Very nice! Nice to see demonstrations of reinforcement learning being used to solve non-trivial tasks.