Randomly Generated Environments

该数据集包含了1000万个随机生成的环境，其中方块的数量从0到60不等，这些数据被用于训练基于扩散的环境生成器。此外，该数据集还用于深度强化学习中的智能体训练，以评估它们在具有挑战性环境中的鲁棒性。规模达到了1000万个环境，任务包括部分可观察导航任务和二维双足行走任务。

This dataset comprises 10 million randomly generated environments, with the number of blocks ranging from 0 to 60. It is utilized to train diffusion-based environment generators. Furthermore, this dataset is also employed for agent training in deep reinforcement learning to evaluate the robustness of agents in challenging environments. Covering 10 million environments, the dataset includes two types of tasks: partially observable navigation tasks and 2D bipedal walking tasks.