SLAM Challenge Dataset 2022

Advancing the field with a common benchmark

The Hilti-Oxford Dataset has been collected on construction sites as well as on the famous Sheldonian Theatre in Oxford, providing a large range of difficult problems for SLAM.

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 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.

Challenge sequences

	Sequence	Rosbag Truth
	RExp01 Construction Ground Level Easy	18 GB 3DOF
	Exp02 Construction Multilevel Medium	34 GB 3DOF
	Exp03 Construction Stairs Hard	21 GB 3DOF
	Exp07 Long Corridor Medium	11 GB 3DOF
	Exp09 Cupola Hard	33 GB 3DOF
	Exp11 Lower Gallery Medium	12 GB 3DOF
	Exp15 Attic to Upper Gallery Hard	19 GB 3DOF
	Exp21 Outside Building Easy	11 GB 3DOF

Additional Sequences

	Sequence	Rosbag Truth
	Exp04 Constr. Upper Level 1 Easy	10 GB 3DOF
	Exp05 Constr. Upper Level 2 Easy	10 GB 3DOF
	Exp06 Constr. Upper Level 2 Medium	12 GB 3DOF
	Exp10 Cupola 2 Hard	27 GB 3DOF
	Exp14 Basement 2 Medium	6 GB 6DOF 3DOF
	Exp16 Attic to Upper Gallery 2 Hard	15 GB 6DOF 3DOF
	Exp18 Corridor Lower Gallery 2 Hard	8 GB 6DOF 3DOF
	Exp23 The Sheldonian Slam * Hard	80 GB 0 1 2 3DOF

* Includes all sections and revisits the ground hall several times for loop closures

Calibration

All ground truths are in the IMU frame. You can download the IMU Noise Calibration here. The calibration sequence for the board (see image below, download description) can be found here.

GitHub

Publication

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

@article{9968057,
    author = {Zhang, Lintong and Helmberger, Michael and Fu, Lanke Frank Tarimo and Wisth, David and Camurri, Marco and Scaramuzza, Davide and Fallon, Maurice},
    journal = {{IEEE} Robotics and Automation Letters},
    title = {{Hilti}-{Oxford} Dataset: A Millimeter-Accurate Benchmark for Simultaneous Localization and Mapping},
    year = {2023},
    volume = {8},
    number = {1},
    pages = {408-415},
    doi = {10.1109/LRA.2022.3226077}
}

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-2022/issues

Other topics should be forwarded to challenge@hilti.com

High-accuracy laser scans

We release the full 3D Terrestrial Laser Scans of the Construction Site , as well as the Sheldonian Theatre in .e57 cloud format. This enables map-based algorithm debugging insights.

We used the Z+F Imager 5016 , a survey grade terrestial laserscanner, and Scantra for the registration of the individual stations. The reported standard deviation for the registration (sigma_t) is below 1mm for all stations.

Hardware

Our sensor suite consists of a Sevensense Alphasense Core camera head with 5x 0.4MP global shutter cameras, and a Hesai PandarXT-32 .

The sensors are mounted rigidly on an aluminium platform for handheld operation. The synchronization between the cameras is done by a FPGA. The cameras and the LIDAR are synchronized via PTP. The time between all sensors is aligned to within 1 ms. An external steel-pin is attached to the setup

Calibration

The calibration file in Kalibr-like format can be downloaded here. Alternatively, you can use the provided calibration sequence to compute your own calibration or use the CAD model as an initial guess. The yaml-file for the calibration board can be downloaded here. The CAD Models can be downloaded as STP or STL. Datasheets for the sensors:

Additionally, we provide a 3D model of the sensor which has be generated by the GOM Atos Q Scanner (accuracy in micrometer range).