1. Re: Pixel Bot Wars -Reply -Reply
C & KL-
>>> C & K L <candk at TICNET.COM> 11/06/98 06:54pm >>>
>I don't quite completely comprehend the idea of using the "least real
>processing time." (see below for more)
>Okay, so a slower algoritm on a faster computer would beat your fast
>algoritm on a slow computer? Is that what you're saying?
No, not at all. What I am saying is that there are two different ideas to
consider. First of all, if you simply implement an algorithm directing a pixel
from start to finish, you can time it by using the time() function. This is
"real" processing time. This is a measure of the actual time the CPU took
to allow that algorithm to solve the maze.
The second one is the idea that in a "real" robotic situation, if I created a
mouse robot with a fast little motor that followed walls, and you created
a dog-sized robot that analyzed the position of all the walls in its sensor
range before making a movement, my little wall-follower would whomp(!)
your big dog. The processing power (real) required to analyze positions
and plot goal-directed behavior for a mobile robot is astounding. This is
why Rodney Brooks subsumption architecture made such a splash back
in the early 90's. Goal-seeking behaviors could be evolved instead of
programmed in, which made the resulting computations much much much
quicker. I believe it was Shakey back in the 70s that was a robotic arm
designed to balance and analyze a room full of geometric forms. It
always broke when they took it out of the simulator and put it in the real
world. So, my point was that if we are really simulating robots, if I
implement a simple wall-follower with no defense or offense, it has to
appear to move faster than a robot with 4 lasers, 20 points of armor, and
16 sensors.
-Jay
