Solving the Rubik s Cube with Approximate Policy Iteration