MiniCrafter

该数据集是为了测试继任特征（SFs）在不同场景下的处理能力而设计的，它包含了 MiniCrafter 这款游戏，游戏中有关于资源收集、制作以及陷阱等多种任务。游戏场景分为单一物品、两种物品、随机、随机惩罚以及制作任务等不同类别，为强化学习智能体提供了多样化的挑战。该数据集在一个12x12的网格中包含了各种对象类型和资源，任务重点在于强化学习中的多任务迁移学习。

This dataset is developed to evaluate the performance of Successor Features (SFs) when handling tasks across various scenarios. It incorporates the game MiniCrafter, which encompasses multiple types of tasks including resource collection, crafting, and trap-related missions. The in-game scenarios are categorized into distinct classes: single-item tasks, two-item tasks, random tasks, random penalty tasks, and crafting tasks, providing diverse challenges for reinforcement learning agents. This dataset contains various object types and resources within a 12×12 grid, with its core task focus on multi-task transfer learning in reinforcement learning.