Combining Reinforcement Learning with a Multi-level Abstraction Method to Design a Powerful Game AI | IEEE Conference Publication | IEEE Xplore