Team Ten Journal

Team Ten Journal

January 28th, 2005

In lab:

• Added code so we could push a button once to start our program and again to actually run it
• Tweaked a few really minor things (wings, bump sensors)
• Tightened lots of screws
• Printed Hefty Smurf logo on stickers for decorating our robot and distributing at the contest

January 27th, 2005

In lab:

• Fixed some bump sensors, moved IR sensors back, and added anti-get-stuck wings. Hopefully no more physical changes!
• OK, we lied, we also had to glue a wheel back on, and raise an IR sensor. And we might have to fix a wing tomorrow.
• Made wall-following actually work! Many many mock runs.

January 26th, 2005

In lab:

• Attached ball-storage tray to top of bot
• Made bracket to hold battery
• Fixed lots of physical things (wheels, bump sensors, chain, etc.)
• Worked on wandering/escape code. Also, goal finding goal. If very very rigged, it will find a ball and score!

January 25th, 2005

Mock competition:

• Everything broke! (More specifically, our orcpad didn't work at all until we replaced a chip, and then it only worked if we put pressure in the right place. Other things were slightly less serious.
• Robot still confused about what a ball looks like... Need to make sure it doesn't recognize a thin column of red as a bal

In lab:

• Tried to fix orcpad further. No success until we dropped a ball on it, at which point it started working.
• Got chain to (mostly) work!

January 24th, 2005

In lab:

• Mounted bump and IR sensors
• Attached orcpad to orcboard
• Mounted camera
• Deal better with ignoring balls that are already in goals.
• Work a little on recognizing bar codes
• Implement too-close IR stuff, and start dealing with bump sensors
• Trash the multiple-strategy code, and put in the 3 minute time limit

January 23rd, 2005

In lab:

• Lots of little code tweaks.
• Added a calibration helper program. Added yellow goals. Benchmarked and figured out that using the DataBufferInt? thing, even though it doesn't work right with ImageIO, is ten times faster than im.getRGB. Added timeouts to the turnRadian things. Fiddled with lots of parameters.
• Can choose whether or not to waste time sending images to the channel (and drawing on them): "sudo ant run" vs "sudo ant images".

January 22nd, 2005

In lab:

• After significant frustration, managed to epoxy our gear such that it won't slip.
• Worked on attaching chain to gears/rollers.

January 21st, 2005

In lab:

• Figured out how to attach bump sensors, and promptly discovered that there were no more of the sensors we like (the small kind with the metal switch). Anyone know where these came from, or how we can get more?
• Mounted chute, rollers, gear for chain.
• Rebuilt one-way gate for front of bot yet again.
• Lots of code tweaks. Got "turn X radians" kind of working.

January 20th, 2005

Mock competition:

• Problem 1: Couldn't get real IP (fixed on site)
• Problem 2: Bot would barely move (fixed on site)
• Problem 3: Bot recognized nearly everything as red (fixed later in lab)
• Problem 4: Wheels slipped, acted weird, etc (fixed later in lab)

In lab:

• Fixed aforementioned problems.
• Worked more on tray, chute.
• Mounted computer on base.

You know you've been thinking about MASLab and mystery hunt too much when

• You have a dream about figuring out, "No, we don't have to follow the walls. That's all wrong. We just have to look at every picture that has a ball in it and take the first letter..."

January 19th, 2005

In lab:

• A bunch of code tweaks. Make it search the image on-demand, instead of continuously in a separate thread. Mess with motor powers. Pass checkpoint 2!
• Made tray and chute for the top of our robot. This is where the balls are collected and then released into the field goal.
• Made ramp leading into PVC pipe to lift the ball off of the ground. The flights on our chain were made less wide as well.

January 18th, 2005

In lab:

• Took entire robot apart with hopes of finishing all cutting/drilling of base. Didn't even come close to finishing. Made some progress in attaching gear/rollers for chain, and dealt with logistics for attaching computer, etc.
• Rewrote a lot of the image code. Downsampling implemented.

At pika:

• Reassembled bare minimum amount of robot in order to hopefully make checkpoint two tomorrow.

January 17th, 2005

In lab:

• Made better one-way gate.
• Made/attached flights for chain.
• Made rollers for chain.

January 13th, 2005

In lab:

• Worked more on assembling actual robot. Made a one-way gate that was slightly sketchy but worked.
• Banged a lot of code together. Implemented the CameraController?. Way too slow. Combined RandomWalk?, FindRedBalls?, and NotTooClose? kind of manages checkpoint 2, only not. I guess we'll miss checkpoint 2 since we'll all be Mystery Hunting tomorrow.

At pika:

• Tried to weld but found that pika's welder was broken. Found another way to attach flights to bike chain without welding.

January 12th, 2005

In lab:

• Gave design review.
• Attached castors, wheels, and "funnel" for directing balls onto bottom of new base. Found that funnel design was not terribly effective (often causing balls to bounce out), but decided to leave it as it was for the time being and experiment with the effects of adding a gate.
• Made SensorController? work, implemented NotTooCloseStrategy?, added the chain to the MotionController?.

January 11th, 2005

In lab:

• Went to look at belts and chains. Home Depot didn't have much. Colmar Belts in Boston was less useful than expected. Hubba-Hubba had lots of chains and belts, but nothing that seemed quite right. Ended up buying a chain and gear from Cambridge bicycle. Also checked out the MIT museum for ideas.
• Designed and cut base for non-pegbot.
• Worked on attaching gear to motor.
• Made slide show for design review.
• Finally got code organized, and set up in an actual working svn repository, and building with ant. Created the Strategy, MotionController?, and SensorController? classes; the latter two have both a "real" Orc version and a "fake" testing version.

January 10th, 2005

In lab:

• Remounted pegboard to allow clearance for balls and experimented with walls underneath to direct balls.
• Spent more time discussing Archimedes screw idea and decided that we favored using a conveyor belt to move and lift balls.

January 9th, 2005

At pika:

• All team members meet together for first time. Much discussion of strategy ensues.
• Made a model Archimedes screw out of soda bottle, cereal box, dowel and duct tape. Found that it works (sort of) for lifting laundry detergent. Wondered about feasibility of making a similar screw to elevate balls.

January 7th, 2005

In lab:

• Ali is back!
• Ali works on code for checkpoint while Maggie and Yuan experiment with servos and sensors.
• 4:15: First somewhat-successful try with new program to make robot turn until it sees a ball.
• 4:35: More successful try after some editing: Robot turns slowly and stops upon seeing ball!
• 7:10: Got checked off! Still need to work on stop mechanism, blue line filtering, etc.

January 6th, 2005

In lab:

• Dave helps explain code till last minute ... as he rushes off to catch a flight.
• Do some colour recognition.
• Yuan and Maggie flail hopelessly while trying to understand/update code, eventually decide to discuss plans for mechanical design instead.

January 5th, 2005

In lab:

• Camera sees object successfully. Botclient sees what camera sees successfully.
• Dave works on code for green box and red ball recognition.
• Yuan and Maggie work to become more familiar with Java.

At pika:

• Green box (well we used a different shade at home) recognition worked upon.

January 4th, 2005

In lab:

• Checked in yesterday's lab
• Assembled short-range IR sensor cable
• Got OrcSpy running on Athena (needed to add --allowremote to /etc/init.d/orc-daemon)
• Used OrcSpy to examine characteristics of short-range IR
• Somehow managed to independently break our LCD and our Orc Pad. Rewired them from scratch.

At pika:

• Got everything back together! (Found the potentiometer we needed in the Murphlet, phew.)
• NotTooClose, our "stop when we're within six inches" demo, works!
• Figured out how BotClient worked, made NotTooClose send messages to it and receive commands from it.

January 3rd, 2005

First day of Maslab

In lab:

• Looked through our kit.
• Started soldering Orc Pad.
• Joy when given LCD display; shock as realisation slowly dawned that soldering was not quite right. An hour passes: problem fixed -- but nothing shows up on the LCD screen. Contrast fixed and all is right with the world.
• Finished soldering gyro (after an awful lot of desoldering)
• Assembled pegbot

At pika:

• Figured out that in order to turn the compute on, it needed to be connected to the Orc Board.
• Figured out (after a long time) that the Orc Board and Eden did not solely communicate through the power cable, and plugged in the USB cable. Found that our Eden was hefty-smurf.mit.edu. Set up ssh for X-forwarding (as suggested by TeamSeven). Created an RSA thing for passwordless login as documented at http://maslab.lcs.mit.edu/2004/wiki/SSH.html. apt-got sudo and set up /etc/sudoers.
• Tried to get video to work. <pre>ln -s /dev/video0 /dev/video</pre> seemed necessary but not sufficient, as the Java video stuff barfed on linking with a C library, I think.
• Wrote our Hello World, complete with a little bit of dancing.

Before IAP

• Dave: tried the odometry tutorial, got the instructions to work but had no luck on getting any of the calibration to be decent.
• Dave: I worked on the color tutorial, got an OK start on finding multiple red and blue blobs, though the coding needs improvement (need to read the Java book!) and it's not doing anything smart with the blobs yet other than finding connected components and ignoring ones that are too small. Also, haven't actually done blue-line filtering yet (though if I wanted to I could implement the dumb code from the example without too much trouble).
• Maggie: checked out some books on Java, tried writing some simple code.