SuperX SLAM builds real-time, robust SLAM systems for Spatial AI — from AirLab, Carnegie Mellon University.
Our systems let robots perceive the world, remember its evolution, and support reasoning and action:
| Project | What it is |
|---|---|
| TartanIMU | The first attempt to achieve IMU foundation Model for different robot platforms. |
| SuperMap | A spatio-temporal SLAM system for visual-language navigation (RSS 2026). Builds a persistent, queryable 4D scene graph: a living spatial memory for embodied AI. |
| SuperOdom | Highly robust and accurate LiDAR-only / LiDAR-inertial odometry, field-proven in degenerate environments (Science Robotics 2025). |
| Robustness_Metric | A robustness metric for evaluating odometry and SLAM systems. |
| SubT Dataset | Degraded SLAM dataset in challenging environments. |
Before you push any code, please follow these rules:
- Open an issue first. Every contribution starts with a GitHub issue — a bug report or a feature request — so we can discuss it before you write code. See the SuperMap issue list and SuperOdom issue list.
- Follow the Google code style. All contributed code must follow the Google style guides:
- C++ (SuperOdom): Google C++ Style Guide
- Python (SuperMap, Robustness_Metric): Google Python Style Guide
- Link your pull request to its issue. Every PR must reference the issue it resolves. How to do this?
- Assign a reviewer. Request a review on your PR before it can be merged. How to do this?
Check our progress:
- TartanIMU — code will be released by the end of this week.
- SuperMap — code release is coming after RSS 2026; we're finalizing it.
- SuperOdom — actively maintained and accepting new issues and pull requests.
- Project website — demos, papers, and the SuperMap interactive demo
- SuperMap paper (RSS 2026)
- Teaser video on YouTube
- Join our Discord — questions, discussion, and announcements
- Contact: shiboz@andrew.cmu.edu