The Bending of The Rules
The coarse course rules changed course a little bit over the
course of the course.
Some things I recall:
- zip-ties were allowed for structural purposes.
- the boss was a little lenient on the 1 ft3 volume restriction.
We should have had a steep fine (not disqualification) for being over
the limit. We should also have a box (like they have at airports for
carry-ons :v) to make sure folks actually fit the limit.
- The tables that made up the course were bowed, so there were spots
with very big bumps that threw off odometry and caused dragged food blocks
to get stuck. Next year, I'd advise "paving" the course in taped-seam
foamboard, with the seams always at the middle of a table, not at the joint
between tables.
- Very few robots needed all 15 minutes. Instead, most robots could
have used a second run. Some robots had last-minute algorithm bugs that
would have made a second identical run worthless. For next year:
Have two competitions (again). At the first one, give each robot
one run of 10 minutes. At the second one, give each robot two runs of
10 minutes. (Many robots stop early, so it shouldn't be too boring.)
The robot's score is best-of-three. That would give preliminary standings
at the first competition, and give people a reason to work hard for the
second one. It would also ensure the second one is good enough to invite
local media. Finally, including the first competition's score in the
best-of-three would ensure that people would have incentive to have their
robot ready for the first one; otherwise they miss a scoring opportunity.
- Many robots had drained batteries and had to run tethered to get
enough power (although, even tethered, there wasn't as much current
available as from full batteries). Of course people want to run their
robots all night before the competition. We should certainly allow
tethers (as we did this year). It would be even better if we could come
up with six spare batteries, and keep them fully-charged for the
competitions.
- IR and photosensors stunk. Smack!Vision helped, I hear. (A red
cellophane filter for the photoresistors). Hoods to protect from
ambient light helped; low hoods to protect from light bouncing off white
tables helped too. IR could probably be improved with a modulator(?).
It was hard to test IR; it would be nice if the lab built a simple
IR testing apparatus that could sense if your IR emitter is working.
- The one-foot rule for detecting unmapped food was silly. That only
tested how well the LED was mounted in the food. My suggestion: two
points for detecting unmapped food (only awarded once per food; and it
must be clear that the robot is not passing off lots of false positives).
Two points for actually capturing the unmapped food. The usual points
for returning the food.
- We didn't actually penalize anybody for excess beeping. In fact,
one robot got style points for musical beeping.
- Hall effect sensors broke off like antennas in a car wash.
Suggestion: Solder and glue each sensor to a fragment of prototyping
board, and solder the wire to that. Then screw the fragment to some
lego for mounting.
- Poison food actually only had one (huge) magnet on top.
That magnet was fragile and slightly dangerous. Should we consider a
new way to do poison food next year?
- We instituted a last-minute score for getting back early. However,
we realized that could be abused (if you just sat at home and won
"early" points). So we modified the rule to only take effect if you
got all four mapped nonpoisonous foods home. In retrospect, it was a clumsy
rule. A better rule might be to allow you to make as many runs as you
like in your allotted time, with your score being the best of all runs.
It would reward fast robots by giving them more opportunities to get
a higher score. Unfortunately, you'd have to be at least twice as fast
as the next robot to take advantage of it. (Well, I guess you could
force-stop a run at your discretion, hoping to do better in the remaining
time. But stopping a robot without him realizing he's done should cost
you something.)
- Nudges and bumps. We allowed you to nudge your robot up to two times,
at a cost of 2 points per nudge. This helped avoid damaging or getting
stuck robots that were off-course. (The right solution, of course, is
to have better error correction or compliancy.) It wasn't a bad rule,
but some "nudges" got pretty extreme. A nudge shouldn't be more than an
inch or a few degrees to free a stuck lego from a wall. Sliding the
robot six inches or turning it 90 degrees is beyond nudge. :v)