Here’s some more photos from our testing on Saturday for your viewing pleasure.
End-cap trepidation
Fun with Barreleye!
"Coolest AUV evar!"
hmm...
Lunch break!
Following it around... in circles...
While we all (and everyone else) watch.
Professor Ben explains timing to us.
Lounging around the AUV
Working as a team!
That’s all for now, I can’t wait until next week!
We finally headed over to the campus pool today for our first day of live underwater testing! We have been working hard on the code to control Barreleye, and we finally had it to a point at which we could test it in a realistic environment. We all gathered in the lab and loaded up Ben’s station wagon with the AUV and associated equipment (BTW, Smart Bungees are really useful for tying stuff down as well as tethering), then made the short drive over to the gym.
Ready to go!
As we unloaded Barreleye and wheeled her down to the pool, there was a group of 30-35 people watching us through the pool room windows! We had told a few tours about our plans, but we weren’t expecting such a crowd! Alas, they turned out to be an orientation tour and left three minutes after we got there, but about four stuck around to watch.
The previous pressure test hadn’t gone too well and we weren’t really sure that the hull wouldn’t leak, so the first thing we did was send the hull and frame to the bottom by themselves to double-check our latest round of spreading silicone all over everything. We left it fourteen feet under for a half-hour, and were relieved to see only three drops of water, which could even have been condensation. With a little nervousness we lined the bottom of the hull with a thin absorbent sheet and loaded in the E-rack and batteries. We hooked everything up and sealed it with the end cap, forgetting to turn on our servos in the rush. We decided to skip all of the planned dry-testing and threw her straight into the water, very gently.
Just throw 'er in!
At first we were worried that the wi-fi connection wouldn’t work automatically and we would have to re-configure something, but the last-minute design change we had made the day before worked flawlessly. For the first 20 minutes, anyway. I eagerly launched the RC software I had written the few days before, and yelled at Joe, who was in the water next to the AUV: “We’re going forward!”. Barreleye sprung to life and smoothly powered her way forward.
Forward motion!
“Try going down!” was Joe’s reply, and I traded the forward thrusters for the vertical one pulling down. I waited a few seconds after maxing out the control, and yet again heard Joe yell “Try going down.” “It is!”, I countered, and we both saw the large ripples on the surface created by the vertical thruster trying its hardest to make the AUV descend, and obviously displacing a lot of water while it was at it. Yet as hard as it tried, she was still sitting on the surface. I shut it off, and Joe and Stephen tried adding a little bit of lead ballast to make it easier. We very quickly noticed (well, they did anyway), that we were no longer positively buyant! She was sinking!
Thar she goes!
An 0.8lb weight had succeded where a $900 thruster had failed. Something is wrong here! We knew that we have to end up positively boyant (i.e. not sink by default), so they removed the weight. Instead of going down, we tried driving around on the surface for a while. Going in a particular direction proved to be difficult, but she had no trouble blasting from one end of the pool to the other. Slightly encouraged, I tried combining forward and downward motion. Slowly she eased below the surface, and continued to drop as I kept her going forward. I stopped, and slowly she surfaced (say that five times fast). I quickly discovered that the best way to go under is to try to go backwards and down, because the heavy end cap tends to angle the hull back, and is thus tends to hydrodynamically drive the whole thing down when driven in reverse. Once we figured out how to get her under, the first thing we tried to do was make it through the validation gate.
I started Barreleye about 20ft from the gate, and eased her into a forward dive. Course correction proved to be difficult, because the controls and the AUV’s momentum worked against each other. After over-correcting enough to end up going in exactly the wrong direction, I slowly turned Barreleye around and tried again. This time I was able to keep her in a straight line until I got to the gate, and about a foot underwater when I got there. By that time, however, Joe was directly in the path and I cut the thrusters to avoid a collision. This proved to be a mistake, because we surfaced directly under the center joint of the validation gate, which came apart and promptly sank.
Get out of the way! Oh wait...
Not discouraged by our path of destruction, we decided to aim for the “flare” instead. The red buoy is moored with a 18lb weight and floats about four feet under the surface. At this point I decided that it would be a good idea to get a video stream from the camera (the “eye” part of “Barreleye”), and managed to get a dozen frames or so before crashing my remote connection. After a quick re-connect I tried to go at it by sight: a difficult task because of the surface distortion (note to self: get a waterproof laptop so I can control it underwater). After several failed attempts, I finally managed to slowly bump into the flare, and even wrapped the safety rope around it. We all celebrated and grabbed some lunch (which my Dad very generously provided), and then wanted to try out some of our PID controllers.
PID, for those not in the know, stands for “Proportional, Integral, and Derivative”, which are the three terms in an equation that is supposed to make our software know how to make our AUV go in a straight line. Stephen had created some code using Simulink and we had downloaded it to the AUV computer the night before. I launched the program, and we yelled at Joe to hit the “On” button (I had decided that it was a good idea for the person in the water to have control over the software. Unfortunately, it left me powerless to start or stop it). After Joe found and pressed the button, Barreleye immediately started spinning in circles. I was monitoring the compass reading and I could clearly see the number falling quickly from 360 to 0 and starting over again. “It’s going in circles!”, I yelled to Joe. “I can see that!” he yelled back. “Where’s Stephen?” I yelled at no one in particular. “Eating lunch.” said Ben as he sat down next to me with his sandwich. “Stephen!” I yelled at Stephen, who was enjoying his sandwich on the steps in the corner of the (very large) room. “Your code isn’t working!” No response. After several attempts and a few amused looks from Joe, we got Stephen to join us under the diving platform and discuss his algorithm. We tweaked some gains for a few minutes and got Barreleye to spin in circles slower, and even to oscillate in wide arcs around the desired heading. All of us eventually figured out that she was overshooting the mark, and was trying to go all the way around again each time. We knew that the way to fix that was to adjust the “Derivative” part of the PID controller, so we tried to adjust that gain value. After a few minutes of searching, Stephen and I realized that we must have left that part out of the code somehow! All we really had was a PI controller, and not a very good one at that.
This story is so long it needed an intermission. Here you go.
Another half hour of spinning in circles and yelling “hit the Go button!” and “Stop it!” at whoever was in the water at the time, and our time was up. We only had 20 minutes left before we had to vacate the building, and right on cue, the main computer batteries died. We hurried to get her out of the water, because we knew that the system could quickly drain the batteries past the point of no return, and our off switch unfortunately doesn’t work after the voltage drops too low. I popped off the end-cap, we quickly removed the batteries, and we packed everything up for the return trip to the lab.
After a hard afternoon’s work, we now know exactly how she handles in the water, and more importantly, that we can’t drive in a straight line, or even go underwater without some hydrodynamic trickery. We have a long way to go, and not a lot of time to get there. Unlike last week, however, we know exactly where we are, and what we have to do to get to where we’re going. Wish us well!
After careful deliberation, AUVT has finally selected a name for our new AUV: “Barreleye”. The name comes from a very strange fish of the same name, a fish that we feel is very similar to our AUV. The barreleye has a clear skull, with swivelling eyes inside it’s head. Although strange for a fish, it is a similar design to our forward camera, which is mounted inside a clear dome on a servo gimbal so that it can move to see at different angles. You can learn more about the barreleye here. Just to clarify something: those two dark bits at the front aren’t it’s eyes, those are noses! To clarify:
Mmhmm
Last night Joe and I (Micah) finally sealed up the hull with all of the wires connected and the e-rack fully operational. We took it out to our small test tank, dunked it, and turned it on. For the first time, the thrusters came to life under internal computer control and propelled the AUV forward! For about two feet, then we hit the end of the bathtub-sized tank. Excited, we drove it around the tub for 10 minutes to see how she handles.
Two things quickly became obvious: our vertical thruster isn’t as powerful as we thought, and our batteries die a painful death if you don’t get them out of the hull in time. The computer attempted to turn itself back on several dozen times after it first died, and apparently there wasn’t enough voltage for the killswitch to work. One of our battery packs seems to be permenantly damaged, but its ok, you live and learn. The hull didn’t leak, and we discovered that having a wi-fi ap means that we don’t have to run Ethernet cables out the lab window anymore. Oh, and being able to RC the AUV via iPod touch/iPhone is pretty cool.
We hope to do a lot more testing this Saturday, so stay tuned for more updates and some photos too!
Today we finally put the finishing touches on the e-rack and slid it into the hull for the first time se we started rebuilding it. Seeing the entire AUV in one piece is inspiring and motivating. It finally looks more like a finished product! Soon we will be able to test it underwater with the computer actually controlling everything! Here’s a little tour:
Glowin' Good
The E-Rack fits like a finger in a glove
The "eye" looking down at us.
The end-cap with lots of expensive brad-harrison connectors (but they're cheaper than seacon!).
This is where the end-cap was in the previous photo
This is what it looks like when we take the end-cap off.
This is all the stuff we have to fit in there.
The batteries just fit right under the e-rack, surrounding a mess of wires.
That mess of wires from above.
And right next to it, the computer.
The mandatory killswitch, which is used to turn off the thrusters should we want to stop.
Let’s go ahead and hit that button. See you next time!
Today we are testing our main hull in deep water for the first time. We are in the War Memorial Gym pool at VT, and the hull is already leaking. We have it in the 14′ deep section to simulate the depth of the TRANSDEC facility as closely as we can. We had to rig up a harness using bungee cords to lift and lower the AUV, and we are using simple weights tied to it to sink it to the bottom.
To get the AUV to the pool we loaded it onto our old wagon, since it was the best thing we had at the time.
Walking over the the pool on campus.
Once we got there we had a half hour to see if it would leak. We carefully dropped it in and lowered it to the bottom. Then we got a little bored.
We sunk the AUV and waited for it to leak.
Joe went down to check for water in the hull.
Unfortunately, there was water everywhere.
That's water where it shouldn't be.
The walk back to the lab wasn’t quite as hopeful.
A bunch of nerds crossing campus with $3000 worth of AUV in a little red wagon.
But we got straight to plugging the leak.
Lots of silicone in a confined space makes for a strong smell.
Overall the trip was a success, although we can’t help but wish we could have started testing months (or years) ago.
Today we finally wired up everything on the electronics rack (E-rack) to see how it fit together. The first thought to enter my mind after we got it together was “holy cow that’s a lot of wires”. We had decided to trade surface-mount soldering for a whole lot of crimping, and this is the result:
If there were any more wires we wouldn't be able to figure out where they go.
Some of the wires are a lot longer than they will be (that beige one that loops a few times and the red SATA cable), and everything will be a lot neater once it’s all cable-tied. The E-rack is actually sitting on the batteries, which are a bit easier to see in this next photo:
BTW, the white thing on the right is the camera with the dust-cover-made-out-of-a-napkin-and-some-electrical-tape on it.
Hopefully the whole thing will fit in the hull without too much trouble, and we will be able to get to the batteries to charge them without removing the entire e-rack. On the e-rack you can see our computer on the right and the electronics stack on the left. The end-cap will be to the left, with the depth sensor taking up the empty space to the left of the e-stack. There are four 4600mAh Li-Ion batteries under the rack, as well as four AA’s for the servos (on the left). Two of the big batteries power the computer and the e-stack, and the other two power the thrusters. Some of the wires are easy to pick out in these images, for example, the white molex connector up front powers the hard drive and the Micro-strain IMU from the computer power supply, and the red/black connector on the bottom left is the servo batteries. You may be able to make out the red Ultra-plug floating abover everything; that’s the main thruster power going from the power switch board to the motor controllers. The gray cable disappearing out the bottom left corner is ethernet, and the big beige loop is a USB cable leading to the Arduino.
Organizing and labelling all of these wires will be a lot of fun for tomorrow, and hopefully we won’t show up at the competition with a nighmare rats nest of wires all over everything.
Yeah...
Some of you reading this know about AUVT. More of you think you know about AUVT. Still more have never heard of AUVT. If you knew the AUVT from before the fall of ‘08, you do not know AUVT. We’re all new, and rebuilding with a small but agile team; focused on organizing the spec-ops of AUV builders.
We have spent the last year as a freshman team. Not a single member of the six-man team has any experience with AUVs prior to 2008. Even with a much larger (15+) team, AUVT did not compete in 2008, and the 2007 entry just barely made it into the water. That’s where we started, and we’ve been learning everything from scratch. The previous iteration of the team left us with a pile of parts, some of which resembled AUV parts. We started by cleaning out the work area and sorting out the boxes and shelves of mysterious parts to see what we had. Exploring and finding all of the parts took almost exactly a year, with every week highlighted by discovering something that we wished we’d had a month ago on a forgotten shelf. We knew we had this thing called an “E-Rack” with a computer on it, and the hull with its “End Cap” seemed pretty straightforward, but that was all the knowledge we started with. The learning curve was long, but slowly and surely we educated ourselves in the ways of AUV, and we became prepared to tackle the task before us.
One of the first issues we faced with a new team was a lack of leadership. Everyone who knew how to lead the team had graduated the year before, and only one member remained who had witnessed some of the original design process. Fall of 2008 and Spring of 2009 went by with only minimal progress, and then both the experienced team member and the new team “captain” dissapeared. We knew the team was fragmented, but we tried to find the best ways we could to work around it. Whiteboards and wikis offered a solution, but resulted in little practical use. We are still a strange variant of “team”, with some of the members rarely even seeing each other, but it seems to work.
It was not long into the summer before we realized two things. One was that we don’t have time to write code from scratch, and the other is that the “Signal Processing” circuit board that we inherited from the previous version of the team was lacking in useability. It was decided to switch to code-generating software to help with the programming, and to ditch the all-in-one custom circuit board for off-the-shelf parts. Initial tests with both aspects were promising, but integration is proving to be tricky. Every day presents a new challenge from the AUV that has yet to be named, and every day we conquer a new hill on the oh-so-mountainous road to San Diego. We will rise to the occasion, solve every problem, and be ready when the day comes to show the world what we can do. We are AUVT, and we will have a completed AUV by July 25, or we will say a collective “dang it” and just use the AUV at next years competition and spend an entire year debugging.
The AUVT website is currently under construction. Please forgive any typos or broken links as we prepare for the final stretch of frantic engineering before we head out for the competition in just a few weeks.
