On Mar 14, 2007, at 6:01 AM, Lisa Meeden wrote:
Belinda,
Years ago when I did a lab like this with real lego robots, I used a
stop watch and timed all the robots from several different starting
points.
Lately, I haven't been so careful about trying to compare their
performance in a quantitative way. Instead, the students typically
agree which behavior is best and are actually more interested in
talking about how the various approaches work.
Lisa
Lisa,
I agree that a more qualitative approach is probably the way to go.
I really like the lab, perhaps in part b/c I've (outside of
commercial control systems venues) never done any "real" robotics
programming. The simulator, as you already know, just makes the
concepts so much more FUN.
My students are already having fun trying to write "wall following"
behaviors and one noted that avoid isn't even needed given how she's
written her wall following code.
I'm sure there are many fancy algorithms that use line fitting to
estimate what's a wall. For now, I've tried various games with
statistics (e.g. average and variance of front left, sampled five
times consecutively) that produces some interesting behavior (I'm
currently only using the size of the variance to estimate walls). My
students are now clamoring for more high-level reporting
capabilities. I now can report number of steps from start to found
light, but they'd like stats on how many times each behaviors been
active. This got me to thinking:
How hard would it be to change the color of the trace that's drawn so
it (roughly) indicates which behaviors were controlling which line
segments?
Or, how easy would it be to have a callback of some sort reset the
stats when a user drags the robot to a new location with the mouse?
I have some other high-level Pyro questions too, but I'll include
that in a different email.
--b
_______________________________________________
Pyro-users mailing list
[email protected]
http://emergent.brynmawr.edu/mailman/listinfo/pyro-users
