Competition:

The Örebro 4D-radar SLAM Challenge involves developing and benchmarking radar–inertial SLAM on the Örebro dataset recorded with a Husky robot in January 2026. The environment combines forested terrain and campus/urban paths, with multiple opportunities for loop closure.

The top 3 submissions will be invited to deliver a 5-minute talk about their submission at the workshop. The talks will be delivered in person or over zoom during the workshop on Monday, June 1, 2026 at around 15:00 CET. We encourage participants to open-source their implementations after the competition for the benefit of the community. 

 Restrictions

• Submissions must run “online” (causal): estimates must not use future measurements.

• Loop closure is permitted when detected online

• Allowed inputs :  Radar point clouds and IMU only:

- Radars: /hugin_[1,2,3]/radar_data

- IMU: /imu/data (and /imu/mag if you choose to use magnetometer)

• Submission trajectory requirements:

- Submit a text file of tum format (timestamp x y z qx qy qz qw) trajectory of poses of the base_link frame.

- The submitted trajectory must be sampled faster than 10 Hz.

UPDATE (02/10):

• Batch optimization is allowed without restrictions on the batch size.

• The SLAM system must be able to function without human intervention, i.e., a posteriori manual editing of the map, such as manual addition of loop closures, is forbidden.

Downloading the dataset:

• The dataset can be accessed here: https://github.com/RNP-lab/orebro_4d_radar_slam_challenge_data  

• The dataset is split into:

- Training : '01_campus_training_localized' — all sensors and references are available for development and validation.

- Testing : '02_campus_eval_filtered'— only radar + IMU are provided; ground truth is withheld for competition evaluation.

Tip: Some sequences revisit previously seen areas. Using loop closure may help improve performance.

Submitting to the benchmark:

• Upload your submission through the Leaderboard website (TBD).

• Re-submissions will be limited for the competition leaderboard to prevent overfitting to the hidden test set.  

• Use the training leaderboard for iterative development and comparison with others.

• Evaluation will be based on ATE (RMSE, unit : m) against the RTK-GNSS reference after transforming the submitted base_link poses to the GNSS antenna frame using provided extrinsics.

Additional submission requirements:

Participants are also required to submit a 2-page+references write-up on their approach. Format is dual-column single-space, IEEE conference format.

Evaluation Policy & Fair Play

Our evaluation relies on the trust and fair play of the community. While the test and training sequences partially overlap, the use of prior maps generated from training data is strictly forbidden during the evaluation.

If you encounter any problems:

If you have any problems, please open a github issue at  https://github.com/RNP-lab/orebro_4d_radar_slam_challenge_data  

Disclaimer:

We reserve the right to change the rules of the competition at any point and to disqualify individuals who break any rules mentioned above or cheat in any other manner or who violate the IEEE code of ethics.