Flatland Competition Dataset

该数据集涵盖了Flatland竞赛中使用的各种环境，旨在评估在网格世界设定下车辆重调度问题（VRSP）的解决方案。此外，该数据集支持评估多种算法和训练调度的方法，包括集中式和分布式方法。规模上，它包含了数百个代理（即列车）在各种环境中的数据。任务方面，专注于多代理强化学习（MARL）在列车协调中的应用。

This dataset encompasses a variety of environments used in the Flatland competition, designed to evaluate solutions to the Vehicle Rescheduling Problem (VRSP) in a grid-world setting. Furthermore, it supports the evaluation of diverse algorithm and training scheduling strategies, including both centralized and distributed approaches. In terms of scale, the dataset contains data from hundreds of agents (i.e., trains) across various environments. For the task scope, it focuses on the application of Multi-Agent Reinforcement Learning (MARL) to train coordination.