Soccer Game Environment
收藏arXiv2025-09-30 收录
下载链接：
https://github.com/ppwwyyxx/tensorpack
下载链接
链接失效反馈官方服务：
资源简介：
该数据集模拟了一个可控智能体在足球游戏环境中的互动，旨在评估DPIQN和DRPIQN模型的性能。足球场由32×32像素的多个RGB网格组成，智能体可以采取五种可能的动作，环境支持多种竞争与合作场景。规模上，该数据集涵盖了多场游戏（1对1和2对2），在10万集的游戏中进行了测试。任务则是针对足球游戏中的多智能体强化学习。
This dataset simulates interactions between controllable AI agents in a football game environment, designed to evaluate the performance of DPIQN and DRPIQN models. The football field is constructed using multiple 32×32 pixel RGB grids. Agents can execute five distinct possible actions, and the environment supports various competitive and cooperative scenarios. Regarding dataset scale, it encompasses multiple game types including 1v1 and 2v2 matches, and has been tested across 100,000 game episodes. The task focuses on multi-agent reinforcement learning within football game settings.
提供机构：
Authors of the paper
