There's no "right" answer, but here's my suggestion. First, I'd start with unsegmented until you get the rest working right. In general, I wouldn't go higher than 6 dimensions, or maybe 7 if the others are small enough, because you'll simply run out of memory. And in general, I think 3 or 4 are the most crucial: LateralVelocity, distance, accel (0 = deccelerating, 1 = same speed, 2 = accelerating), and time since velocity change or WallDistance. The more segments / slices you add, the more you're trading learning speed for long-term accuracy. Play around and see what works! =) Also, welcome to the wiki! Please feel free to make a page for yourself and your bots. Cheers, -- Voidious |