Half field Offense (HFO)
收藏arXiv2025-09-30 收录
下载链接：
https://github.com/mhauskn/HFO/blob/master/doc/manual.pdf
下载链接
链接失效反馈官方服务：
资源简介：
该数据集是对RoboCup 2D游戏的简化抽象，特点在于混合动作空间，可以测试不同的多代理场景，包括1对2和2对1的配置。该数据集使得算法能够在具有混合动作空间的环境中接受评估，并包含了对代理观察的详细状态信息。它涉及多个具有不同角色和目标代理的规模，任务是实现在体育模拟环境中的多代理强化学习。
This dataset is a simplified abstraction of the RoboCup 2D game. It features a hybrid action space, supporting the evaluation of various multi-agent scenarios including 1v2 and 2v1 configurations. This dataset enables algorithms to be evaluated in environments with hybrid action spaces, and contains detailed state information for agent observations. It covers multiple scales involving agents with different roles and goals, with the core task of implementing multi-agent reinforcement learning in sports simulation environments.
提供机构：
RoboCup community
