Hamiltonian circuit for the entire 2x2x2 cube group

Discussion in 'Puzzle Theory' started by cuBerBruce, Dec 26, 2011.

1. StefanMember

7,287
27
May 7, 2006
WCA:
2003POCH01
StefanPochmann
But if the algorithm is very long?

2. qqwrefMember

7,830
39
Dec 18, 2007
a <script> tag near you
WCA:
2006GOTT01
qqwref2
If you only check the position after each run of the algorithm, it doesn't matter how long it is.

3. StefanMember

7,287
27
May 7, 2006
WCA:
2003POCH01
StefanPochmann
Ah yes, I misunderstood, sorry. Probably didn't realize what you meant (and what I now see he had said) because I already knew it's impossible

Last edited: Dec 30, 2011
4. Lucas GarronSuper-Duper ModeratorStaff Member

A couple of letters in the algorithm are "G", but only "g" is defined. Are these intended to be the same?

5. qqwrefMember

7,830
39
Dec 18, 2007
a <script> tag near you
WCA:
2006GOTT01
qqwref2
g=ncabVR

6. Lucas GarronSuper-Duper ModeratorStaff Member

Oh, I'm silly. I was looking at the algs so hard that I forgot the {U, V, R, S, F, G} definitions at the top (which I only read the first time I saw this thread).

Mm-hmm. Hence my confusion, because I only saw that, but not the definition of G.

7. qqwrefMember

7,830
39
Dec 18, 2007
a <script> tag near you
WCA:
2006GOTT01
qqwref2
Oh, sorry. I misread that pretty hard, and assumed that G must have been the one you'd seen, since it was right there at the top. Never mind. (It WAS kinda weird to use S, V, and G - considering that other sequences are inverted throughout the definitions...)

Any feedback on my bounds and/or idea for generating a full 3x3 Hamiltonian path? Is the idea wrong, or right, or potentially useful?

PS, a little rewriting to (hopefully) make things slightly easier to understand:
Code:
P=UR
Q=U'R

a=P^5
b=P^3
c=Q P^14 Q
d=Q P^5 Q P^5
e=Q P^5 c P
f=P c P^5 Q
g=Q P^5 c P^8 Q
h=Q P^8 c P^5 Q
i=P^7
j=cc P^5 Q
k=P^6 Q
l=Q P^5 cc
m=Q P^6
n=Q P^5
o=Q P

s=nbgaQdblinQeckdbcanbQmcPmcidoP
t=PPnUUaPdQihkhcecncabdQdcPfPPhaQdlfabnbQdeckcfQigkPdPcfccaPdQdejhhejaoPdcPPefaQe
ijaQhcQijaQnbefaodPPglfaQeckdcPfabQmhidnUUaPdeefknbhaodcPfccncabdodcPjabQdccfcgdc
dbcanbQmcPmciddUUaPdQihkhcecncabdQdcPfPPhaQdlfabnbQdeckcfQigkPdPcfccaPdQdejhhejao
aPdnQdmccQijaQhcQijaQnbefaodPPglfaQeckdcPfabQmhidnUUaPdeefknbhaodcPfccncabdodcPja
QdblinQeckdbcanbQmcPmciddUUaPdQihkhcecncabdQdcPfPPhaQdlfabnbQdeckcfQigkPdPcfccaPd
cPfPPlPQicaPdnQdmc
w=cQijaQhcQijaQnbefaodPPglfaQeckdcPfabQmhidnU

u=krQsPtwF
v=u^8 aQrQsPtwUUF'
p=FU'w't'R'U's'RRRUR'R'UR'U'b'R'FRbPQrU'R'R'R'sPtwUF'
q=F'U'w't'R'U's'R'UR'R'Ua'R'FU'U'w't'R'U's'R'UR'R'Ua'FRrQsUR'R'R'twUaF'
x=krQsPtcQijaQhcQijaQnbefaodPPglfaQnQUpbPUpbPUpbockdcPfabQQkaboPUpbPUpbPUpPdPPdnU
FkrQsPtQUpaabQQijaQhcQicnaUpbdQnbefaodPPgnoPUpbPUpbPUpPQcfaQeckU'paQUpbPcPPcPUpbn
bQQkabnbPUpbPdPPdnUFaQrQsPtQUpaabQQijaQhcQijaQnbePoUpaidodPPglfaQenbPUqbPQknQUpbP
nabPU'pPfabQQkPUpUpbUpUpcUpPUpUpPQbUpbQbPUpQUpbPoUpbUUUF'

z=v^6 u^6 x

Last edited: Dec 30, 2011
8. cuBerBruceMember

914
14
Oct 8, 2006
Malden, MA, USA
WCA:
2006NORS01
cuBerBruce
I thought (as far as my description is concerned) I would reserve capital letters for single moves and lower case letters for sequences. To try to make parsing it by a computer program as simple as I could, I avoided using exponentiation or other form of repetition counts. However, using an inverse operator avoided having to define a lot more variables/sequences, and I had used almost all the lower case letters already.

Last edited: Dec 30, 2011
9. stannicMember

120
2
Jun 16, 2010
The country of riddles
If each of two undirected graphs G1 and G2 have a Hamiltonian path then their Cartesian product also has a Hamiltonian path. We can draw |G2| copies of G1 and add edges between copies to get Cartesian product. On all copies of G1, we will mark the same Hamiltonian path. Then we can connect all these paths into one Hamiltonian path.

Is it true that the cube group is the Cartesian product of G and H?

Let A=<U,D,R,L,F,B>, B=<U,D,R2,L2,F2,B2>. Let C be the group that arises when we merge positions in A that differ by permutation in B.
Then |A| = 4.32*10^19, |B| = 1.95*10^10, |C|=2.217*10^9.
Suppose there is a Hamiltonian path on graph {B, <U,D,R2,L2,F2,B2>}, and there is a Hamiltonian path on {C, <R,L,F,B>}. Can we conclude that there is a Hamiltonian path on {A, <U,D,R,L,F,B>}?
Is this interpretation correct?

10. qqwrefMember

7,830
39
Dec 18, 2007
a <script> tag near you
WCA:
2006GOTT01
qqwref2
Yeah, that's what I was thinking too.

That's what I was going for. I'm not completely sure it's mathematically valid since it's been a while since I did any serious group theory, but if it does work, the relative sizes of these subgroups should make it a lot easier to find a Hamiltonian path.

11. stannicMember

120
2
Jun 16, 2010
The country of riddles