SACSoN

SACSoN 数据集由加州大学伯克利分校的研究团队创建，专注于机器人在复杂室内环境中与人类交互的自主导航任务。该数据集包含在多个办公环境中采集的大量数据，旨在促进机器人社交导航策略的学习与优化。数据集包含超过 75 小时的机器人导航数据，覆盖 5 个不同的办公环境，记录了机器人在执行导航任务时的视觉观察、激光雷达扫描、速度信息和碰撞信号等数据。数据来源于真实的机器人在办公环境中与人类的自主交互，涉及超过 4000 次丰富的人类-机器人交互。The SACSoN dataset was created by a research team at the University of California, Berkeley, focusing on the autonomous navigation task of robots interacting with humans in complex indoor environments. This dataset contains a large amount of data collected across multiple office environments, aiming to facilitate the learning and optimization of robotic social navigation strategies. It includes over 75 hours of robotic navigation data covering 5 distinct office environments, and records various types of data such as the robot's visual observations, LiDAR scans, velocity information and collision signals when the robot performs navigation tasks. The data is sourced from real autonomous interactions between robots and humans in office settings, involving more than 4,000 rich human-robot interaction instances.

搜集汇总

背景与挑战

背景概述

SACSoN数据集是一个用于社交导航策略训练的大规模数据集，包含丰富的自主机器人导航轨迹和人机交互数据，旨在实现机器人在人类环境中的非干扰性导航。