Motivated reinforcement learning for adaptive characters in open-ended simulation games