StarCraft Micromanagement Scenarios

该数据集包含了《星际争霸》中多种微操场景，其中训练有素的巨兽对抗不同的敌方单位，包括狂热者和跳虫。这些场景被分为小规模和大规模微操两类。此外，数据集还包括了性能指标，如训练期间的胜率、回合步数和平均奖励。该数据集涵盖了从小规模到大规模场景的多种尺度，其任务是研究在《星际争霸》中强化学习在微操策略上的应用。

This dataset comprises multiple micro-management scenarios from *StarCraft*, where well-trained behemoth units engage against various enemy units including Zealots and Zerglings. These scenarios are categorized into two groups: small-scale and large-scale micro-management operations. Additionally, the dataset includes performance metrics such as win rate, episode steps, and average reward during training. Covering a broad spectrum of scales from small-scale to large-scale scenarios, this dataset serves as a testbed for researching the application of reinforcement learning to micro-management strategies in StarCraft.