[Icfp04-discuss] Ant optimizers?

Yutaka OIWA oiwa at yl.is.s.u-tokyo.ac.jp
Fri Jun 11 05:55:13 EDT 2004


>> On Thu, 10 Jun 2004 12:14:16 -0700, David Brown <icfp04 at davidb.org> said:

David>    n+1   flip 1   target target

This kind of instructions (Goto or Nop) seem to be useless,
but actually they are useful in some cases to control
the speed of ants.

At the beginning stage of the game before ants hits foes, foods
or rocks, the states of the ants can be considered deterministic.
Therefore, we can program each ants to detect her relative position
in the anthill. Using this, my program assigns a special task for the
six ants in the corners.  However, sometimes their job is unfortunately
obstructed by their friends, because they walk slower than other
random friends. To overcome this problem, I implemented another special
routine for ants on the edges of anthill and makes them walk slightly
slower than they can do. For this purpose I use
"Sense Here next next Friend" instructions as a "nop".
Of course this delayed routine is only used in the beginning.

-- 
Yutaka Oiwa              Yonezawa Lab., Dept. of Computer Science,
      Graduate School of Information Sci. & Tech., Univ. of Tokyo.
                    <oiwa at yl.is.s.u-tokyo.ac.jp>, <yutaka at oiwa.jp>
PGP fingerprint = C9 8D 5C B8 86 ED D8 07  EA 59 34 D8 F4 65 53 61


More information about the Icfp04-discuss mailing list