Brax Environments - Humanoid, Walker2D, Halfcheetah, Ant

该数据集包含了在Brax环境中评估的政策，这些环境包括Humanoid、Walker2D、Halfcheetah和Ant。研究的重点是在稳定的前向运动的同时，尽量减少能量消耗。每个环境具有不同的维度：Humanoid为227个观察维度和17个动作维度，Walker2d为17个观察维度和6个动作维度，Halfcheetah为18个观察维度和6个动作维度，Ant为87个观察维度和8个动作维度。这四个环境具有特定的观察空间和动作空间维度。该任务的目的是评估质量多样性强化学习中的政策性能和行为多样性。

This dataset contains policies evaluated in the Brax environment suite, which includes Humanoid, Walker2D, Halfcheetah, and Ant. The study focuses on minimizing energy consumption while maintaining stable forward locomotion. Each environment has distinct dimensional specifications: Humanoid has 227 observation dimensions and 17 action dimensions; Walker2D has 17 observation dimensions and 6 action dimensions; Halfcheetah has 18 observation dimensions and 6 action dimensions; and Ant has 87 observation dimensions and 8 action dimensions. These four environments feature specific observation and action space dimensions. The objective of this task is to evaluate policy performance and behavioral diversity in quality-diversity reinforcement learning.