Research on the StarCraftll Decision Method Based on Hierarchical Reinforcement Learning
Liang Chen, Tong Liu, Yunting Liu · 2020
In order to solve the problem of "dimensional disaster" in traditional reinforcement learning algorithms when making decisions in a complex environment, this paper proposes a decision-making algorithm based on hierarchical reinforcement learning. Select the famous real-time strategy game (RTS) StarCraft II (SCII) as the research environment. The game contains a lot of strategies, these strategies affect the trend of the game, at the same time strategies change over time and as circumstances change. In order to solve the problem of decision-making difficulties caused by the huge strategy space and state space during the experiment, state space and strategy space are treated in different hierarchical manners by using hierachical reinforcement learning. It is verified by experiments that agents using hierarchical task structures have achieved good results in solving high-dimensional decision-making problems.