The discussion of local maxima for the Q measure of distance led to an informal
use of symmetry. It is not clear to me just what symmetry is needed to carry
through the maxima argument but I suggest the following is sufficient (although
perhaps too restrictive).
Let C by the rotation group of the cube (closure of IJK: order 24)
Let G be Rubik's group (closure of UDRLFB: order 10^19 or so)
Both groups can be represented as a permutation group on [0, 1, ...53] for some
arbitrary numbering of the 54 faces. We can also use the names UDRLFB for the
six colors; where the association is made once and for all for any given physical
puzzle. Like U=red, F=blue, etc.).
The elements of g are 1-1 with the observable configurations of the standard
cube; and in fact are the recipes to reach the configurations from "home". g' is
the "solution" that returns g to home.
The elements of G*C are also 1-1 with the observable configurations except now
the correspondence must also take into account the observed orientation of the
cube.
Each g in G is represented by a permutation of the cubelet faces. Each face in g
is a fixed color.
For color X, let X[g] be the set of faces of g colored X. |X[g]| = 9. Let Coloring[g] = {U[g], D[g], R[g], L[g], F[g], B[g]}.
Then g is totally symmetric if for all c in C, Coloring[gc] = Coloring[g].
----
It is true that "home" and UUDDRRLLFFBB are totally symmetric by this
definition. "home" is a minimum (special case). UUDDRRLLFFBB is a local
maximum.
Questions:
Is there a simpler equivalent definition?
How many totally symmetric configurations are there?
Is there a less restrictive definition that guarantees local maxima?