These are all complex systems where figuring out how to simulate them is a much difficult task than even reinventing reinforcement learning from scratch.