RTS‐GT Dataset 2023

Description of the RTS-GT Dataset 2023

Note: still in progress

This page gives a description about the RTS-GT Dataset. The following table describes the deployment done with some characteristics which are the prism data distance collected, the type of calibration evaluated, the robotic platform used and the environment where the experiments were done.

Deployments done

Table of deployments

Deployment	Number of experiments	Distance [km]	Two-points resection	Number of static GCPs	Dynamic GCPs calibration	Inter-prism calibration	Robot	Environment	GNSS	Lidar scan
24/02/2022	1	1.58	❌	8	❌	✔️	Warthog	Outdoor campus	✔️	❌
07/03/2022	1	0.98	❌	8	❌	✔️	Warthog	Outdoor campus	✔️	❌
12/03/2022	1	2.06	❌	8	❌	✔️	Warthog	Outdoor campus	✔️	✔️
14/03/2022	1	1.98	❌	8	❌	✔️	Warthog	Outdoor campus	✔️	❌
16/03/2022	1	1.24	❌	20	❌	✔️	Warthog	Outdoor campus	✔️	❌
31/03/2022	2	1.18	❌	10	❌	✔️	Warthog	Outdoor campus	✔️	❌
24/04/2022	2	1.97	❌	12-20	❌	✔️	Warthog	Forest	✔️	❌
05/05/2022	2	0.56	❌	12	❌	✔️	HD2	Tunnel	❌	❌
13/05/2022	6	8.63	❌	12-15	❌	✔️	Warthog	Outdoor campus	❌	❌
23/05/2022	4	0.99	❌	12	❌	✔️	HD2	Tunnel	❌	❌
25/05/2022	2	5.78	❌	12	❌	✔️	Warthog	Outdoor campus	✔️	❌
22/06/2022	2	1.75	❌	10	❌	✔️	Warthog	Outdoor campus	✔️	❌
30/06/2022	2	1.67	✔️	38	❌	✔️	Warthog	Outdoor campus	✔️	❌
11/07/2022	2	1.37	✔️	40	✔️	✔️	Warthog	Outdoor campus	✔️	❌
15/07/2022	4	0.28	✔️	10	✔️	❌	Warthog	Outdoor campus	✔️	✔️
17/07/2022	5	1.58	❌	15	✔️	✔️	HD2	Tunnel	❌	✔️
10/09/2022	9	0.86	❌	85	❌	✔️	HD2	Tunnel	❌	❌
03/11/2022	3	0.43	❌	12	❌	✔️	Warthog	Outdoor campus	✔️	❌
09/11/2022	3	1.58	❌	12	❌	✔️	Warthog	Forest	✔️	✔️
10/11/2022	1	1.58	❌	12	❌	✔️	Warthog	Forest	✔️	✔️
16/11/2022	3	0.87	❌	12	❌	✔️	Warthog	Outdoor campus	✔️	✔️
23/11/2022	1	0.48	❌	12	❌	✔️	Warthog	Outdoor campus	✔️	✔️
24/11/2022	2	0.67	❌	12	❌	✔️	Warthog	Outdoor campus	✔️	✔️
29/11/2022	5	1.46	❌	12	❌	✔️	Warthog	Outdoor campus	✔️	✔️
05/12/2022	3	1.4	❌	12	❌	✔️	Warthog	Outdoor campus	✔️	✔️
07/12/2022	2	0.7	❌	12	❌	✔️	Warthog	Outdoor campus	✔️	✔️
25/07/2023	3	2	❌	12	❌	✔️	Warthog	Outdoor campus	✔️	✔️

Notes about deployments

1. Deployment of 07/03/22 was done under a freezing rain weather.
2. Deployment of 12/03/22 was done under a snowy weather. Accuracy may have been affected because of the snow reflections.
3. Deployment of 31/03/22 was partially done in an area with trees (first experiment). Accuracy may have been affected because of the occlusions.
4. Deployment of 24/04/22 was done in a forest environment. Due to the high level of snow on the ground, it was not possible to place the RTS on a stable surface. The levelling was not correct and differ by more than 2 mm between the beginning and the end of the experiment. The second experiment done in the forest was during the night on a hilly ground. Due to the difference of altitude of the path used, prisms were not in direct line of view and few data were recorded.
5. Deployment of 13/05/22 was done during the day (experiments 1 to 3) and during the night (experiments 4 to 6). During the day, the temperatures were higher than 30 degrees Celsius. A difference of more than 10 degrees was observed between the setup time of the RTS and the gathering of the data. These difference of temperatures may have affected the accuracy of the measurements. We also noticed that the levelling was shifting to a wrong value during the experiments, probably because of this difference of temperatures. For the night experiments, temperatures were stable, and no incidents were noticed during the gathering of the data. Experiment 5 was done with very strong motion of the Warthog (high accelerations, speeds and rotation velocities).
6. Static GCPs were always taken at the beginning of a deployment. If many experiments were done in the same area, no new GCPs were taken, except when the time difference between two experiments was higher than 4h.
7. GCPs were taken with one prism on the robot to speed up the time to gather the data.
8. Data for the two-points resection were gathered on the campus ground according to four geodesic pillars (2 pillars used of the experiment on the 30/06/22), which relative positions are known with a millimetre accuracy. Maximum distance between two pillars is around 500 m.
9. Experiments in tunnel were done on a slippery surface. To place the RTS in a stable position, cracks on the ground or the use of weight were used.
10. RTS were always set at a different height to minimize occlusions during an experiment. The same for the prisms on the robotic platform. The lowest RTS was always tracking the highest prism on the robot, and the highest RTS was always tracking the lowest prism on the robot.
11. Because the prisms were active, i.e. meaning that they have an ID, their position on the robots were always the same to ensure efficient processing of the data after each experiment.
12. Positions of the RTS were set to be next to each other during a deployment, so in case of an occlusion it was possible to quickly access to the RTS and point it again to the prism. Therefore, one person was able to deploy the all setup to gather the data, although a minimum of two people is preferable.
13. Install of the RTS setup can take between 30 min to 1h depending on the number of people involved. Then the static GCPs gathering can take between 15 min to 1h depending on the number of points taken and their spread. Totally, a deployment can take between 2h and 4h.
14. The sensor extrinsic calibration was done after each deployment to ensure the data taken are close to the one of the experiments.
15. GNSS data were gathered when possible during experiments done outdoor. Three GPS antenna were located on the Warthog, and one GPS was used as reference antenna for the RTK GNSS correction.

Robots

Two robots were used during the deployements:

• A Warthog robot from Clearpath. This robot was used during all the outdoor experiments. The robot was on wheels during the summer and automn, and on tracks during winter and spring to go on snow. Due to its size and weight, this robot had more sensors on it during deployments, including lidars, GNSS, IMU and wheel encoders. Its speed can go up to 2.5 meter per second.
• A HD2 (Heavy Duty 2) robot from SuperDroid. This robot was deployed only in underground and indoor environments due to its small size. It has tracks and its speed was slower compare to the Warthog ranging from 0 to 1 meter per second.

Pictures

Pictures of the different deployment are available 👉 here

Download dataset

RTS-GT dataset will be uploaded soon ! The dataset is divided into four part:

• GNSS data can be downloaded 👉 here (zip file of 1.6GB)
• ROS1 rosbags (total of four zip files of 107.3GB):
  1. Eight deployments (January 2022 to May 2022) 👉 here (zip file of 7.6GB)
  2. One deployment (May 2022) 👉 here (zip file of 67.6GB)
  3. Six deployments (May 2022 to July 2022) 👉 here (zip file of 21.2GB)
  4. Eleven deployments (July 2022 to December 2022) 👉 here (zip file of 10.9GB)
• ROS2 rosbags (total of five zip files of 130.5GB)
  1. Eight deployments (January 2022 to May 2022) 👉 here (zip file of 7.0GB)
  2. One deployment (May 2022) 👉 here (zip file of 67.8GB)
  3. Six deployments (May 2022 to July 2022) 👉 here (zip file of 23.2GB)
  4. Four deployments (July 2022 to November 2022) 👉 here (zip file of 19.2GB)
  5. Seven deployments (November 2022 to December 2022) 👉 here (zip file of 13.3GB)
• Calibration information (available soon)

The dataset of ICRA 2023 paper can be downloaded 👉 here (zip file of 37.9GB)

GNSS data processing

GNSS data were gathered when possible during experiments done outdoor. Three GPS antenna were located on the Warthog, and one GPS was used as reference antenna for the RTK GNSS correction. GNSS data were also post-processed by the RTKlib library available 👉 here

The GNSS raw file data given by each GPS are available in the dataset. The four GPS used are 4 Emlid Reach RS+, or 4 Trimble R10-2

Rosbag data

The dataset was recorded with Melodic on ROS 1 for the RTS data. Rosbags recorded with the robots are done with Humble on ROS 2. Important topics are displays in the following tables for both robotic platform:

Relevant topics Warthog

Name	Description
/theodolite_master/theodolite_correction_timestamp	Topic which records the time corrections between the master and the clients
/theodolite_master/theodolite_data	Topic which records each RTS data collected by the Raspberry Pi
/imu/data	Topic which records the imu data of the robot
/warthog_velocity_controller/odom	Topic which records the odometry of the robot

Relevant topics Marmotte

Name	Description
/theodolite_master/theodolite_correction_timestamp	Topic which record the time corrections between the master and the clients
/theodolite_master/theodolite_data	Topic which records each RTS data collected by the Raspberry Pi
/MTI_imu/data	Topic which records the imu data of the robot
/odom	Topic which records the odometry of the robot