SuperLoc: The Key to Robust LiDAR-Inertial Localization Lies in Predicting Alignment Risks
*Equal contribution
Carnegie Mellon University
Code will be released in early March

Predict Alignment Risksâś…
Real World Degraded Environmentâť“
ICP Failure under Degeneracyâť“
Overview Video
About SuperLoc
Map-based LiDAR localization, while widely used in autonomous systems, faces significant challenges in degraded environments due to lacking distinct geometric features. This paper introduces SuperLoc, a robust LiDAR localization package that addresses key limitations in existing methods. SuperLoc features a novel predictive alignment risk assessment technique, enabling early detection and mitigation of potential failures before optimization (check out Why Predictive Alignment Risk). This approach significantly improves performance in challenging scenarios such as corridors, tunnels, and caves. Unlike existing degeneracy mitigation algorithms that rely on post-optimization analysis and heuristic thresholds, SuperLoc evaluates the localizability of raw sensor measurements. Experimental results demonstrate significant performance improvements over state-of-the-art methods across various degraded environments. Our approach achieves a 49.7% increase in accuracy and exhibits the highest robustness. To facilitate further research, we will release our implementation along with datasets from eight challenging scenarios.

SuperLoc is to our knowledge the first open-source LiDAR-inertial localization system that not only predicts alignment risks, and estimates the observability of scan, but also can actively incorporate pose priors from other odometry sources before failure occurs. The entire process doesn't require predefined heuristic thresholds to detect degeneration and it has been evaluated in various challenging environments including caves, long corridors, flat open areas, and staircases.
Cave
Multi-floor
Long Corridor
Comparison with other methods
Cave
Multi-Floor
Long Corridor
To benefit the open community, our localization package also includes following features,
Robust Initialization
Transition between mapped and unmapped region
Ground Truth Map
To benefit the open community, we release the following ground truth maps for localization:
Dataset
All datasets from our paper is released as follow,
Name | Source | Location | Robot | Sensor | Trajectory | Duration | Rosbag | Calibration (Extrinsics) | Calibration (Intrinsics) | GT Map | GT Traj. |
---|---|---|---|---|---|---|---|---|---|---|---|
Cave01 | SuperLoc | Laurel Craven | Handheld | RGB,LiDAR,IMU | 416 | 838 | link | Google Baidu | Google Baidu | link | link |
Cave02 | SuperLoc | Laurel Craven | Handheld | RGB,LiDAR,IMU | 475 | 986 | link | Google Baidu | Google Baidu | link | link |
Cave03 | SubT-MRS | Laurel Craven | Handheld | RGB,LiDAR,IMU | 490 | 768 | link | Google Baidu | Google Baidu | link | link |
Cave04 | SuperLoc | Laurel Craven | Handheld | RGB,LiDAR,IMU | 597 | 959 | link | Google Baidu | Google Baidu | link | link |
Corridor01 | SubT-MRS | Hawkins | RC2 | RGB,LiDAR IMU | 617 | 279 | link | Google Baidu | Google Baidu | link | link |
Corridor02 | SuperLoc | Hawkins | RC1 | RGB,LiDAR IMU | 690 | 893 | link | Google Baidu | Google Baidu | link | link |
Floor01 | SubT-MRS | Hawkins | SP1 | RGB,LiDAR,IMU | 270 | 480 | link | Google Baidu | Google Baidu | link | link |
Floor02 (bonus) | SuperLoc | Hawkins | SP1 | RGB,LiDAR,IMU | 410 | 2190 | link | Google Baidu | Google Baidu | link | link |
We also provide initialization poses for each dataset to support localization modules that allows initial pose configuration. You can find corresponding initial pose config for each dataset here.
Ground truth trajecotry follows TUM format,
timestamp x y z q_x q_y q_z q_w
Acknowledgments
The authors would like to express sincere thanks to Guofei Chen for helping us collect the dataset. The authors would like to thank Professor Ji Zhang for letting us borrow the Diablo platform and offering constructive advice.
Citation
Zhao & Zhu, et al. “SuperLoc: The Key to Robust LiDAR-Inertial Localization Lies in Predicting Alignment Risks.” 2025 IEEE International Conference on Robotics and Automation (ICRA). IEEE, 2025. Paper
@inproceedings{zhao2025superloc,
title = {SuperLoc: The Key to Robust LiDAR-Inertial Localization Lies in Predicting Alignment Risks},
author = {Zhao, Shibo and Zhu, Honghao and Gao, Yuanjun and Kim, Beomsoo and Qiu, Yuheng and Johnson, Aaron M. and Scherer, Sebastian},
year = {2025},
booktitle = {2025 IEEE International Conference on Robotics and Automation (ICRA)},
url = {https://arxiv.org/abs/2412.02901}
}
Zhao, et al. “SubT-MRS Dataset: Pushing SLAM Towards All-weather Environments.” 2024 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR). IEEE, 2024. Paper
@inproceedings{zhao2024subt-mrs,
title = {SubT-MRS Dataset: Pushing SLAM Towards All-weather Environments},
author = {Zhao, Shibo and Gao, Yuanjun and Wu, Tianhao and Singh, Damanpreet and Jiang, Rushan and Sun, Haoxiang and Sarawata, Mansi and Whittaker, Warren C and Higgins, Ian and Su, Shaoshu and Du, Yi and Xu, Can and Keller, John and Karhade, Jay and Nogueira, Lucas and Saha, Sourojit and Qiu, Yuheng and Zhang, Ji and Wang, Wenshan and Wang, Chen and Scherer, Sebastian},
year = {2024},
booktitle = {IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)},
doi = {10.1109/CVPR52733.2024.02137},
url = {https://superodometry.com/datasets},
video = {https://youtu.be/mkN72Lv8S7A}
}
Contacts
If you have any question or want to contribute this work, please feel free to send email to Shibo Zhao (shiboz@andrew.cmu.edu). Thank you! :)
Results
Impressive Results
Our approach demonstrates significant improvements in localization accuracy and efficiency across various environments
Enhanced Accuracy
Our method achieves superior localization accuracy compared to traditional approaches, with a significant reduction in error rates.
Real-time Performance
Efficient processing enables real-time localization updates, crucial for dynamic environments and mobile applications.
Robust Adaptation
The system demonstrates remarkable adaptability to different environmental conditions and sensor configurations.
Indoor Environment Results
Our method achieves precise localization in complex indoor environments, handling occlusions and dynamic changes effectively.
Outdoor Environment Results
Robust performance in outdoor settings, maintaining accuracy under varying lighting and weather conditions.
Dynamic Environment Results
Exceptional handling of dynamic scenarios with moving objects and changing environmental conditions.


