# Degradation Resilient LiDAR-Radar-Inertial Odometry

**Authors:** Morten Nissov, Nikhil Khedekar, Kostas Alexis  
**arXiv:** [2403.05332](https://arxiv.org/abs/2403.05332)  
**Date:** 2024-03-08  
**Summary page:** https://www.summarizepaper.com/en/arxiv-id/2403.05332v1/

## Key points

- Ensuring robust operation of autonomous robots in challenging environments is crucial for advancing autonomy in the field of robotics.
- Reliability of estimation and odometry is a common challenge faced by autonomous systems, which can lead to catastrophic consequences as a single point of failure.
- Researchers Morten Nissov, Nikhil Khedekar, and Kostas Alexis proposed a novel method for enhancing odometry performance through LiDAR-Radar-Inertial fusion.
- The key innovation involves integrating radar and inertial sensor data to complement LiDAR measurements and mitigate LiDAR degeneracy within a factor graph-based windowed smoother framework.
- Sensor information-specific factor formulations enable conveying partial information along non-degenerate axes to address degeneracy issues.
- Real-world tests with a flying robot operating under degraded conditions showed improved odometry performance and robustness, highlighting the potential impact on enabling autonomous systems to navigate complex scenarios more reliably.
- The researchers have made their datasets publicly available on GitHub to foster collaboration, reproducibility, further research validation, and advancement of robotic technology through transparent practices.

## Summary

In the field of robotics, ensuring the robust operation of autonomous robots in challenging environments is crucial for advancing autonomy. One common challenge faced by autonomous systems is the reliability of estimation and odometry, which can serve as a single point of failure with potentially catastrophic consequences. Addressing this issue, a group of researchers including Morten Nissov, Nikhil Khedekar, and Kostas Alexis have proposed a novel method for enhancing odometry performance through tightly-coupled LiDAR-Radar-Inertial fusion. The key innovation of their approach lies in mitigating the effects of LiDAR degeneracy by integrating radar and inertial sensor data to complement LiDAR measurements. By combining these different sensing modalities within a factor graph-based windowed smoother framework, the researchers aim to leverage the strengths of each sensor type while compensating for their individual weaknesses. Specifically, they introduce sensor information-specific factor formulations that enable partial information to be conveyed to the graph along non-degenerate axes in cases where degeneracy occurs. To validate the effectiveness of their proposed method, real-world tests were conducted using a flying robot operating under degraded conditions such as geometric self-similarity and obscurant occlusion. The results demonstrated improved odometry performance and robustness in challenging environments, showcasing the potential impact of this research on enabling autonomous systems to navigate complex scenarios with greater reliability. Furthermore, in a gesture towards fostering collaboration and knowledge sharing within the robotics community, the researchers have made the datasets from their experiments publicly available on GitHub (https://github.com/ntnu-arl/lidar_degeneracy_datasets). This open access approach not only facilitates reproducibility and further research validation but also underscores the commitment of the authors to advancing robotic technology through transparent and collaborative practices.