From reid@math.berkeley.edu Sun May 17 22:10:48 1992
Return-Path:
Received: from math.berkeley.edu by life.ai.mit.edu (4.1/AI-4.10) id AA20296; Sun, 17 May 92 22:10:48 EDT
Received: from phnom-penh.berkeley.edu.berkeley.edu by math.berkeley.edu (4.1/1.33(math))
id AA08289; Sun, 17 May 92 19:10:33 PDT
Date: Sun, 17 May 92 19:10:33 PDT
From: reid@math.berkeley.edu (michael reid)
Message-Id: <9205180210.AA08289@math.berkeley.edu>
To: Dik.Winter@cwi.nl, cube-lovers@life.ai.mit.edu, keng@zcar.asd.sgi.com
Subject: Re: My program is too fast ;-).
stop it, you're killing me! i also have the same idea for combining
the "coordinates" in pairs, but i'm not getting too far implementing
it. :-( i wouldn't suggest using singmaster's notes for pattern
maneuvers. have you seen bandelow's book? it has very short
maneuvers for most patterns, including two different ones for
"walker's worm" in 14 turns (assuming i've got the right pattern in
mind). bandelow counts "slice turns" as one move, though, so
his maneuver for 6X (order 3) is 24 face turns.
what amazes me about this whole business is that the algorithm
finds very short moves when they exist. i would have expected
the program to produce maneuvers of approximately the same
length for all patterns. i would say that this is a major step
forward.
you'll probably be swamped with patterns to test, but here's a
couple:
stripes: (18) F3 U1 F2 U3 R1 B2 R3 U1 F2 L2 U3 L1 B2 L3 U1 L2 U3 F1.
python: (15) R1 U3 F3 B1 L1 F2 L3 F1 B3 U3 R3 L1 F2 U2 L3.
since i found these by hand, i'm curious to see how close they
are to optimal. hopefully i'll have my program running soon.
mike