Multi-Agent Particle Environments Benchmark Suite
收藏arXiv2025-09-30 收录
下载链接：
https://github.com/schroederdewitt/multiagent-particle-envs
下载链接
链接失效反馈官方服务：
资源简介：
该数据集包含了在2D空间中模拟的多机器人任务，包括协作导航和捕食者-猎物场景。该数据集允许测试在不进行通信的情况下智能体之间的协调能力，并包含了稀疏奖励的场景。规模上，多个智能体在各种各样的任务中互动。该数据集的任务主要针对多智能体强化学习。
This dataset contains multi-robot tasks simulated in 2D space, including cooperative navigation and predator-prey scenarios. It enables testing the coordination capability between agents without communication, and includes scenarios with sparse rewards. In terms of scale, multiple agents interact across a wide range of tasks. The tasks in this dataset are primarily designed for multi-agent reinforcement learning.
提供机构：
Multi-Agent Particle Environments
