novastar112/visgym_pacman_2d_remap

该数据集包含用于自定义VisGym Pacman2D环境的行为克隆轨迹。每个轨迹代表一个成功的oracle episode，包含`image_prev`、`image`和`image_next`等历史条目，图像是以JPEG base64字符串形式存储的。环境设置包括不同的网格大小和难度级别（Easy、Hard、Ultrahard），以及实体（Pacman、食物、墙壁和一个确定性幽灵）。幽灵通过BFS距离追逐Pacman，动作空间包括四个方向和停止动作。数据集分为训练和测试集，且初始状态通过`init_state_hash`确保不重叠。此外，README还提到了一个remap变体，该变体重写了历史步骤中的图像帧。

This dataset contains behavior-cloning trajectories for the custom VisGym Pacman2D environment. Each row is one successful oracle episode. The history entries include `image_prev`, `image`, and `image_next`; no synthetic thinking traces are stored. Images are JPEG base64 strings rendered from the greyblue9/pacman-python visual assets used by the environment. Environment summary includes different grid sizes and difficulty levels (Easy, Hard, Ultrahard), and entities (Pacman, sparse food, walls, and exactly one deterministic ghost). The ghost chases Pacman by BFS distance with fixed tie-break order up, down, left, right. The movement action space is four directions; VisGym records can also end with `(stop, stop)`. Splits are provided for train and test sets, and initial states are checked to be disjoint by `init_state_hash`. Additionally, a remap variant is mentioned that rewrites each history step so `image_prev` is the same rendered frame as `image`.