SLAM Challenge Dataset 2021

Advancing the field with a common benchmark

The dataset contains indoor sequences of offices, labs, and construction environments and outdoor sequences of construction sites and parking areas. All these sequences are characterized by featureless areas and varying illumination conditions that are typical in real-world scenarios and pose great challenges to SLAM algorithms that have been developed in confined lab environments. Accurate sparse ground truth, at millimeter level, is provided for each sequence. The sensor platform used to record the data includes a number of visual, lidar, and inertial sensors, which are spatially and temporally calibrated.

The purpose of this dataset is to foster the research in sensor fusion to develop SLAM algorithms that can be deployed in tasks where high accuracy and robustness are required, e.g., in construction environments.

	Sequence	Rosbag Truth
	RPG Drone Testing Arena	5 GB 6DOF
	IC Office	10 GB 3DOF
	Office Mitte	16 GB 3DOF
	Parking Deck	27 GB 3DOF
	Basement	6 GB 3DOF
	Basement 3	16 GB 3DOF
	Basement 4	20 GB 3DOF

	Sequence	Rosbag Truth
	Lab	8 GB 6DOF
	Construction Site Outdoor 1	10 GB 3DOF
	Construction Site Outdoor 2	20 GB 3DOF
	Campus 1	23 GB 3DOF
	Campus 2	20 GB 3DOF

Hint: To speed up playback, use rosbag decompress *.bag.

File format

We provide all datasets in binary format as rosbag. The rosbag files contain images and IMU measurements using the standard sensor_msgs/Image and sensor_msgs/ Imu message types, respectively. The Ouster data uses the standard sensor_msgs/PointCloud2 format while the Livox data is stored in the custom livox_ros_driver/CustomMsg message type. The TF tree contains all transformations between the sensors.

The 3DOF groundtruth is measured using a total station while the device was held stationary, while the 6DOF data is recorded using an Optitrack motion capture system. The accuracy can be assumed to be better than 5mm. The frame of the groundtruth data is indicated by the file name (Construction_Site_2_prism.txt means the groundtruth is in the /prism frame).

GitHub

Citation

When using this work in an academic context, please cite in the following manner:

@misc{2109.11316,
    Author = {Michael Helmberger and Kristian Morin and Beda Berner and Nitish Kumar and Danwei Wang and Yufeng Yue and Giovanni Cioffi and Davide Scaramuzza},
    Title = {The {Hilti} {SLAM} Challenge Dataset},
    Year = {2021},
    Eprint = {arXiv:2109.11316}
}

License

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.

Contact

Any dataset-related questions and concerns can be raised as issues at github.com/Hilti-Research/hilti-slam-challenge-2021/issues

Other topics should be forwarded to challenge@hilti.com

Hardware

The sensor suite consists of:

Sevensense Alphasense
Ouster OS0-64
Livox MID70
ADIS16445 IMU

The sensors are mounted on a surveying pole for handheld operation. The CAD Model in STEP format can be downloaded here. The synchronization between the sensors is done by a FPGA for the cameras and the ADIS16445 IMU. The cameras and the LIDARs are synchronized via PTP. The time between all sensors is aligned to within 1 ms.

Calibration

The calibration file can be downloaded separately here (updated Aug 26th). The rosbags contain the TF tree with all transformations between the sensors. The CAD Model in STEP format can be downloaded here. The imu noise parameters can be found in the corresponding data sheets: