Visual Gridworld
收藏arXiv2025-09-30 收录
下载链接：
https://github.com/samlobel/CFN
下载链接
链接失效反馈官方服务：
资源简介：
该数据集是一个基于网格的环境，其中智能体从左下角导航至右上角，目标是在有限的步数内到达终点，同时获得稀疏奖励。观察数据由84x84像素的图像组成，智能体的表现将根据其对访问次数预测的准确性进行评估。该数据集的规模为42x42的网格大小，任务是基于计数的强化学习探索。
This dataset is a grid-based environment where an AI Agent navigates from the bottom-left corner to the top-right corner. The objective is to reach the destination within a constrained step count while receiving sparse rewards. All observations are 84x84 pixel images. The performance of the AI Agent is evaluated based on the accuracy of its predicted visit counts. This dataset uses a 42x42 grid, and the task is a counting-based reinforcement learning exploration task.
提供机构：
Visual Gridworld
