George Chustz1 and Srikanth Saripalli1
9/15/2021 Initial release
Localization is one of the fundamental problems in robotics. Visual Inertial Odometry is a set of algorithms which attempt to estimate the position and orientation of a subject using only camera(s) and an inertial measurement unit (IMU). We release the RELLIS Off-road Odometry Analysis Dataset to fill a void in available VIO datasets to provide high-quality, accurately time stamped off-road traversal data sequences for VIO researchers and developers. Across our data sequences, there is over 20,000 images, 250,000 IMU readings, and 6,000 RTK + heading measurements.
Dimensions in cm.
- Basler Pylon Camera - 1920x1200 @ 30FPS, PTP enabled, Driver here
- Vectornav VN300 IMU - 400 Hz, GPS denied, Driver here
- Ardusimple simpleRTK2B kit - 10 Hz, RTK GPS + heading, Driver here
Data included in raw ROS bagfiles:
| Topic Name | Message Type | Message Descriptison |
|---|---|---|
| /pylon_camera_node/image_raw | sensor_msgs/Image | Images from the Basler Pylon Camera |
| /vectornav/IMU | sensor_msgs/Imu | Imu data from VectorNav-VN300 |
| /UBX/hpposllh | ubxtranslator/hpposllh | GPS data from the ground truth RTK GPS |
| /UBX/relpos2D | ubxtranslator/relpos2D |
ROS BAG file, ground truth, and calibration results download links:
| Dataset | Bag | Result/Ground truth |
|---|---|---|
| rt4_calib | Bag Download [6GB] | Kalibr Results TXT |
| rt4_gravel | Bag Download [8GB] | Ground Truth CSV |
| rt4_rim | Bag Download [5GB] | Ground Truth CSV |
| rt4_updown | Bag Download [12GB] | Ground Truth CSV |
| rt5_calib | Bag Download [6GB] | Kalibr Results TXT |
| rt5_gravel | Bag Download [7GB] | Ground Truth CSV |
| rt5_rim | Bag Download [5GB] | Ground Truth CSV |
| rt5_updown | Bag Download [10GB] | Ground Truth CSV |
To complement our data contribution, we have also evaluated two leading VIO implementations on our datasets, OpenVINS and VINS-Fusion. Their estimated trajectories and videos of their performances can be found below.
In order to determine the efficacy of Kalibr's extrinsics calibration tool, we collected an additional 11 ROS bags of camera-IMU extrinsic calibration motion.
- Calibration Datasets Download 38GB
- The IMU, camera, and target parameters are given here.
@misc{chustz2021rooad,
title={ROOAD: RELLIS Off-road Odometry Analysis Dataset},
author={George Chustz and Srikanth Saripalli},
year={2021},
eprint={2109.08228},
archivePrefix={arXiv},
primaryClass={cs.RO}
}
All datasets and code on this page are copyright by us and published under the Creative Commons Attribution-NonCommercial-ShareAlike 3.0 License.
RELLIS-3D: A Multi-modal Dataset for Off-Road Robotics
A RUGD Dataset for Autonomous Navigation and Visual Perception inUnstructured Outdoor Environments