Jackal Dataset

Clearpath Robotics的Jackal UGV被用作数据收集平台。这款滑移转向四轮驱动车辆配备了车载IMU，两个带有编码器的直流电动机 (可测量车轮角速度) 以及电流传感器 (可测量电动机电流输出)。在机器人的每一侧，前轮和后轮都与变速箱连接，因此以相同的速度和方向一起旋转。IMU根据欧拉角以及车身在三个欧几里得轴上的线性加速度和角速度提供了车辆姿态测量。诸如RS D435i和RS T265之类的相机系统安装在连接到机器人平台顶部的铝制框架上。当跟踪摄像机T265朝前时，D435i深度摄像机的定位和倾斜方式使摄像机可以清晰地看到地形补丁。贴片尺寸为〜 “680毫米 × 340毫米”，相对于车辆底盘的前方距离为150毫米。D435i用作本研究的常规RGB相机，并且D435i重建的深度图像未包含在任何数据集中。T265摄像机被用作视觉惯性里程计 (VIO) 解决方案，可提供车辆的自我运动估计。

记录了六个不同的传感器源: 1) 电流反馈，2) 来自车辆两侧的车轮编码器读数，3) 来自T265的6个DoF VIO测量，4) 三轴线性加速度，5) 来自IMU的姿态测量，以及6) D435i拍摄的RGB图像。除RGB图像外，所有其他传感器信号均用作本体感受功能。调查了七个地形类别，包括沥青，砖路，草，砾石，路面，沙子和涂层地板。

有关更多详细信息，请参阅我们的手稿: https://ieeexplore.ieee.org/document/9507312

The Jackal UGV from Clearpath Robotics was utilized as the data collection platform. This skid-steering four-wheel drive vehicle is equipped with an on-board inertial measurement unit (IMU), two DC motors with encoders (for measuring wheel angular velocity), and current sensors (for measuring motor current output). On each side of the robot, the front and rear wheels are connected to the gearbox, so they rotate simultaneously at the same speed and direction. The IMU provides vehicle attitude measurements via Euler angles, as well as linear acceleration and angular velocity of the vehicle body along the three Euclidean axes.

Camera systems including the RealSense D435i and RealSense T265 are mounted on an aluminum frame affixed to the top of the robotic platform. With the tracking camera T265 oriented forward, the D435i depth camera is positioned and tilted to enable clear capture of terrain patches. Each patch has dimensions of approximately 680 mm × 340 mm, and is located 150 mm in front of the vehicle chassis. The D435i functions as a standard RGB camera for this study, and the depth images reconstructed by the D435i are not included in any of the datasets. The T265 camera is employed as a visual-inertial odometry (VIO) solution to deliver ego-motion estimation of the vehicle.

Six distinct sensor sources are recorded in total: 1) current feedback, 2) wheel encoder readings from both sides of the vehicle, 3) 6-degree-of-freedom (6-DoF) VIO measurements from the T265, 4) triaxial linear acceleration, 5) attitude measurements from the IMU, and 6) RGB images captured by the D435i. With the exception of RGB images, all other sensor signals are utilized as proprioceptive measurements. Seven terrain categories are examined in this work, including asphalt, brick pavement, grass, gravel, paved road, sand, and coated floor.

For additional details, please refer to our manuscript: https://ieeexplore.ieee.org/document/9507312

搜集汇总

背景与挑战

背景概述

Jackal Dataset是一个用于无人地面车辆（UGV）地形分类的数据集，基于Clearpath Robotics的Jackal UGV平台收集。它整合了多种传感器数据，包括IMU、车轮编码器、电流传感器、视觉惯性里程计（T265）和RGB相机（D435i），覆盖七种不同地形类别（如沥青、草、沙等），旨在支持视觉-本体感受融合方法的研究。数据集发布于2021年，由伊利诺伊大学厄巴纳香槟分校提供，适用于机器人学和自动化领域的应用。