Bipedal Robot Locomotion Dataset

该数据集包含了从两足机器人在不同地形上行走时收集的观察-动作对，这些地形包括平坦地面、5.7度和10.2度倾斜的斜坡，以及未知地形，如崎岖地形和离散地形。数据集还包括了之前控制时间点的状态，并覆盖了不同的速度和斜坡，为在不同条件下扩散策略的泛化能力提供了洞见。该数据集规模为500对，其中每个速度（0.3米/秒和1米/秒）各包含250对，其任务聚焦于腿部运动控制与策略学习。

This dataset consists of observation-action pairs collected during a bipedal robot's walking on diverse terrains, including flat ground, slopes with 5.7-degree and 10.2-degree inclines, and unknown terrains such as rough terrain and discrete terrain. The dataset also incorporates states from prior control timesteps, covers a variety of walking speeds and slope configurations, and offers insights into the generalization capability of diffusion policies across different conditions. The dataset contains 500 pairs in total, with 250 pairs for each of the two tested speeds (0.3 m/s and 1 m/s). The targeted task focuses on legged locomotion control and policy learning.