HumanoidBench
收藏资源简介:
HumanoidBench是由加州大学伯克利分校开发的一个模拟人形机器人基准数据集,包含15个全身操作任务和12个移动任务。该数据集旨在通过模拟环境加速人形机器人的算法研究,解决复杂动态控制、身体各部分协调和长期复杂任务等问题。数据集包含多种任务,如货架重新排列、包裹卸载和迷宫导航,适用于评估和推动人形机器人在多样化环境中的灵活性和适应性。
HumanoidBench is a simulated humanoid robot benchmark dataset developed by the University of California, Berkeley. It includes 15 full-body manipulation tasks and 12 locomotion tasks. This dataset aims to accelerate algorithmic research on humanoid robots within simulated environments, addressing core challenges such as complex dynamic control, coordinated movement of individual body segments, and long-duration complex tasks. The dataset covers diverse tasks including shelf rearrangement, package unloading, and maze navigation, and is suitable for evaluating and advancing the flexibility and adaptability of humanoid robots across varied environments.




