Why so long?
Surely this is less AI and more well, physics calculations. (Disclaimer: not looked into it more than this article)
If you know the values of breaking the things on screen and the locations of the required objects (image analysis) then how is this AI? Someone has told it what those values were and it calculates the exact trajectories based on those values.
Or is it trying the level multiple times to see what works and what doesn't? If it is doing this why does the article not mention it?