Monday, September 29, 2014

Robotic Fabrication Class

This semester, I started taking an "Intro to Robotic Fabrication" class because since architecture is finding more ways to incorporate technology into design, development, and fabrication, I though why not.  We use Rhino as our base software but also use the Grasshopper and Kuka PRC plugins, pretty much as a user interface, to help us program and visualize the robot movement on the computer.

At first, I was pretty lost because whereas almost the all of my classmates were grad students, I was a sophomore.  I was just beginning to learn how to use Rhino and I hadn't actually programmed before.  A month into classes, though, I'm feeling like things covered in class aren't flying so far above my head anymore.  I'd like to thank my group members and the teachers for that because I've gained most of my knowledge from simply watching them.



Our latest project involved attaching a pen tool to the robot to draw lines on a piece of vacuum-formed plastic.  Since the small robots that we were using have a smaller range of motion due to the wires and air tubes bundled up inside, axis 4 (the rotating action of the "forearm") frequently runs into errors from the program trying to turn the axis too far.  This happened mostly when the lines crossed the midway-line of the robot's work space.  The text in the beginning of the video helps explain what we did to help remedy that problem, although patching of the program was still necessary.  

I feel like this project was the one where I know most of what I was doing so far.  I did less sitting around and watching and more actually working the robot and doing various other tasks besides video editing.  I also started to understand what was going on on the computer screen and why some things worked and some things didn't.  I'm not sure why, but I have this preconceived notion that digital technology is not very reliable.  I'm more comfortable with mechanical devices where if something doesn't work, you can locate the problem visually and go at it with a screwdriver.  Whenever any of my digital devices, whether my phone, computer, or alarm clock don't work properly, I'm not surprised.  That doesn't make me more patient though, as I have to use this technology if I simply want to keep up with school.  I was hoping to stick with a non-smart phone for as long as possible but the rate at which the school's technology is moving forward made that impossible for me earlier this year.  Anyway, when it came to digital robots, this turned into a feeling of helplessness when something doesn't work quite right.  It's almost as if I was not motivated to fix a bug in the program because of my distrust in digital technology.  Thankfully, I'm now finding out  that it's pretty easy to do with the equipment we're using for class.  



Our previous project was to attach an endoscope on the end of the robot (getting the tool to show up in the program was a task in itself) and use it to pan around a small model.  In order to begin drawing the tool path in the program, we had to scan our model, which in my group's case was my toy '65 Mustang, with Autodesk's free app 123D Catch to get it to show up on the computer.  To put it simply, you take 20-40 photos from various angles around the model so the computer can tie them together.  


It took me at least five tries to get a decent scan of the car.  The one above was the third try when my friend, who's pretty computer-savvy, tried to use his phone to scan.  It's a hollow shell of half of the car.  One way that Autodesk tries to aid the process of taking photos at a correct angle was to take advantage of the accelerometers that most devices have.  I had a heck of a time trying to get mine to work.  


My last try, which took the whole night to render, ended up good enough for our group to use.  I found out later that my images were a bit distorted because the car was shiny, making it harder for the computer to detect the depth of the 3D surface from the photo.  Anyway, it worked and the project came out as we intended.