Our final alternative is to do simultaneous breadth first searches from both the initial state and the goal state.

Problems: