Team Thirteen/Journal
Day 1 (1/6/14)
Assembled pegbot mechanics (but not wiring). We met at Next in the evening decided on how to solve problems of ball collection (harvester), lifting (archimedes screw), depositing (drop down ramp + extra roller for green balls) and harvesting from the silo (metal rake). Also decided on the overall strategy that the robot would ideally take (wall follow, and remember where landmarks are).
A lot of parts were scrambled around lab today so we left a little early because we couldn't get anything done. The nuts and bolts didn't seem to be matching.. Will finish kit bot when we have all the parts to do so.
EDIT (1/22/14): DON'T DO A SCREW. NEVER DO A SCREW. HEED MY WARNING.
Day 2 (1/7/14)
Spent time today doing most of the wiring on the Kitbot. We ran into some issues with the connection to the battery, which we will work on tomorrow, but it should be easily fixed once the plugs for the battery are delivered (they were delayed by snow). We made a janky tape solution to keep wires attached to the battery, but it's temporary and not really working.
Uploading code to the Maple and installing the drivers was also a pain today. Eventually we got it working, but it took way too long. The Maple wasn't being recognized as a device so we had to fudge around with installing the drivers a couple of times.
We also started thinking about possible designs for the actual competition robot. Neil begins a CAD.
Day 3 (1/8/14)
Today we finished the kitbot and tested it with the tablet - the sample code works. We also started prototyping some of the parts we may use in the final design, such as the ball harvester. We experimented with different dimensions and tried to find the optimal design - Clare laser cut out a small jig to test different heights of the harvester above the ground (and took it to Shaw's while grocery shopping..)
In addition, we got the visual systems running and tested it, so now the robot is able to track both the red and green balls/lines. We also continued modeling potential designs for the robot.
Day 4 (1/9/14)
Clare began working on the ball lifting mechanism using wire. We looked up examples of Archmedes Screws online and tried making them. We had difficulty prototyping a very durable screw. First the gauge wire we used wasn't thick enough, so the wire would bend under the weight of the ball. We had trouble finding a tube/cylinder to make the screw on too, so that the actual screw wasn't even/concentric. Neil ended up going to racecar and taking a packaging tube. Still stability was an issue. The screw wire would flex around.
Neil continued modeling of robot/parts on Solidworks.
Coding team also developed the visual control system for the robot further.
Day 5 (1/10/14)
Clare and Dabin continued prototyping the Archimedes Screw mechanism using cardboard and a laser cut jig. Lesson learned - laser cut or build a jig so that what you are prototyping is mounted and constrained! It worked wonders in letting us know what was wrong. But still, we continued to have issues with screw diameter and the wire bending.
The coding team continued work on wall following and distance measurements (based on height of the blue strip on the camera)
Day 6 (1/12/14)
Lab was open on Sunday.
In terms of code, developed means of determining how far away we are from the walls.
The Archimedes screw still was having tons of trouble. On Saturday, Clare CADded a screw to laser cut as a solution. We cut it out on Sunday and assembled it. Seems to work? Edited CAD files to incorporate this new design.
Days 7 and 8 (to 1/14/14)
UPOP week - Neil is gone for most of the day this entire week.
The software team focused on improving wall sensing using the camera, and on extrapolating walls in the camera's blind spots. We made progress, but wall extrapolation is still unusable, as of 1/14. We set up ultrasound sensors on the kitbot, as a possible replacement for camera wall extrapolation.
Clare picked up CADing the robot because Neil is gone, and the screw was put on the back-burner in order to finish the CAD. Build team worked on getting the top layer of the robot CADed. We rearranged the ball sorter and inserted holes/pegs to put everything together. Clare became less of a noob at Solidworks.
Day 9 (1/15/14)
On the software side, we gave up on wall extrapolation entirely. Most of the day was spent finding two ultrasound sensors that worked. Added map support for ultrasound.
Clare and Dabin finished CADing the bottom two layers. We are a little behind in terms of real construction, so we pushed to get the bottom finished so it can start to be assembled, and we can start driving around our final robot base.
Day 10 (1/16/14)
First attempt at wall-following - can kind of do it. Ultrasound continues to be a little flaky.
Mechanical team began cutting out base for final robot. Started assembling the bottom portion of our robot. We had issues finding supplies though - seems like other teams beat us to a lot of equipment.. We ended up scrapping Cake-Bot to get the screws and bolts that fit our design. The wheel o-rings are a bitch to put in (set screws suck).
Day 13 (1/19/14)
Not much work the past two days. (Traveling for interviews, Mystery Hunt, weekend, etc)
Began work on localization using sensors. Wall-following is hampered by camera performance at close range. Will add forward-facing ultrasound to fix this problem.
Day 14 (1/20/14)
Building - worked on wiring new robot. Stripped down pegbot so we can use its encoder. Made alterations to the top layer of robot - we decided to make it shorter and get rid of the ramp system, so there was a pretty big overhaul of our CAD. This way we can drive directly up to the reactor and not depend on ramp deployment at the right angle. Hopefully tomorrow we can begin assembling the top half to the bottom half.
The screw comes off the back burner now that we are putting together the physical robot. Unfortunately that doesn't make it any less hard to position right. Clare and Dabin worked on the screw for an hour or two each of the past days. The laser cut MDF screw was too rough and caused a lot of friction with the ball. We tried sanding down the edges, taping on wire, etc etc. But it still wouldn't work.
We decided to basically just scrap our previous work. Clare made a new screw out of wire and attached it to a wooden dowel under tension. We made another jig and tested it - seems to be smooth! Instead of using an L as support, we are now using a steel rod we found in lab, and the ball travels upward much smoother. Will finish things up tomorrow?
Day 15 (1/21/14)
Goals for the day: Building: Finish assembling the bottom half of the robot today so that it can drive by tomorrow. This means attaching the ball harvester and getting it to work with our ramp. Power it with a motor.
Made a lot of headway on assembling our robot. We wired our breadboard. Clare also got all of our pieces laser cut. We are in the process of putting everything together. Neil put together our aluminum ramp for the ball harvester. We are also working on making hubs and getting all our our pieces connected and able to move. We cut hubs out of MDF because we were told we were using too many real ones.
Day 16 (1/22/14)
Goals for the day: Building: Finish what we didn't last night with getting the bottom half to move. Assemble the top half of the robot. Hopefully fine tune the position of our screw (plus its rod) and get it working with our motor pulley system. If possible, attach the mechanism for ball sorting and the side walls (out of sheet metal).
SCREW IS A BITCH. But it works by hand now. We drilled holes into our MDF base to mount the screw. We glued on the aluminum ramp for our harvester. It feeds into the screw well now, but we might have to do some troubleshooting with the front part where balls come in from the harvester. The ball exits the screw in the right direction, into where we put our color sorter. Successful day!
We decided to scrap the pulley system for turning our screw. It'd be unreliable and prone to slipping. Plans for now are to attach a motor at the top to spin it. Clare is working on a housing support for this top motor.
Day 17 (1/23/14)
Goal for the day: Building: Add the motor for the screw!! Insert a color sorting sensor at the top. We also need to add walls to the top (general ball constraints). Coding: Get localization tuned up - add gyros or IR.
Screw motorized and reliably picks up the ball (as long as running <4 volts), along with kicking the ball out at the top using a piece of wire that disturbs the ball from its path just right. Color sorting sensor LEDs set up with newly drilled holes, but still needs to be coded for. We decided on about 2.5 volts to make sure that the screw is going slow enough to pick up the ball. Otherwise, the screw just bounces the ball right back out.
Day 17, Vipul's Course 20 status remains unnoticed.
Day 18 (1/24/14)
Goal for the day: Troubleshooting - finish what we got stuck on in the previous days. Wiring, sanding down, etc.
Today, we soldered and wired up the robot. Lab is running out of wires, so our nice red/blue/black color scheme is quickly disintegrating... Looks like our choices as of now are yellow, yellow, and yellow... Wiring is a hot tangled mess..
The harvester wheel and ramp seems to be the last mechanical roadblock we are running into. The harvester wheel is scraping the top of our robot so we filed it down a little. It looks like we accidentally used a larger wheel that was previously cut out. Oops. The ramp is working when the ball is fed in through the center but not the sides. We are considering building a wall so that the ball can't enter from the sides (as an alternative to changing the spacing of the wheels, which wouldn't work with how we constructed them)
We may have back driven current into our Dagu while testing our motors with the shop power source - hopefully nothing is fried. eep.
Vipul will be writing the code for the color sorter over the weekend.
Day 19 (1/25/14)
Felix came into lab to test code. Localization is working. Also, we discovered that putting the tablet into our tablet holder changes the center of mass just enough so that our ramp no longer scrapes the floor when the robot moves - hooray.
Day 20 (1/26/14)
Everyone came in today.
Dabin, Neil, and Clare worked on troubleshooting. We added side walls to our harvester wheel so that the ball enters through the center. We also added directing walls to the top of our robot: we didn't use countersink screws, so they stick out of the MDF. With the new aluminum walls, we direct the ball away from the spots where they get stuck behind screws. We also soldered more wires and tested the harvester wheel, ramp, and screw in conjunction for the first time - they work!! We also cleaned our table. Things look much more organized now.
On the coding side, Vipul and Felix also continued work. Vipul is working on the color sorter: we bought some resistors and LEDs from RadioShack to try to get it to work. As of now, we can't detect a significant difference between the red and green balls but we're still working on it! Felix fixed the robot's turning - the issue was that all the gains in our control were set too high.
Robot accidentally scored its first points today - we accidentally drove into a green ball and sent it into a reactor's lower opening. It's a start?
Day 21 (1/27/14)
It was discovered that balls were getting stuck on our top ramp so that they couldn't dispense properly if there were over 2-3 balls being stored. Clare and Dabin constructed aluminum walls so that the balls line up single-file and don't get stuck as easily. They were hot glued onto the top of our robot. We also wired the harvester wheel and screw and tested them in conjunction - they work with our battery. Some of the hot glue holding the screw together is coming apart so we will reglue that tomorrow. Lastly, the wires are less of of a mess now.
Felix continued on the coding front.
Tomorrow we need to hook up the last servo - lab ran out of them so we had to order our own. Hopefully it'll be in tomorrow.
Day 22 (1/28/14)
Clare and Dabin wired more and tried to stabilize components of the robot (aka more hot glue.)
Vipul is working on the color sorter. We found that one of the servos is broken and can only turn clockwise. Thanks to staff for coming in (from watching paint dry..?) and giving us the servos we need! We accidentally fried a servo today, but we also got it replaced later in the evening. It's almost 1am, but color sorting seems to be working really reliably.
Felix has been tuning navigation code to look for landmarks.
Things to keep in mind: there seems to be some sort of depression in our aluminum ramp that the ball can come to rest in and never reach the spiral. Hopefully when the robot is moving this won't be an issue. Also, our spiral seems to have gotten misaligned. The ball is going up, but it's definitely taking more torque to get past some regions. Not sure how we can remedy this at this point - we may just have to cut our losses and hope nothing continues to act up.
We still need to code the flap servos so we can deposit balls into reactors/over to the other side. That should be quick and simple, hopefully (trying not to jinx it.)
Tomorrow is seed. eep!
Day 23 (1/29/14)
We got second seed, holy crap. We couldn't get color sorting to work so we tore off the aluminum housing and just collected balls, getting all of the red ones and all but one green one. If more than one ball goes into the spiral at once it gets pretty badly stuck... The robot went really fast while getting the balls so we're going to slow it down. The tablet also flew off the holder so now we are using rubber bands to hold the tablet.
But it seems other teams were unprepared and will have their shit together by tomorrow/Friday. There's still work to be done.
Note for the future - have some sort of queue for the ball sorter. As of now, the spiral goes directly into the ball sorter, and if we pick up a second ball then it gets stuck in our spiral... That's a design flaw that probably can't be altered at this point.
Day 24 (1/30/14)
Mechanically we are basically fixed in place. There could be some work done, but at this point it works and it's good enough for us (hopefully). Clare repositioned the bottom of the screw and glued it in place, and the spiral seems to be less stuck now because it went a lot faster when we ran the bot.
Vipul continues to work on color sorting with the photoresistor. As of now it is relatively consistent, but we will have to definitely take advantage of the calibration period tomorrow.
