Level 15 reminds me of the P=NP problem, that making a solution is much easier than finding one. While the designer can randomly input a sequence of moves and see how the robots move, then pick one that has interesting path and only passes through a cell once (box is put here); the players have to try out different combinations and logical deduction seems to fail here.
The first 14 levels are really good; making good use of the terrain and the waiting-and-colliding mechanics. Hope that there will be levels having two robots pushing two different boxes.