unsolved
Member
I'm just copying the ideas of earlier computer solvers like Thistlethwaite's algorithm, Kociemba's, Shuang Chen's, etc. where the pieces are just "partially solved" in an early stage and fully solved in the final stage; this strategy generally seems to lead to a low-ish move count.
What are the intermediate goals? I never fully understood what the objective was, other than reducing the number of moves spawned by the move generator by selectively getting certain pieces to certain locations so as to not ever have to worry about having to make those moves later.
This can be decomposed as [R' : [2D, F' D F]] (3-cycle of wings) with [D2, 3R2] (a pair of 2-cycles of midges) inserted to cancel one move.
It's amazing what brute force will uncover without any intelligence behind it though!
I coded the third phase over the past two days, and I expect to have the full solver done in a few more days. I'll describe it in more detail if my five-phase reduction algorithm doesn't turn out to be fatally flawed.
I'll have to take a look into these multi-phase reduction algorithms at one point.