attribue(__interrupt__) and attribute(__noinline__) may appear similar but the former will totally hose your stack pointer.  Like to the point where you're executing out of SRAM. Bad news on a harvard machine.  Only took a day to find that typo... yeesh.

We're currently gearing up for our dunk.  We'll have a 15minute run at 11AM PST, and I'll be reporting back immediately after.  We're hoping to be able to take first in the losers bracket as it stands now.  Gate, Bouys, Hedge and a few Paths are what we're hoping for.  Wish us luck!

Time to have fun, but I'll give you the results:

Our run didn't go so hot.  We didn't make it under the gate, I think we still maintain our place from yesterday, more info coming tomorrow probably.

After the vision guys performed some tweaking, we hit the water and prepared to see what all SeaWolf could do.  After a few runs of being sent towards various sides of the gate, the diver finally gave us a good start towards the center. We veered a little to the left while we were covering the distance from the dock to the gate, which required a little gymnastics to align with path, but we managed it.  Seawolf headed for the bouy's, bound directly for the red one.  About 5 feet from the bouy, Seawolf went into a sweeping partern... sadly, something like as if it was looking for the next path marker.  She slowly inched her way forward, sweeping from left to right. We were down to 2 minutes on the clock, no time to reset and were just praying for the bouy bounce.  

With a minute left we saw the divers sign that we had hit red, but there was a little doubt which Seawolf saw fit to correct.  After a few seconds of intermittent contact, Seawolf basically sat down on the bouy.  Clearing up any remaining questions of her contact. 30 seconds left.... time.  Seawolf III completed more of the course than she ever had before. That's enough to keep us pleased and as of the standings last night, it was enough to put us in the top 10.

We also opted to try to grab a few extra points yesterday by implementing JAUS, a DoD driven project that acts as a protocol for communicating with unmanned vehicles.  Chris set to work shortly after we were out of the pool and managed to have us up and running to nearly complete state within 2 hours.  The University of Florida had developed an opensource solution to JAUS called JAUS++ and with a little finagaling Chris managed to get it to compile against our own LibSeawolf making implementing the protocol near painless.  In no time we were able to send data and let a JAUS sever connect with us properly, change states correctly, and most importantly, pass data out to the server.  There was still one little bug preventing us from completing the Interop Challenge completely, but we'll have another shot between 8-10 on Sunday.

While the 9th place announcement was reason enough for excitement, Dave Novak had one last surprise for us.  Sunday would play host not only to the finals which the top 7 teams would be competing in, but also to a wildcard round in the morning consisting on a single 15 minute run.  There's a mystery check waiting for the team who is able to do the best, and with only one teams ranked above us by a thin margin so far, we're hoping for the best.

After a long night for the club 5:30 came early, a necessity after a long nights work if we were going to get everything packed up and make it to TRANSDEC early enough to get a nice early practice run in.  Thankfully everyone is pretty on task and we were making tracks for the pool by 7.  When we arrived we put in for the 8:30 pool test slot and Chris headed right out to start taking some accoustics measurements after some major retooling of how the correlation block was being executed on the blackfin.  Accoustics seems to be growing ever closer to that magical point of actually being able to drive Seawolf!  We're hearing pings loud and clear and are now just working on effectively measuring the delays between when the ping reaches Seawolfs four ears.  These delays will enable us to locate the pinger and hopefully track it down!

The 8:30 pool test wound up being more like a 9:15 pool test due to some delays at the pool and kind of dampened our hopes.  It turns out gatorade bottles don't exactly hold their shape when you fill them with air and submerge them 8 feet under water, meaning our ballasts stopped giving us much flotation.  Seawolfs a bit on the heavy side and losing boyancy was wreaking serious havok on our ability to accurately find and maintain a depth.  While this wasn't much of a problem for the gate mission it proved disastrous for trying to bump the bouy.  Our emergency breach code kept trying attack Micheal 
(our diver for the time) after seawolf would start sinking at depth due to compression and our control code not ramping up quick enough to compensate.  Thanks to Maryland we've now replaced those pesky bottles with some pink foam that seems to be working well (if maybe not a little too well as we've now discovered).

While inspecting Seawolf after the test run, we noticed that one of our Fischer connectors (our class A waterproof connectors) was beginning to corrode.. readers from last year, this might sound a little familiar, and yes, Seawolf was once again performing chemistry experiments.  A solid round of testing by Rabbit found that the aluminum frame that holds Seawolf together was carrying a voltage! After a good deal of debugging we found that one of the connectors had a shield wire touching the outer casing.. we'd found our culprit, the source of our attempts to once again electroplate Seawolf.

After a bit more acoustics testing and some  buoyancy adjustments in the dolphin pool (small pools set up for teams to test in) 12:30 rolled around and it was time to see old Seawolf get into action.   Tetherless Kirk (resident Navy sailor dude for lack of a better description) lowered Seawolf into the water and after explaining our mission start/killswitch system to the diver, the swipe was made, the status light blinked and Seawolf dove.  It quickly became evident that we had overcompensated a bit in the boyancy department, even at when she was at depth you could still see the jets from the struggling thrusters breaching the surface. Due to a cross breeze there appeared to be a current in the Transdec pool, combined with the mistake of leaving her forward speed set a little low and the immense thrust the vertical thrusters were exhibiting Seawolf barely missed making it through the gate deciding to split the right side pole instead.  After two more attempts with various degrees of success and failure we yanked her our of the water, Michael plugged in the tether and corrected our speed issue.  We reset and tried again, this time succeeding in making the gate!  Sadly we never saw the path on the bottom of the pool and got a little hung up looking for it, forcing the restart once again.  We again managed to make the gate but path align failed again.

We were greeted to day two of the competition once again by the delightful wakeup call/advertisement hotline the Kona Kai has opted to employ.  I believe that they fail to realize that waking up to an advertisement is a sure fire way to leave a sour taste in my mouth for whatever it is they're trying to convince me to spend my money and  "extra" time on while we're here.

We once again made it to TRASNDEC ahead of our 7:15 unload time and quickly registered for our first pool test of the day.  That's about where things started falling apart.  Once we were in the water we started losing communication with just about everything on random occasion... Our little honey moon with the new and predictable Seawolf was coming to a close and we were starting to see problems of old start creeping out on us.  After a bit of debugging, it turns out, running thrusters and electronics on a single bank of batteries is either a project for grad students or worth avoiding by isolating your thrusters.

It turns out, big inductive loads, make equally big noise.  This meant that from time to time, the USB hubs on Seawolf were eating some crazy voltage spikes, resulting in the IMU and our cameras freaking out in a rather spectacular fashion.  Seawolf was regularly going blind deaf and dumb and worse than that, was having trouble putting it all back together once things had settled down.  Sadly this little problem has cost us our brand new power board... but a quick reworking of our old harness and a DC-DC converter has given us legs to stay in the competition!

While this hasn't completely fixed the issue completely, we're banking on a few zener diodes on the last few points noise could work its way back up the lines to the USB hub.

I wish there was more to tell of the flights, and there probably is, I however managed to catch up on some much needed sleep over the course of the two flights being rattled from unconsciousness only when the passengers next to me required my movement.  Pitfalls of the isle seat I suppose, but well worth it.

Once landed in San Diego we were glad to find SeaWolf, although throughly searched, was still in tact and all of our spray painted boxes had survived the hands of the "throwers" behind the scenes.  Kevin and Dave set off quickly to locate out 15 passenger van, which much to our surprise came back as a much more modest Kia Serento with seating for 7.  A quick game of Tetris and we hit the road and headed for point Loma with a minor distraction of food planned along the way.

We made it to our favorite resort, the lovely Kona Kai, around 2 and just in time for our orientation.  Orientation consisted of the typical, "You're on a navy base, in California, there are plenty of ways to get hurt, please avoid them and don't tell all of our secrets" followed by an unusually short round of Q and A time with Dave as engineers from competing schools grilled him on the specifics of how the course was going to be set up, missions were to be run, and a fair share of questions that didn't need to be asked.  We stayed quiet for the most part, I believe our stance on such issues as course lay out would be more aptly described as reactionary than proactive.

Following orientation, a quick check-in and we were on to assembling our craft.  Thankfully, and much to our amazement, even with the reduced load, there were no major missing components, nothing severely damaged, and other than a utility knife (which.. in reality we had, it was just in hiding) no missing tools!  With assembly and some work on our new cable hiding skirts, Seawolf was ready for action and looking the best she has since we've moved on from SeaWolf II.

The remainder of the evening was spent tooling around for a few parts to complete our new powerboard, finding some food for dinner, and finally locating some much needed sleep with wake up calls coming at 6AM.

This morning certainly came quickly, but thanks to that wonderful 3 hour west coast lag we were all up and moving fairly quickly. After removing a few thrusters from Seawolf to get her to fit in the trunk, we did the impossible (note: this may mean the end of the world is in fact upon us) and showed up to Transdec 5 minutes ahead of our 7:15 unload time.

A quick unload and some short tuning had us up and running well enough to hit the water for the first time this year.  Our first brush with fabled Dave Novak gave us the 8:15am testing slot in the practice half of the pool. Following Cornell we were the 3rd team in the water.  While it wasn't all that visually impressive of a run due to a small bug, we couldn't be happier with the way SeaWolf was performing.  Long gone are the days of futile attempts at making her go in a straight line! (now replaced with less futile attempts to make her dive... but at least it was much more resolvable)  A short retooling period following the first test led up to our second.  Again, while not that impressive to those watching, we collected good data on what the objects we're trying to see actually "look" like to SeaWolf.  After having some navy divers drug her around for a bit we pulled her out and have set out on giving her her last major upgrade of the season, a brand new power board designed by Kevin Wolf and Byron Beddingfield.  A few short reworks of some bad parts and a few load tests later we're all ready to see the days of the famed rats nest we call our wiring harness come to a close and the introduction of a sleek and clean power solution make that last step toward a clean and managed case

Once again AUVSI has managed to fail to provide reliable internet at Transdec, and while not particularly surprising, (I see how providing internet to an outdoor secured navy testing pool could provide some challenges) is atrociously annoying at a robotics competition.  This blog post has been the result of a quick trip back to the Kona Kai for some parts orders and some the need for some datasheets.

I'm sure we'll be in the water at least one more time over the afternoon and hopefully we'll start collecting data for our accoustics system!  I'm planning to update this once a day, but you may also want to check out robosub.org for more up to date information and video recaps starting on Friday!

Thanks for reading,

Matthias Welsh

So SeaWolf... It's that time of the year again when we can feel San Diego beginning to lurk around the corner and everyone starts ramping up their efforts.  We currently have 3 major projects under development. We're testing vision pretty handily, all the time finding out that our PIDs and control systems weren't quite tuned or working as we had intended but luckily we're discovering this now and not at Transdec this year.  We'll take it.   A couple of minor things we're currently looking to rework vision wise

1. Wide angle lense. This is hurting us.. vision keeps losing site of our target far before we get there leaving us in a dead reckoning state.  If I remember correctly.. our test gate is a little smaller than the real gate, meaning this problem will only compound during competition.
2. Better cameras.  Our current cameras are webcams in fancy waterproof enclosures.  While this is convenient in some aspects, in others it's a bit of an annoyance.  A complete lack of control over auto-gain, auto-whitebalance and auto contrast can wreak havok on the image we're getting back.  This boils down to the fact that off the shelf webcams weren't intended to deal with underwater lighting conditions, surprise surprise.  We're looking for cameras currently that will allow for more fine grained control under linux, if you have any reccomendations, please shoot me an email (contact stuff in the side bar).

One of our members, Kevin Wolf is hard at work on a new power-board for SeaWolf.  The current "power-squid-thing" is functional (loosely defined) but is easy to mis connect, looks like hell and generally needs to be cleaned up.  In addition to tidying things up quite a bit Kevin is also preparing out switch to some form of modern battery technology.  Currently we're in the process of investigating LiPoly as they seem to be the balance between energy density and explosiveness (li-ion being equated to placing incendiary grenades in our case with the pins pulled and hoping for the best) especially due to the high-current pull of our thrusters.

Finally, I've currently taken on the objective of ripping out a new motorboard for Seawolf.  There's a few quirks in the current board that make it a bit difficult to deal with.  The design objectives for this project are to reduce the USB cable count (we were going to try to cut down on the number of controllers... but this works for now).  We're planning to achieve this through one of these guys, FT4232H by FTDI. This will allow us to communicate with 4 Arduinos.  3 will be used for motor control and a 4th will be broken out for use with sensors or anything else we decide we need (status lights etc).  The current plan for the breakout (both analog and digital) is to provide:

signal1    voltage    ground

signal2    voltage    ground

This allows for 3 pin headers to be slapped in instead of having to run individual wires all over the place.  A few fuses to protect our controllers and call it a day.

So lets start with the practice run, which was largely an impressive success, due to tether length issues we were only able to complete the gate mission but after looking back over the logs we were able to see that Seawolf indentified both the bouy and orange marker pointing to the bouy.  After seeing Seawolf do something totally intelligent, we headed back to our tent with high hopes and ready for our next chance at qualifying.  In large we decided to leave Seawolf as she was, fixing a few funky bugs that Chris found but nothing major.  When he time came we quickly headed to the pool, ready to make the most of our last chance.  Once she was on the crane we all crossed our fingers and hoped for the best as a culmination of ten very short months and four very long days was about to hit the water for the final run.  We instructed the diver, blue RFID card to blue tape to start and kill in the event of a restart, red magnet to red tape in the event of a system failure.  Once in the water, the diver swept blue, and Seawolf sprung to life, but with no forward motion, she aligned to the gate, sank to depth, but refused to go for it.  After sitting for a few seconds she began to drift backwards and attempted to pin our poor diver under the dock, and we signaled to kill. After being dragged out from beneath the dock we quicky grabbed a hold of the crane operator Kirk, and hoisted SeaWolf onto the dock.  Throwing our teather down to Brooks, Chris plugged in and started debugging.  After 10 long minutes of bug hunting he stumbled on a bug he had corrected early which was adding an offset to the yaw directional thrusters and pushing Seawolf forward!  After a quick fix we dropped her back in the water and hoped for the best, the driver once again swiped blue and once again she saw the gate, aligned herself, sunk to depth and with only a minute to spare took off toward the gate. B-lining directly through the gate she make a small course correction to align with the orange plank sending us to the bouy, and headed for it.  Sadly, we shortly detoured and missed our target and after some debugging work later we found that a single line of code had left us stuck in the "search for path" mission.  After all we had accomplished this year, that was the least of our worries, we have a working platform from which to build on last year, and we had sucessfully qualified this year, placing us somewhere in the middle of the pack for this years competiton.

More to come.. but I'm too tired right now.  Hope you've enjoyed following us and I look forward to blogging our design process and competition again in the following year!  Thanks to everyone who's shown an interest in SeaWolf and please continue to, we're only going up from here!

I'll start with yesterday thought, kind of a frustrating day for us.  Seawolf was having bugs creep up that were about as off the wall as they come.  During our qualification run Seawolf was having trouble doing anything but spinning in circles and trying to attack our poor diver.  It's quickly become apparent that when Seawolf goes into a spin the extruded aluminum coming off the rear of our craft acts as a rather scary weapon in addition to the handles they were intended.  The divers are nimble though and averted any injury but after trying to take him out a few times, we pulled our ornery bot out of the water and set to debugging.  While we're not 100% sure what was causing the problem it appears it was related to attempting to use our magnetic kill switch as a mission start.  Surprise, surprise, electronics + magnets = problems.

We quickly scheduled a practice run and sadly... we found similar results.  Seawolf was in a mood to fight and little else.  After a bit of messing around and some debugging at TRANSDEC we headed out in somewhat low spirits.  On our way out of Point Loma we happned to see another van, and it was on.  After the necessary egging on back and forth, we found that it was virginia tech, both in our 15 person rental vans.  We took off and they followed suit.  With engines roaring (a rather impressive sound, not that that translated to much in terms of speed) we barreled through San Diego.  We have the superior craft when it comes to vans though and we were eventually forced off the road as our destination approached, but there's nothing like a roaring engine (regardless of speed.. and with loaded vans there wasn't much of that) to introduce a little injection of some much needed fun into the evening. 

After grabbing a bite a Baha Fresh we headed back to the Kona Kai only to discover a full scale Luau taking place right outside our door. We took in some sights and sounds for a few minutes before packing up Seawolf and heading to the pool to correct our rather disabled craft.  After returning to our RFID card mission start system we found some improvement, at least now all of our motots were staying on, we still had some issues to debug with our yaw PID (control code).  Chris started correcting problems one by one and pretty soon, SeaWolf knew what straight was again.  With our problems solved we moved onto testing our vision bouy bumping code.

Because of the lighting in the pool, the orange bouys were showing up white, not something we were prepared to deal with, however due to Micheal and Jeffery's hard work on vision and interesintg "find the closest thing to what color we're looking for" code the red on my swim trunks quickly became a suitable target.  Torro, torro, torro!.  After some minor corrections we were quickly heading the right direction and with good time.

After all of Chris's contributions for the day he took the chance to grab some sleep while our vision guys were hard at work, but once they were done it was time to mix our "go straight" and "find the gate" code.  After waking Chris we attempted to re-test the "go straight" code with scary results, again seawolf was lost, and going in circles.  Frustrated, cold from being in the water and beat Kevin and I hopped into the hot tub to relax for a few minutes.  I was quickly requested back in the pool, and in no time, seawolf was once again on the right track.

So here we are, back at transdec, with our practice run coming in less than an hour.  We're currently mixing the vision and go straight code hoping for the best results.  Wish us luck, and we'll be back shortly!