Ant environments
收藏arXiv2025-09-30 收录
下载链接：
https://github.com/cosynus-lix/STAR
下载链接
链接失效反馈官方服务：
资源简介：
该数据集专为评估目标表示算法而设计，包含了一系列在模拟四足机器人环境中具有挑战性的任务。这些任务具有分层动态特性，需要不同的状态表示以达到最优性能。数据集在30维状态空间和8维动作空间中进行扩展，任务内容聚焦于复杂连续控制任务中的分层强化学习（HRL）。
This dataset is specifically designed for evaluating objective representation algorithms. It contains a series of challenging tasks in simulated quadruped robot environments. These tasks possess hierarchical dynamic properties, requiring distinct state representations to achieve optimal performance. The dataset spans a 30-dimensional state space and an 8-dimensional action space, with its tasks focusing on hierarchical reinforcement learning (HRL) in complex continuous control tasks.
提供机构：
OpenAI Gym (simulated environments)
