Sparse-reward 3D visual navigation benchmark

该数据集是一个旨在测试元强化学习中的探索和利用策略的基准，它专注于具有高维视觉输入的复杂探索问题。该基准是目标导向的，要求探索者在面对任何潜在目标时都能发现相关信息，而不仅仅是为了单一任务。因此，这一数据集涉及的是具有高维视觉观察的复杂探索问题，其任务是关于元强化学习中的探索与利用。

This dataset is a benchmark designed to evaluate exploration and exploitation strategies in meta-reinforcement learning, focusing on complex exploration problems with high-dimensional visual inputs. This benchmark is goal-oriented, requiring the exploring agent to discover relevant information when facing any potential target, rather than only being optimized for a single task. Thus, this dataset addresses complex exploration problems with high-dimensional visual observations, with its task focused on exploration and exploitation in meta-reinforcement learning.