We introduce an innovative dataset specifically crafted for challenging natural agricultural settings, such as a greenhouse, where achieving precise c
Description
The sensorised platform used has been created at the University of Almería. It has a base of 900 x 500 x 700 mm with a load capacity of 180 kg, equipped with four straight wheels and two handles for user pushing. The sensors are mounted on the pine wood platform, taking into account the specific constraints of each sensor:
- Bumblebee BB2: The stereo camera is from Bumblebee, model BB2-08S2. It has two lenses for capturing synchronized recordings. It is connected through a FireWire IEEE 1394 interface, providing a data transmission speed of 800 Mb/s. The stored data has a resolution of 1032 x 776 pixels, with an accuracy of 80 %. The field of view is 97º horizontally and 66º vertically within a range of 0.3 m to 20 m. Recording was done at 10 Hz, with a maximum frame rate of 20 fps. This camera is placed on the front so that the images captured are ideal for obstacle identification for a robot.
- Velodyne VLP16 LiDAR: Velodyne’s LiDAR VLP16 with a maximum range of 100 m, a 360º horizontal and 30º vertical field of view is used. It is placed at the top of the runway to have the whole field of view available, except for the operator. It was recorded with a frequency of 10 Hz.
- Ouster OS0 LiDAR: Data are also collected from the Ouster model OS0 for a meaningful comparison. This is a 32-channel LiDAR with 360º horizontal and 90º vertical field of view, ranging between 0.3 m and 50 m. The data are also recorded at a frequency of 10 Hz. In addition, it has a reflectivity sensor capable of identifying objects in a wide range of light. Finally, an IMU is installed in this element to link the complete recording point cloud to a local source. This sensor will be placed in a structure 20 cm lower than the Velodyne, obstructing its view and only collecting data at 275º horizontally.
- HISTTON PC: Taking this into account, to acquire data from the sensors, the platform is equipped with a HISTTON computer featuring an Intel i7 – 8550U processor at 4 GHz, an Intel UHD 620 graphics card with 24 CUDA cores, and 32 GB of DDR4 RAM. This equipment, with an operating temperature range of 0 – 70 ºC and with a humidity range between 0 and 85 %, was chosen for the future robot implementation. It only consumes 15 W to meet environmental aspects, resulting in minimal power consumption to consider for battery capacity. This computer runs a partition with Ubuntu 20.04 for using ROS Noetic and another with Ubuntu 22.04 for using ROS2 Humble, providing the opportunity to work with both versions.
It is important to note that NovAtel GNSS – GPS, model IMU-IGM-A1, is also installed with the antenna ANTCOM 42G1215. In this case, it will be used to share the same timestamp.
The following video shows a complete process simulation of a robot. Afterward, the result of the post-processing with the MOLA algorithm is shown.
The results of the point cloud are shown separately:
a) Pointcloud Ouster OS0
b) Pointcloud Velodyne VLP16
Download links
- Dataset
The dataset is available for download as ROS 1 Noetic bags and as MRPT rawlogs (for ROS2 Humble):
| Dowload sequence | Length [m] | Duration [s] | Size (GB) |
| SEQ_01_2022-10-05 | 459.2 | 696 | 20.4 |
| SEQ_02_2022-10-14 | 457.36 | 701 | 34.1 |
| SEQ_03_2022-10-19 | 1321.21 | 1432 | 69.5 |
| SEQ_04_2022-10-26 | 1432.08 | 1463 | 27.6 |
| SEQ_05_2022-11-02 | 1233.87 | 1486 | 71.7 |
| SEQ_06_2022-11-09 | 1293.29 | 1532 | 73.5 |
| SEQ_07_2022-11-17 | 1332.58 | 1752 | 84.3 |
| SEQ_08_2022-11-23 | 1428.45 | 1692 | 66.1 |
| SEQ_09_2022-11-30 | 1440.32 | 1730 | 83.4 |
- Calibration
The files used to calibrate the sensors are available here.
- Tools
The tools necessary to perform the MOLA SLAM algorithm correctly are available here.
Copyright
All datasets and benchmarks on this page are copyright by us and published under the Creative Commons Attribution-NonCommercial-ShareAlike 3.0 License. This means that you must attribute the work in the manner specified by the authors, you may not use this work for commercial purposes and if you alter, transform, or build upon this work, you may distribute the resulting work only under the same license.
Citation
@Article{canadas2024greenbot,
AUTHOR = {Cañadas-Aránega, Fernando and Blanco-Claraco, Jose Luis and Moreno, Jose Carlos and Rodriguez-Diaz, Francisco},
TITLE = {Multimodal Mobile Robotic Dataset for a Typical Mediterranean Greenhouse: The GREENBOT Dataset},
JOURNAL = {Sensors},
VOLUME = {24},
YEAR = {2024},
NUMBER = {6},
ARTICLE-NUMBER = {1874},
URL = {https://www.mdpi.com/1424-8220/24/6/1874},
ISSN = {1424-8220},
DOI = {10.3390/s24061874}
}
Changelogs
- 15/03/2024: Our GreenBot Dataset paper is available for download now!
- 08/02/2024: Calibration data problem solved.
- 31/01/2024: Initial data release.
Privacy
This dataset is made available for academic use only. However, we take your privacy seriously! If you find yourself or personal belongings in this dataset and feel unwell about it, please get in touch with us and we will immediately remove the respective data from our server.
- Fernando Cañadas-Aránega
: fernando.ca@ual.es - J. Luis Blanco-Claraco: jlblanco@ual.es
- J. Carlos Moreno: jcmoreno@ual.es
- F. Rodriguez: frrodrig@ual.es
Credits
This work has been partially financed by the ‘CyberGreen’ Project, PID2021-122560OB-I00, and the ‘AgroConnect.es’ infrastructure used to carry out this research, grant EQC2019-006658-P, both funded by MCIN/AEI/ 10.13039/501100011033 and by ERDF A way to make Europe, with the support of the Regional Ministry of Economic Transformation, Industry, Knowledge and Universities and the European Regional Development Fund (FEDER) with the projects UAL2020-TEP-A1991 and PY2020_007-A1991. Also, author Fernando Cañadas-Aránega is supported by an FPI grant from Spanish Ministery of Science, Innovation and Universities.