>Date: 5/12/2004 4:45 PM Central Daylight Time
><snip>
>>Aside from the steering controller is a throttle/brake controller. I've
>tried
>>2 different things:
<snip>
No, method 1 didn't try to figure out the optimum speed. This was for RC cars
where basically they run full steering lock in the corner, then use the
throttle to keep the line. My thinking was basically that if I apply full or
less than full steering lock and the car is having no problem following the
line, then you should increase throttle. Once it's trailing off the line,
reduce it. Nice and simple theoretically, but in the really tight corners it
slowed down too much.
Method #2 used target speeds instead and could run through the same corner much
faster.
If so, then
I played a bit with the throttle/brake gain rates. Even if the AI car reaction
time was 1/300 of a second and it had a lightning fast trigger finger, method 1
just didn't work as well.
Maybe you're right, I don't know. Method 2 was just so much easier and it
worked better straight off so I kept it. Granted, method 1 did work well
everywhere except on the tightest corners. Plus it was just cool to watch the
cars making essentially better and better lines by themselves (until they
actually got on the line I defined, that is). I'd go take a smoke break and
come back to check their laptimes. It was pretty fun and interesting to watch
because it looks like they're really learning (they were, I suppose), but the
performance just wasn't as good as letting them chase my laps around,
especially in the slowest corners.
Come to think of it, a lot of it might have been spline inaccuracy. If there's
so much as a kink anywhere the car will think it's on a tight turn and let up.
Also, the car wasn't really measuring from the center of the car, but from a
point out in front of it that moved further down the track as the car went
faster. Perhaps at low speed the point was too far forward so it thought it
was going off when it really wasn't and slowed down too much. But if I didn't
have the point out far enough, it would criss cross the line and dart around in
a pretty fake way. It was just such a tough balance to work out that I tried
the other way and it ended up working better all around, and it was much faster
to come up with the lines in the first place; more driving lines to pick from.
With 20 second lap times it doesn't take long to get 50 or so lines :-)
Todd Wasson
Racing Software
http://PerformanceSimulations.com
http://performancesimulations.com/scnshot4.htm
