Team Two Journal

Team Zombie Nazi Daemons with three arms and butcherknives for hands and machine guns coming out of their chests that they fire by going like this

January 28th, 2005

• True to form, the robot won't start from the orcpad. -Abraham

January 27th, 2005

• Back for one last hurrah. -Abraham
• Set screws coming loose like mad. Ready to punt robot out of window and go home. -Abraham
• I knew it would happen eventually. Our orcboard decided to release a little magic blue smoke during a routine power up.
• Setscrew can be one word. We had some successful goes at navigating and maybe going towards balls, but targeting goals still won't work properly, and balls may not even work. we are struggling with raising balls up ramp now. -jasonkb
• Currently waiting to recover from our meltdown earlier. Once that's done, we're going home. I really don't care anymore. -Abraham

January 26th, 2005

• Testing on-going. It's going to be a long night. -Abraham
• Linear Actuator burned out. Disaster. -Abraham
• Wall avoidance seems to have broke.

January 25th, 2005

• Testing the code we didn't get a chance to test yesterday as I type. Only got 4 hours of sleep so I'm a bit dsypeptic. Hopefully all will go well, if not I'll just fall asleep in a corner somewhere. -Abraham
• Heading back to to my room for a phone call about a job offer. Jimboo got stuck on a wall during the 2nd mock contest and it made me sad. Right now I await death's sweet embrace. -Abraham
• Good God the transdrive rules. Now to actually use it properly... -Gabe

January 24th, 2005

• Did some optimization for my multiple ball finding, making the search space tighter. Still waiting for the team members who went to buy screws for the high torque motors. I think they're dead in a snowdrift and without them, we can't test our new movement code. Alas. -Abraham

January 23rd, 2005

• Did a quick optimization involving the accessing of image data. Took all of 10 minutes. Waiting to get the camera remounted on the bot so I can do further optimizations. -Abraham
• Did a bit of fiddling with the camera settings to get sharper images. Huzzah for noise reduction. -Abraham
• Round out our state machine so it will scan at proper times, and go to goals at proper times, and butt into goals, deposit balls, know when it has balls after capturing them. It's all 100% untested. Let's hope the previous bugs haven't perennated themselves... jasonkb

January 22nd, 2005

• Re-wrote and finished our barcode reading code. It's pretty robust with random sampling and all that jazz. Tomorrow is reserved for estimating distance to walls visually and a grab bag of optimizations on my image processing code. -Abraham

January 21st, 2005

• 5:45PM: Well, the new dual-shortranged IR sensor system does make us avoid walls. However, if we go to a corner, the robot is stuck turning from one wall to the next, paralyzed. --GabeWB
• I updated yestreen, but it was not saved. Things were bad with sensors yesterday but I have written a new exploration mechanism that had a good run today not hitting walls. code for scanning and getting balls is there, but probably buggy as hell. :wq, jasonkb

January 20th, 2005

• The mock contest was today and we actually didn't suck. I was pleasantly surprised. Going into cruch mode this next week to make sure things all come together for next week. Tired but not so dispirited anymore. -Abraham
• We've decided not to use global headings from the gyros, due to drift among other problems. We tried using two long-range IR, but it just wasn't useful enough to be worth it. We've decided that short-range IR, for the specific purpose of wall-avoidance and close-quarters navigation, will suit us better. We're going to try dual shorts tomorrow. We're considering having them forward facing or cockeyed. Because of our recent sensor decisions, we're looking into alternative navigation systems for finding balls efficiently and, most importantly, getting back to a mousehole in time to score. --GabeWB

January 19th, 2005

• This weeks just keeps getting worse. On the bright side, the rgbToHsv conversion now magically works again. Go figure. However, we seem to be having some firmware issues and as such cannot actually operate the robot. We're waiting on another #2 chip. Until then I'm just waiting for the entire thing to catch fire randomly and throw itself out of a window. That would be a fitting end. Updates as they happen. -Abraham
• We swapped out the orcboard and most everything seems to work properly with the exception of the gyro. It basically goes crazy when it reaches the pi/-pi state and doesn't know which direction it's supposed to be turning so it just loops endlessly. I think we might just ditch the gyro completely. We did finally manage to get checked off for CP2 after much wrangling. I doubt if we'll put up much of a showing in the mock competition tomorrow, maybe collect a ball or two. I get the distinct feeling we're behind the other teams and have to play catch up now mainly becuase we had so many problems with things just mysteriously not working. Anyway, I'm headed to lab early tomorrow in an attempt to get things working better. - Abraham, discouraged.

January 18th, 2005

• Was minding my own business, working on some of the image processing libraries and suddenly we can no longer detect anything. It looks likes a new jar got rolled out this evening and it killed my feature detection code. As near as I can tell, everything's reading as yellow. The balls when converted to hsv look a sickly greenish yellow. Add that to one of our motors going south and an encoder port not working and we're getting fucked from all directions. I'm awaiting word from the staff as to whether this rbgTohsv conversion is working and my parameters have suddenly become invalid or if the new jar just fux0rs things up. -Abraham
• Improve drive system and state machine, which in theory are robust but in test runs are as good as psychedelically bad. Why are the jars updated automatically? We all would prefer to have the updates be manual. Anyway we scored a point a while ago, but 'twas twigging in an empty forest. And irreproducible. But should the motors work correctly and the IR sensor (we had it replaced, and now have to return the old apparently broken one) also operate with reasonable reproducible accuracy we would be showbizzy. sigh jasonkb

January 17th, 2005

• Worked a little bit on the image processing and drive system code. Mainly just minor changes. The bot does drive although we need to take its clearance into account when turning. Surprisingly enough, it drives straight without any encoder info, a testement to Greg's skills. Tomorrow we'll be fine tuning things and probably getting the 2nd checkpoint done although technically we did it this evening, no one was around to see it. -Abraham

Weekend, January 15-16th, 2005

• Mystery Hunt consumed us. Spat us out after two days. We'll be back in lab on the morrow.

Friday January 14, 2005

• Finally calibrated the camera so we can get the an the relative angle between the camera and ball objects. I refined our ball finding code a bit and am in the process of estimating distance based on the size of the balls. We haven't tested the robot must yet today and we've been working pretty hard all week. This weekend is mystery hunt and we could all use a break so we're going to cut out around 8pm and call it an evening. -Abraham
• Didn't get checked off but we're tired and my back hurts. We'll be back another day. -Abraham
• Make State Machine fully operation. Finish testing classes for ball collector and imager, and make some debugging tests. Other things, but my memory is shot lately. - jasonkb

Thursday January 13th, 2005

• Today, I devoted my time to image processing, multiple ball detection to be exact. I got a pretty good way of detecting multiple balls close up and I'm working on a method that detects balls that are far from our camera as well. I've just wrote that code, all that remains to do is test it. Hopefully it'll work and I can call the day a success. -Abraham
• Rewrite a superflexible state machine. make drive system and IR sensor into an interface and abstract class. - jasonkb
• W00t, got multiple ball detection that will detect multiple balls up to about 2-3 meters away from our camera. Not the world's fastest algorithm but it works. -Abraham

Wednesday January 12th, 2005

• Today was a long yet somewhat productive day from the mechanical side of the robot. Basically I worked on the rest of the chassis and ball collection mechanism, and mounted the optical encoders. Unfortunately most of the day went towards waiting for the glue to dry on the top "ramp" portion of the robot (this basically guides the balls which we collect towards the depositing mechanism in the front). Since the size of this dictates the spacing between the two support walls, I have to wait until the ramp was finished before mounting the support walls (and the "shelves" which hold the computer, battery, camera, and several sensors). But nevertheless, the ramp is done, and I've decided to mount it to the chassis with velcro-angle brackets that I made today, which should be pretty spiffy (the original plan was to actually glue the ramp to the support walls, which of course isn't the best plan in terms of maintanance and assembly). Ideally, I'll have everything mounted midday tomorrow, except the ball collection and deposit mechanisms, both of which should be done tomorrow evening. Practically, I don't see any major complications arising except perhaps a flaky chain-drive system or an ineffective ball collection mechanism (specifically the belt), both of which should be solvable with clever engineering. I guess I'll find out tomorrow. - Greg
• A short day for me today since I had to leave early for a dinner. I attempted to work on a algorithm that would detect multiple balls. Unfortunately it was accessing data it shouldn't have and was crashing. I'm sure I just have a test condition somewhere that's not correct and once I fix that I'm confident multiple ball detection will be cake. -Abraham
• Begin restructuring a more robust and flexible state machine that isn't just a stupid switch statement. make debug easy to use. put optical sensor stuff into wheel, test it. write an irsensor class and semi-calibrate it, semi-decently. learn how to serialize objects, which I'll use for testcases. learn a nice nine-letter word that vicissitudinously describes our robot: superpimp. - jasonkb

Tuesday January 11th, 2005

• Finally finished checkpoint 1 after a bit of an ordeal. Discovered that one of our motors was bad and as causing both of our motors to turn off. Once we fixed that we were able to once again rotate and drive correctly. The bottom level of the chassis is coming along nicely and we should have a ball collection mechanism tomorrow. Since our code is general, only a few changes to be made so we can score points. More later tonight. -Abraham
• So I finally finished constructing the drive system, and after a few tests everything seems to work fine (woohoo). I spent most of the day mounting the motors (cutting chain, drilling holes, getting the motors at the right height. etc), something that I expected to take a lot less time than it did. All in all, I think I spent about 20 hours in the machine shop this past week making all of the bits and pieces to make this drive system work. Hopefully it was worth it, but at the very least it looks cool. So, tomorrow's agenda includes mounting the optical encoders and constructing the rest of the chassis. I'm not entirely sure of all the details yet, but I intend on making the frame of our ball storing mechanism, ball collecting mechanism, and a place to mount the electronics. This might be tricky due to some of the slight angles in the ball collection mechanism. Anyways, I'll take care of it tomorrow. - Greg

Monday January 10th, 2005

• Did some work with imagining and stoping in front of balls. Had everything working to get checked off for CP1 but then the bot stopped stopping when it got close to the ball. Annoyed the hell out of me, something seems to have broken in our state machine. Our chasis is coming along nicely though, lab closed early but supposedly it'll be open late tomorrow and that's where I'll be. -Abraham

Sunday January 9th, 2005

• Put back together the peg bot so we could iron out the bugs we have in our rotation code. I discovered earlier this week that the gyro rapidly drifts, I'm not sure how to stabilize this, I'll probably have to write a much better gyro callibration routine. Either way, we can now complete checkpoint one as soon as the methods to make decisions based on processed images are written. -Abraham
• Did some more work on the image processing code but testing is delayed because the wireless connection to the bot seems to be acting flaky and I don't have a spare cable on me right now. I guess I'll just test it tomorrow afternoon. Jason and Gabe did some work on the state machine that controls our bots behaivor, seems promising.

Saturday January 8th, 2005

I guess I should update this a little more frequently, but here's where we stand on the mechanical front:

• The first story of the chassis has been constructed, and I've mounted our three drive wheels and casters. In terms of our previous problem regarding all three wheels making contact with the floor, I machined the pillow blocks with enough accuracy so that when the robot is carrying the battery, the chassis deflects enough to make things work out (horray for non-complicated solutions). All I have left to do is mount the motors and a few sensors.
• The second story of the chassis need a bit of work (more glue and another surface to mount the camera), but I'll take care of that tomorrow. Also, I'm going to rig up a nifty camera mount so we can adjust the angle to whatever works best.
• I took a little trip with Gabe over to a place called "Bikes not Bombs" in order to aquire some gears and a chain. The mission was successful (only cost $5.25, w00t), but this also means I'll have to spend another day in the Hobby Shop attaching these things to our wheel axles and motors. I figure I'll buy a 1.5" aluminum rod, cut it into disks, drill a center hole with a lathe, then press-fit the gears to the wheel axles, and make set screw holes for those that attach to the motors. Unless something funky happens we should be able to drive monday night, I promise.
• Also, I'm trying some Silicone Rubber Sealant to frictionize the wheels a bit more. The Contact Cement helps, but I'm not satisfied with the result.
• Hopefully by Tuesday I'll be ready to put together our ball collection mechanism. At the very least, It should be done by Wednesday or Thursday. I don't forsee any funky parts that need to be ordered/machined. Oh, that reminds me, I should check out that belt place. Yes, belt place..... - Greg

Friday January 7th, 2005

• Managed to get a decent amount of work done today. Wrote the beginings of our drive system and got ball detection working. Unfortunately just when I was going to put them together the camera mysteriously stopped working. After about 45 minutes of cursing and head stratching Yuran suggested switching off x forwarding. Sure enough this solved our problem but by then it was too late to get checked off really. Oh well, we'll get it done at the next earliest time. Overall, things have progressed likely and I see us having a near fully functioning robot next week. -Abraham

Thursday January 6th, 2005

• Was out of town for an interview, delayed both ways and exhausted. I plan on doing some massive coding all tomorrow though. -Abraham
• Fourth team member returns. Gets a feel for java and Abraham's extant code. cheers on Greg as he machines wheel mounts and fictionizes wheels. Changes Demons->Daemons in team name.

Wednsday January 5th, 2005

Vision is an unholy nightmare. I've been tinking away with my detect barcode code all afternoon, trying to come up with threshholds that work. I just managed to get it to differention between black and green fairly well. Now I just need to come up with a half-way decent method of figuring out how many "squares" are green. I've thought about it a bit but everything I'm trying doesn't seem very robust. More news to come I'm sure. -Abraham

• Finally got the bot to read the barcodes. Not the worlds fastest procedure but it works. Makes me not want to use barcodes for navigation though. -Abraham

Tuesday January 4th, 2005

We're currently thinking of using a combination of short and long-range distance sensors. The long range sensors would be useful in helping with wall avoidance and perhaps mapping in conjunction with the camera. The short range sensors would be useful for ball detection and close range wall avoidance. We'll definitely need optical encodes for our wheels with some misc bump sensors thrown in for good measure. We're most certainly using the gyro, I love those little things(Abe).

• I tried writing a little chunk of code that would forward what the robot's camera was seeing to the botclient. Unfortunately, this caused the application to crash. The staff is looking in to getting that fixed. Otherwise we pretty much just did what was required for the day. -Abraham

Monday January 3rd, 2005

Day One, Morale is high. The summit is in sight.

Things to note:

• Forth (sic) team member missing. Presumed dead in some scrabble-related incident.
• Abe has black electrical tape on his back, which he still hasn't noticed.
• We got everything on the list finished this evening with a few hiccups such as getting an unprogrammed lcd chip and general java issues. I do hate java but I suppose I'll get used to it again. -Abraham