MuSoHu

MuSoHu是一个大规模、多模态、以自我为中心的社交人类导航数据集，由乔治梅森大学创建。该数据集涵盖约100公里、20小时的社交合规导航演示，由13名人类演示者在多种公共空间中收集，包括室内外环境。数据集通过穿戴式传感器套件收集，包括3D LiDAR、立体和深度相机、IMU、麦克风阵列和360°相机，旨在为机器人学习提供高质量的自然社交导航数据。MuSoHu的应用领域包括机器人导航行为的模仿学习，以及研究社交机器人导航的多样性和文化依赖性。

MuSoHu is a large-scale, multimodal, egocentric social human navigation dataset created by George Mason University. This dataset encompasses approximately 100 kilometers and 20 hours of socially compliant navigation demonstrations, collected by 13 human demonstrators across various public spaces including both indoor and outdoor environments. The dataset is collected via a wearable sensor suite comprising 3D LiDAR, stereo and depth cameras, IMU, microphone array, and 360° camera, aiming to provide high-quality natural social navigation data for robotic learning. Application areas of MuSoHu include imitation learning of robotic navigation behaviors, as well as research on the diversity and cultural dependency of social robot navigation.

搜集汇总

背景与挑战

背景概述

MuSoHu是一个大规模、多模态、以自我为中心的社交人类导航数据集，由乔治梅森大学创建，涵盖约100公里、20小时的社交合规导航演示，在室内外环境中收集。它通过穿戴式传感器套件（如3D LiDAR和相机）提供高质量数据，旨在支持机器人导航的模仿学习和社交导航的多样性与文化依赖性研究。