Minigrid

该数据集是MiniGrid-SimpleCrossing环境的一部分，要求强化学习代理在处理更为复杂的观察空间时，能够在一个网格中导航并达到目标，同时管理多种动作。观察值是一个包含147个维度的向量，代理可以从六个动作中选择，奖励系统基于达到目标，同时会对所采取的步数进行惩罚。这一任务属于强化学习领域。

This dataset is part of the MiniGrid-SimpleCrossing environment. It requires reinforcement learning agents to navigate in a gridworld and reach the target while managing multiple actions when facing a more complex observation space. The observation is a 147-dimensional vector, and the agent can select from six actions. The reward system is based on reaching the target, with penalties imposed for each step taken. This task falls within the field of reinforcement learning.