20100112

Design Challenge - LEGO vs Sleep


I had an ambitious idea that I would be posting regular interpretations of my MIT SDM experience on this blog.

Ha.

The january program is intense enough that there isn't enough time for sleep, much less regular writing. Heck, I'm writing this during a lecture right now.

Week 1 was dominated by Design Challenge 1, the goal of which was to build a robot which would perform a variety of different tasks. I have been burned by similar challenges before (2.70!) and learned then that the most important thing is to test, test, test. Our team agreed to physical and code freeze by thursday night, then spend all Friday testing. We also had an idealistic vision that we would not work too late any given night. We didn't meet either of these goals totally, but we did manage to spend most of Friday testing. Unfortunately, we tested in the wrong lighting conditions. At the competition, our robot kept seeing its own shadow and confusing it with the lines on the field. Because we tested in a dimly-lit room without shadows, we thought that our robot was 80-95% reliable depending on the event.

Instead, we ended up failing miserably. We scored NEGATIVE points on 3 of the 5 challenges and came in last place. Given how much time we sunk into the challenge and the strength of our test runs, this was a crushing result. I went home Saturday night sleepdep'd and numb.

Once again, the lesson is test, test, test. Except that if your tests don't reflect field conditions accurately, you're not really testing.

No comments:

Post a Comment