StarCraft II Micromanagement Benchmark

该数据集是一个专为《星际争霸II》环境设计的去中心化协作基准，用于评估多代理强化学习算法。它包含了多种微管理场景，如2c_vs_64zg和MMM2，常被用于评估最先进的多代理强化学习方法。在这个数据集中，多个代理控制单个单位，并共享奖励机制，任务是在协作环境下进行多代理强化学习。

This dataset is a decentralized collaboration benchmark specifically designed for the StarCraft II environment, aimed at evaluating multi-agent reinforcement learning algorithms. It includes various micro-management scenarios such as 2c_vs_64zg and MMM2, which are commonly utilized to assess state-of-the-art multi-agent reinforcement learning methods. In this dataset, multiple agents control individual units and share a common reward function, with the core task being to conduct multi-agent reinforcement learning in a collaborative environment.