# Super Odometry: IMU-centric LiDAR-Visual-Inertial Estimator for Challenging Environments

**Authors:** Shibo Zhao, Hengrui Zhang, Peng Wang, Lucas Nogueira, Sebastian Scherer  
**arXiv:** [2104.14938](https://arxiv.org/abs/2104.14938)  
**Date:** 2021-04-30  
**Summary page:** https://www.summarizepaper.com/en/arxiv-id/2104.14938v2/

## Key points

- Authors propose a high-precision multi-modal sensor fusion framework called Super Odometry
- Super Odometry integrates LiDAR, camera, and IMU sensors for robust state estimation in challenging environments
- Approach adopts an IMU-centric data processing pipeline combining loosely coupled and tightly coupled methods
- Framework comprises three key components: IMU odometry, visual-inertial odometry, and laser-inertial odometry
- Visual-inertial and laser-inertial odometry provide pose priors to constrain IMU bias and receive motion predictions from IMU odometry
- Implementation includes a dynamic octree for real-time high performance with reduced computational overhead
- Successfully deployed on drones and ground robots for DARPA Subterranean Challenge, achieving first and second place in Tunnel and Urban Circuits respectively

## Summary

In their paper titled "Super Odometry: IMU-centric LiDAR-Visual-Inertial Estimator for Challenging Environments," authors Shibo Zhao, Hengrui Zhang, Peng Wang, Lucas Nogueira, and Sebastian Scherer propose a high-precision multi-modal sensor fusion framework known as Super Odometry. This innovative approach offers a simple yet effective solution for integrating multiple sensors such as LiDAR, camera, and IMU sensors to achieve robust state estimation in environments with degraded perception. Unlike traditional sensor-fusion methods, Super Odometry adopts an IMU-centric data processing pipeline that combines the strengths of loosely coupled and tightly coupled methods to recover motion in a coarse-to-fine manner. The proposed framework consists of three key components: IMU odometry, visual-inertial odometry, and laser-inertial odometry. Visual-inertial odometry and laser-inertial odometry provide pose priors to constrain IMU bias and receive motion predictions from IMU odometry. To ensure real-time high performance, the authors implement a dynamic octree that significantly reduces computational overhead compared to a static KD-tree. The Super Odometry system has been successfully deployed on drones and ground robots as part of Team Explorer's participation in the DARPA Subterranean Challenge. The team achieved remarkable success by securing first and second place in the Tunnel and Urban Circuits respectively. This showcases the potential for advanced sensor fusion techniques to enhance navigation capabilities in challenging environments and pave the way for improved autonomous systems across various applications.