Radar-only ego-motion estimation in difficult settings via graph matching

Sarah Huiyi Cen, Paul Newman

Abstract: Radar detects stable, long-range objects under variable weather and lighting conditions, making it a reliable and versatile sensor well suited for ego-motion estimation. In this work, we propose a radar-only odometry pipeline that is highly robust to radar artifacts (e.g., speckle noise and false positives) and requires only one input parameter. We demonstrate its ability to adapt across diverse settings, from urban UK to off-road Iceland, achieving a scan matching accuracy of approximately 5.20 cm and 0.0929 deg when using GPS as ground truth (compared to visual odometry's 5.77 cm and 0.1032 deg). We present algorithms for keypoint extraction and data association, framing the latter as a graph matching optimization problem, and provide an in-depth system analysis.

The Challenge of Radar Odometry

Autonomous vehicles typically rely on a standard sensor suite of vision, lidar, and GPS. However, these sensors are fragile in adverse conditions—cameras fail in poor lighting, lidar struggles in heavy rain or snow, and GPS signals drop out in tunnels or urban canyons.

Radar scanning geometry and common artifacts like false positives.

Radar is a highly versatile alternative that penetrates weather and operates in all lighting conditions. However, radar-only ego-motion estimation is notoriously difficult. Radar scans suffer from low resolution, heavy speckle noise, clutter, and false positives caused by multipath reflections. Existing radar odometry methods usually require heavy parameter tuning, rely on restrictive geometric models, or need strong motion priors. This work addresses these challenges by building a purely radar-based odometry system that works across diverse environments with minimal human intervention.

Robust Pipeline Design

The proposed indirect (feature-based) odometry pipeline is divided into two main novel components: one-parameter keypoint extraction and data association via graph matching.

Overview of the radar odometry pipeline from raw scans to scan matching.

One-Parameter Keypoint Extraction

Instead of using traditional Constant False Alarm Rate (CFAR) detectors which require multiple tuned parameters, the system utilizes a gradient-based approach. A Prewitt operator is applied to the radar scan to find peaks (areas of high intensity and low gradient). The algorithm suppresses redundant returns by keeping only one keypoint per continuous bright region. This requires only a single tunable parameter: the maximum number of keypoints ($\ell_{max}$).

Data Association via Graph Matching

To match keypoints between consecutive scans, the authors abandon iterative local searches (like ICP) in favor of global graph matching. This approach leverages the global geometric structure of the scene, making the system incredibly robust to radar's "ghost" objects and speckle noise, as random noise rarely maintains consistent global geometry between scans.

  1. Unary Descriptors: A new rotation-invariant descriptor is introduced for each keypoint based on histograms of surrounding angular slices and annuli, normalized using a Fast Fourier Transform (FFT).
  2. Pairwise Compatibility: The system checks if the geometric distance and relationship between two points in Scan A match the relationship of their proposed pairs in Scan B.
  3. Optimization: The matching is solved by finding the principal eigenvector of a compatibility matrix. A novel termination condition stops adding matches the moment a new match decreases the overall mutual compatibility of the set.

Experimental Evaluation

The system was evaluated on six diverse datasets collected using a millimeter-wave (MMW) FMCW scanning radar, spanning urban UK streets to unstructured Icelandic boulder fields.

  • High Accuracy: Across all datasets, the Radar Odometry (RO) pipeline achieved a median translation error of 5.20 cm and a rotation error of 0.0929 degrees compared to GPS ground truth.
  • Performance Benchmarks: RO slightly outperformed state-of-the-art Visual Odometry (VO), which had median errors of 5.77 cm and 0.1032 degrees. It also vastly outperformed standard Iterative Closest Point (ICP) matching across all environments.
  • Real-Time Capability: The system runs efficiently in real-time at ~4 Hz, matching the physical rotation speed of the scanning radar.

Conclusion and Future Outlook

This research proves that scanning radar can provide primary, highly accurate odometry that rivals vision, even without a motion prior. While the current system is limited to 3-DOF planar motion estimation—which can lead to slight errors in highly unstructured terrain where pitch and roll are significant—it demonstrates remarkable generalization. Future work will focus on deducing 3D environmental information from 2D scans and further streamlining the candidate proposal step to reduce computational complexity.

Summary updated

Abstract

Radar detects stable, long-range objects under variable weather and lighting conditions, making it a reliable and versatile sensor well suited for ego-motion estimation. In this work, we propose a radar-only odometry pipeline that is highly robust to radar artifacts (e.g., speckle noise and false positives) and requires only one input parameter. We demonstrate its ability to adapt across diverse settings, from urban UK to off-road Iceland, achieving a scan matching accuracy of approximately 5.20 cm and 0.0929 deg when using GPS as ground truth (compared to visual odometry's 5.77 cm and 0.1032 deg). We present algorithms for keypoint extraction and data association, framing the latter as a graph matching optimization problem, and provide an in-depth system analysis.
BibTeX
@inproceedings{cen2019radar,
  author = {Sarah Huiyi Cen and Paul Newman},
  title = {{Radar-only ego-motion estimation in difficult settings via graph matching}},
  booktitle = {IEEE International Conference on Robotics and Automation},
  year = {2019},
  url = {https://openreview.net/forum?id=6PZvhqZut7-},
}
Figure