The software team on the dock thought that either the vehicle did not get within camera range of the yellow buoy (since their elevations were changed about 5 minutes before the run) or that the poor lighting conditions prevented vision from working on the yellow buoy. Post-run, it was found in the logs that the backup-and-sway task started (this task is responsible for making the vehicle back up after the ram and sway to find the next buoy), but the generic-find did not run (this task actually looks for the buoy). When it was run on our recording, the vision software seemed to work well enough, suggesting that it wasn’t the problem. Upon comparing the debug logs to a successful run, one of the team members, Jeff, noticed that the order of execution between these two tasks was swapped in the failed mission. This suggests a possible race condition (i.e. the correct execution of these tasks was dependent on them being added or initialized in a specific order). One mystery that remains is that both tasks should have still run because they were both non-blocking (meaning neither will prevent the execution of other tasks in parallel). This will be investigated in the coming weeks.

The second run went wrong due to human error. Unfortunately, the team on the dock incorrectly thought they had shut down the old mission and started a new one. This meant the two missions ran in parallel and “fought” for control of the sub, making the vehicle miss the validation gate completely. This type of error is somewhat unavoidable, but will be mitigated next year by having even more practice time that simulates the competition. The software team during the summer will practice having the vehicle pulled back and reprogramming on dock as if a run went wrong.

Once the second mission had been stopped, the vehicle started its third run. This was much more successful. Unfortunately, the mission had been changed to only hit the green buoy twice but the direction to search after the buoys had not been changed. This meant the vehicle did not find the correct pipe to get to the next competition element. Despite this, the vehicle timed out and completed the rest of the mission with only a few seconds remaining, getting us enough points for second place.

A few things will be done this year to fix these problems. First, the team will push even harder to get into the water on time to give the software team more time to do full autonomous testing. This amounts to more competition-like practice for the people driving the vehicle, and this will allow them to experience issues that come up when running untethered and hopefully better prepare them for competition runs. Also, our mission system is currently too complex. It makes the timeline of adding and initializing simultaneous tasks at best confusing and at worst non-deterministic. This year the system will be largely reworked. It is currently multi-threaded, but one idea is to rework it into a single-threaded program which will interleave the running of the mission tasks like multi-threading would while retaining determinism. It will be reworked so the mission tasks themselves do not have to be altered; the only ones that will change are those that depend on the quirks of the current system. There are also plans to make a “mission replay system” which will help replicate rare bugs. The entire mission will be recorded (including vision, sensor values, and other variables) and then used as an input to the mission code to see exactly what went wrong. Since race conditions are random, it will not recreate them reliably, but if an error doesn’t occur during playback it would point to a random race-type condition.

Congratulations CUAUV for scoring second place in the 14th annual AUVSI RoboSub Competition, and a big congratulations to our friends over at ETS SONIA for first! Hopefully everyone at home was able to watch the finals via the live webcast, but if not, you can see recap videos of the competition at robosub.org.

The final day of the competition was both exciting and nerve racking. Unfortunately, the weather was not on our side this day as the clouds never fully cleared until after our final run. The day began with a practice run in the competition side of the pool, as the practice side was being used for the Second Chance part of the competition. The practice run was very successful, as we were able to quickly complete almost all of the course. Towards the end of the run as we were attempting the hydrophones task, we encountered the same error which stopped us dead in our tracks during last year's finals. Thankfully, our software team was able to come up with a quick solution to the problem so that if it happened during our final run we could still continue.

The final run started well with us navigating through the gate and hit the green buoy. Unfortunately, we appeared to be hung up on the yellow one. After asking the divers to bring the sub back and changing the code slightly, we started the run again. This time, though, there were two instances of the mission running which again caused problems. The sub was brought back in and the code was changed once more. On its final run, the sub went through the gate, rammed the green buoy twice (this was done on purpose, as we believed the first error may have been a vision problem caused by the cloudy day), and overall navigated several pipes. At this point, the sub timed out on the mission and went directly to the hydrophones task which it completed perfectly with only 10 seconds to spare! After the run, we were able to determine that the cause of the first error was due to a race condition error in our mission code.

Although things did not go exactly the way we wanted them to in our final run, we are still extremely proud of how Drekar performed at RoboSub this year and very happy to take home second place. Again, congratulations to everyone at CUAUV, ETS SONIA, and all other teams at the competition this year! We're looking forward to next year's competition and can't wait to see all of you again!

Congratulations CUAUV on first place in the qualifying rounds! After a very competitive round of qualifiers, we finished on top with the most points. Joining us in the finals tomorrow will be University of Florida, ETS, University of Rhode Island, University of Maryland, U.S. Naval Academy, NC State, and Reykjavik University. For those unable to attend the competition this year, you can watch tomorrow's finals down below in an embedded video or at robosub.org! There will be a live webcast starting at 1:00pm PST/ 4:00pm EST. Our run will start around 1:25, so be sure to tune in!

**Note: The embedded video has been removed from this page as it played as soon as the page loaded. You can find recaps and other videos at robosub.org.

As for today's recap, the day started off with a practice run in the morning. We mainly worked on bins trying to clean up the vision code from the last qualifying run. After that, we had a few hours off before our second qualifier. Similar to the last qualifier, this one was also very successful. We had a few more problems when it came to the hearts and bins today. We were able to successfully fire through the large heart on the first side, but got hung up for a few minutes on the other side. After firing a shot and missing, we moved on to the bins were we dropped one marker and timed out of the mission before the second could be dropped. The first landed again on the white fringe, similar to yesterday. We proceeded to the pinger and finished the final task with 30 seconds to spare. We believe the hearts problem was caused by an error in the code where the sub would not properly back up if it couldn't find the target, and the problem with the bins was caused by pitch control as well as not compiling code. We had a one more practice run shortly after, and the day ended by watching the other great teams in the competition and ironing out a few of the code problems.

Check back tomorrow to see how we did after the finals and be sure to watch the webcast at 1:00pm PST!

A lot has happened during the last three days at TRANSDEC!

The first day started off early with a practice run at 7:30am, during which we were successfully able to prequalify by driving through the gate. After the gate, the rest of the run was mainly spent taking images of all of the tasks for vision testing later. Our second run was fairly similar, but we also tested control and several mission elements including path, the L, and hearts. Our third and final run of the day was mainly spent doing vision testing on the buoys as they proved to be rather difficult outside in the TRANSDEC water. It was here that we found out the sun in the afternoon looks very similar to a buoy and Drekar is great at tracking it.

Like the first day, we were able to get three runs in on the second day. This day saw much of the same work as the first (vision testing, mission elements, and control testing), but saw much more of an emphasis on mission elements. We ran into a problem with the hydrophones due to having pingers on both the practice and competition sides, but we were able to solve this after spending some time in the center of TRANSDEC (aka the "Love Shack"). In the end, we were able to get the hydrophones accurate to less than 1 degree. The day ended with a successful autonomous mission run with the exception of the bins and hearts, as the mission timed out.

Today (the third day) was the first day of qualifiers! Our qualifying time, determined the day before, was at 1:15. We were able to squeeze in a short 15 minute practice run in the morning in which we further fine tuned our vision code. The practice run was followed by static judging around 9:30, and another practice run around 11:30. Due to a delay in the competition, our first qualifying round began at 1:30. We had a fantastic run! We attempted every task, and successfully completed almost all of them. We missed one shot through the hearts (although we did make it through the small heart!) and landed on the white fringes of the bins on both attempts. The run really showed off just what Drekar can do, and we look forward to attempting it again tomorrow at our next qualifier!

Also, in searching for competition pictures and news articles online, we happened across this blast from the past on CNET: an article featuring our old vehicle Proteus at TRANSDEC! The coverage on CUAUV starts on slide 6!

http://news.cnet.com/2300-11394_3-6196395-4.html