灵巧作业臂-手机器人的力觉模态数据集

本数据集由“灵巧作业臂–手机器人技能学习与自主发育研究”项目在双 UR5 机械臂平台上实地采集，重点记录机器人在多场景、多任务执行过程中的真实交互力信息。每个机械臂的左右手腕均安装 Hypersen六维力传感器，采集并同步记录Fx、Fy、Fz与Mx、My、Mz 六维力/力矩数据。数据覆盖接近、触碰、抓取、拖拽、移动及放置等关键阶段，并与视觉数据通过统一时间戳与手眼标定保持严格对齐。数据集经过滤波、异常剔除与基线校正等处理，保证力觉曲线连续、稳定、可复现。该数据集适用于接触事件检测、抓取稳定性分析、力控制策略学习以及视觉–力觉融合等机器人智能研究任务。This dataset was collected on a dual UR5 robotic arm platform through the project titled "Research on Robotic Skill Learning and Autonomous Development for Dexterous Manipulator Arms and Hands", which focuses on recording real-world interaction force data of the robot during the execution of multiple tasks across various scenarios. Hypersen 6-axis force/torque sensors are mounted on the left and right wrists of each robotic arm, collecting and synchronously recording six-dimensional force/torque data encompassing Fx, Fy, Fz, Mx, My, and Mz. The dataset covers key operational stages including approaching, touching, grasping, dragging, moving, and placing, and is strictly aligned with visual data via unified timestamps and hand-eye calibration. Preprocessing operations such as filtering, outlier elimination, and baseline correction have been performed on the dataset, ensuring that the force-sensing curves are continuous, stable, and reproducible. This dataset is suitable for robotic intelligence research tasks such as contact event detection, grasping stability analysis, force control strategy learning, and visual-force fusion.

搜集汇总

背景与挑战

背景概述

该数据集由清华大学深圳研究生院在双UR5机械臂平台上采集，重点记录了机器人在多场景任务执行中的六维力/力矩交互数据，并经过滤波和校正处理。它适用于机器人智能研究中的接触事件检测、抓取稳定性分析和力控制策略学习等任务。