Experiment
1.	Hardware

1.1   Fire-Fighting Robot Contest Maze at Trinity
  See document:  arena.gif
We arrived at Trinity Saturday, the afternoon before the contest.  
There were 3 mazes set up for testing, similar to ours, but
to our surprise, the contest rules had changed at the last minute!  
Due to the overwhelming number of contestants, we now had to 
succesfully complete a pre-qualifying run in the maze to be allowed 
to compete on Sunday.  We set our robot "Willitrun" into their 
maze to try it out, but all it did was veer into walls.  
Luckily we brought extension cords, tools, DMM's, a scope, and lots 
of parts!   Testing on the bench showed that our object sensor 
sensitivity had somehow increased to double of what it was at 
Ward College.  Why?  Trinity had high pressure sodium overhead 
lighting, strictly a DC light source (not modulated by a ballast).  
Since our sensors were no longer swamped by 40 KHz flourescent 
ballasts, their sensitivity leaped!  Since our algorithm was to 
meander towards the open space in the middle of hallways, it 
lost control when all sensors saw walls everywhere all the time!  
We increased R27 to 470 ohms (less IR LED intensity) which helped, 
but we had another problem:  The sodium lighting swamped our flame 
sensors.  Luckily we had some Kodak 87C IR lowpass filter 
material with us.  This material passes 900-1000 nM and longer 
wavelength IR light but rejects all visible light under 800 nM. 
  See document:  kodak87c.gif
We glued small pieces of it over all our sensors,  the 
flame sensors and the object proximity sensors.  After hours 
of fine tuning, we finally got "Willitrun" to complete a 
qualifying run - finding and extinguishing a candle - we were in!  
  See document:  ffc20.jpg
 
1.2.1   Flame Detectors 
The home-rolled flame detectors worked so well, we had a 
couple of people in the business ask about our design.  
The Kodak filter film really helped cut the ambient light 
interference.  The modified AWG#10 splicing connector housings 
provided about a 30 degree aperture, wide enough to sense a  
flame by passing by it in motion, and narrow enough to 
zoom in on it off the front sensor.  

1.2.2   Fan 
Lesson learned: Too much experimenting can be bad!  
After seeing the overkill some of the other robots had for 
fan horsepower, we decide to wire in another D-cell in series 
with our other two for more windpower.  It worked - too well.  
In our first official run in the actual contest, Willitrun 
quickly found the candle in the 4th room, turned towards it, 
gave the fan a short test burst and blew out the candle 
instantly.  The crowd cheered!  The judge got out his ruler, 
our run was disqualified because we now could blow out the 
candle from 13" away - the rules say it must be within 12". 
If we left it alone, our algorithm would have seen the 
test burst leave the candle still lit, and would have moved 
in a few inches further before trying again, which would have 
made for a perfect run.   


See the next section "Optimize" for ideas on how this robot 
could be improved for next years' contest.

Back to Robot Main Page