Raphael takes 4th place in the 2nd annual Autonomous Robot Speedway Competition
Video of Raphael at the competition:
After working through the night into Saturday morning, the R@M: Autonomous Ground Vehicle team was able to program Raphael to successfully navigate the competition course in the 4th fastest time. Raphael lost to the 3rd place team by a mere 4/10 of a second, taking 4th place out of 12 robots, after being penalized 9 seconds for hitting 3 cones.
Mechanical work on Raphael was completed the Thursday before last. A new rear servo was installed to allow the robot to steer both the front and rear wheels. Our turning radius before adding the rear servo was dismal.
Raphael uses Player/Stage to interface with various systems. On Tuesday we finished writing a fiducial driver to filter laser data and only pass cone points through to the deeper control systems. Unfortunately, the k-means clustering that we based the driver on didn't detect cones very well and was computationally expensive. We didn't realize this until Friday, after spending a majority of Thursday testing the k-means driver and a couple other basic obstacle avoiding algorithms.
On Friday we implemented a new cone filtering solution. This time everything worked extremely well, the driver was able to detect every cone with a fairly low amount of noise. However, after testing well into Friday night, we realized that we needed a new control algorithm. Even with near perfect cone detection, Raphael could barely get past a handful of cones.
(...)
Introductory meeting, First Look Fair, Eng. Community Picnic
R@M spent the last two days spreading information about the club and recruiting. We set up a pair of tables at the First Look Fair on McKeldin? Mall both days, with Tortuga on display and Donatello driving around (and hauling people on the grass). After the rain essentially ended the FLF Thursday (but failed to do any damage to the 'bots), we moved the whole display to the field outside Martin Hall for the Engineering Community Picnic, where we found another lot of interested people. (We had Donatello carry Tortuga the whole way, naturally.) The day's third R@M event was the opening new-member meeting, where several leaders introduced potential recruits to the robots' workings.
We will be holding another introductory meeting on Tuesday, 9/22, at 6:30 in the Neutral Buoyancy Research Facility. If you're interested in joining the club or just want to see what we're doing, be there!
Also, we have a new announcements mailing list. If you were on the Yahoo! group, you've been subscribed to this one. Otherwise, you should sign up to keep track of changes in meeting schedules and other big news.
Robotics @ Maryland attends AUVSI Conference and Exhibition
Robotics @ Maryland (R@M) will be at AUVSI Unmanned Systems North America 2009 from August 11th to 13th. We will have a table in the exhibitors area next to the other AUVSI AUV Competition competitors. We will have Tortuga III and demonstrations of our software on display. This is the largest unmanned systems conference and exhibition in the world, and R@M looks forward to seeing what it has to offer and showing off our hard work.
Tortuga III Takes 8th Place in 12th AUVSI AUV Competition
Note: This post is for Tuesday August 4th, but its posted late due to connectivity issues
Tortuga III took 8th place in the 12th annual AUVSI AUV competition. Despite not having as good a showing as we hoped we are still proud of our efforts. We accomplished much at the competition:
Took 2nd place in the qualifying rounds
4th highest scoring round of quals & finals (only the winning team did better)
The only team besides the winner to complete the bombing run objective at Transdec
The only team besides the winner to complete the machine gun nest objective at Transdec
We did all this despite arriving with a mechanical and electrical system which had only 2 hours of water time, non-functioning torpedo launchers and motors which shut down randomly after a few minutes. In addition to fixing those issues, we overcame blowing our stack of electronics the day before the competitions first practice day.
The final run itself has an interesting story. Our final round was after Cornell, and because they failed to do the bins perfectly we could still win if we could do the bins, machine gun nest objectives, and briefcase objective perfectly. We had carefully programmed in timeouts so that if we failed to complete the earlier objectives Tortuga III would move onto the briefcase objective and secure second place. The problem was we went with an untested mission, and Tortuga III ended up against one wall when it timed out. It turns out that the sonar system could not pick up the pinger against the wall. Due to the length of the earlier timeouts, we did not have enough time to recall the robot and go for the briefcase only mission we did for qualifying. This left us with just the gate, pipe and flare on the score board.
We would like to again thank all of our sponsors, without whom we would not have there opportunity to compete in this competition.
Note: This post is for Tuesday August 4th, but its posted late due to connectivity issues
First we would like to congratulate the Cornell AUV Team for their win of the 12th annual AUVSI AUV Competition. For the past several years they brought a vehicle with excellent hardware and electronics, and this year they combine it with better software and lots of testing. They attempted every objective of the course and did all but one perfectly. We would also like to congratulate the 2nd place team from the University of Victoria and 3rd place team the University of Rhode Island. We look forward to competing against Cornell and the other competitors next year.
We would also like to give special thanks to the ETS SONIA team. For the third year in a row, they have proven the most congenial and sportsmanlike competitor. This year they let us use there miniature bins to let us test our bombing run AI the night before the final competition round.
Finally we would like to thank all of our sponsors, without whom we would not have there opportunity to compete in this competition.
Note: This post is for Saturday August 1st, but its posted late due to connectivity issues
After a scratch yesterday Tortuga III moved into 2nd place in the qualifying round by autonomously grabbing the briefcase and surfacing in the octagon. Just like the day before we had an 8:20AM practice with a 9AM qualifying run. During the practice we again hit a last minute issue. This time we discovered our sonar system was failing to work. Utilizing our remote oscilloscope program we discovered one of the hydrophones cables came disconnected from the board inside the vehicle. With only 15 minutes to our qualifying run, we opened Tortuga III up and reconnected the hydrophone.
During the round Tortuga III moved through the gate, followed the first pipeline, missed the red light then used the sonar to grab the briefcase and surface. The vision system confused the red light with the wall but it successfully timed out on the task and moved onto the sonar. After the round we did another, non-scoring autonomous round to test our bin dropping, but that was unsuccessful.
Note: This post is for Friday July 31th, but its posted late due to connectivity issues
Today we had to scratch our first qualifying round. We had signed up for the 9AM slot with a 8:20AM practice slot right before, planning to test everything before doing a conservative qualifying round. In reality a latent bug that had been unknowingly plaguing us for the first few days surfaced. During the practice run we initially had a thruster switched which led to strange control issues, and even after fixing that the vehicle was basically uncontrollable. With too little time to fix it we had to scratch the qualifying run and hope to get things working in time for the next day.
It took several hours of testing before we found the issue. It turned out to be a latency timer that controls communication between the main computer and the embedded electronics was an order of magnitude too large. It had not been set during the haste to get Tortuga III's new computer up and running. This led to large delays in vehicle control signals reaching the thrusters. After fixing the latency issue the vehicle's earlier control issues went away as well.
Missing our first qualifying run was a big let down but we believe we will be ready to move up near the top of the heap tomorrow.
Note: This post is for Thursday July 30th, but its posted late due to connectivity issues
The second practice day was more mixed then the first. After some initial success on the bins we have been having vehicle control and vision issues while maneuvering around them. We also had a bug with the control interface which blocked the ability to do the sonar objective. In addition to that we had issues picking up the pipes, barbed wire and buoy with the vision system. The one positive is that we successfully lined up to fire at the machine gun nest several times. We have yet to work out the firmware and wiring bugs with our launchers but they should be operational soon.
Note: This post is for Wednesday July 29th, but its posted late due to connectivity issues
Our first practice went pretty well with Tortuga III completing several of the competition objectives. Our two first runs we spent gathering data on the practice and competition sides of the course. After gathering the data we calibrated our vision system and updated our AI settings in preparation for our first semi-autonomous run. Our final run of the day was on the competition side of Transdec and we drove Tortuga III between the objectives activating the AI for each. We successfully hit the buoy, dropped the markers in both bins, and then grabbed the briefcase and surfaced.
Note: This post is for Tuesday July 28th, but its posted late due to connectivity issues
Despite a major electrical failure today, we have Tortuga III ready for its first day of practice. We will be joining Cornell, ETS, and Florida in the 7:00 AM arrival slot.
The failure story itself is interesting. The failure was caused when we swapped in a spare board which had a PCB manufacturing defect. The spare board shorted 26V to the 5V bus and blew several boards in our electronics stack. We swapped in a spare board while attempting to fix the firmware; Tortuga III only got in the water hours before it was shipped. Our electronics team worked throughout the day and early morning to get our spare electronics stack running. After lots of hard work we managed to get Tortuga III back up and running with a fully working motor board.
Note: This post is for Monday July 27th, but its posted late due to connectivity issues
We have gotten ourselves fully settled in at the Kona Kai resort in San Diego and our robot has arrived. After dealing with unloading a 500lb crate without a loading dock or lift we got to work. We unpacked Tortuga III and its gear, then got to work on assembly. It took a little while to get everything assembled and iron out some firmware bugs but we eventually got it in the water. All basics system check out and we are going to fix the rest of the issues on Tuesday before the first practice day of the competition.
Top: Unloading the Tortuga III from the crate, Bottom: Testing Tortuga III in the hotel pool.
Tortuga III Shipped, AUV Team on the Way To San Diego
Note: This post is for Sunday July 26th, but its posted late due to connectivity issues
Robotics @ Maryland is on the way to San Diego following Tortuga III and its gear which were shipped Thursday evening. We tested Tortuga III's hardware with our new AI, which we spent all year perfecting on the venerable Tortuga II. We still have some minor issues to address once we reach San Diego, but we are all excited to compete at this years competition and we feel we are better prepared then ever before.
Part of the team waiting for rental cards outside the airport in San Diego.
We have completed an in depth journal paper which describes the design and engineering decisions behind our latest AUV: Tortuga III. You can Download the PDF to take a look. We have also updated the project section our website with additional content about Tortuga III.
During the actual testing Tortuga successfully followed pipes, hit the flare, went under the barbed wire, lined up for the machine gun nest, and surfaced with the brief case in the octagon. The bombing run was not completed due to issues with floor tile patterns, but we have completed it numerous times at our normal testing facility, the NBRF. Below are some pictures taken of Tortuga during the ERC test runs:
A compilation of shots showing Tortuga in action, going clockwise from the upper left: Hitting the flare, lining up for the barbed wire, surfacing with the brief case, lining up for the machine gun nest.
Engineers working on those projects reviewed design presentations given by both our AUV and AGV groups. They provided lots of useful feedback to both teams that will be folded into some of this years work, but it will mostly be applied to next years designs. After the design reviews GDRS was nice enough to provide us lunch and give us a tour of there testing and integration center, prototype USV vehicle, ANS development building, and finally a demo of there ANS system in action.
