ReFeree: Radar-Based Lightweight and Robust Localization Using Feature and Free Space

2024-10-02

IEEE Robotics and Automation Letters

AI Summary

To address the failure of vision- and LiDAR-based localization under adverse weather conditions (e.g., rain, snow, overcast), this paper proposes a lightweight and robust millimeter-wave (mmWave) radar place recognition method. The approach innovatively integrates complementary noise modeling of feature points and free space, and introduces a rotation-invariant circular 1D range histogram descriptor. It achieves low computational overhead (<50 KB parameters, <3 ms inference on ARM Cortex-A72) while maintaining weather-agnostic robustness. Notably, it is the first mmWave radar method to enable heading estimation from the very first frame in unstructured environments, thereby enabling embedded SLAM frontends. Evaluated on multi-session, cross-weather datasets—including OORD—the method achieves >92% recall, significantly outperforming existing radar-based localization approaches.

Technology Category

• Intelligent Robots: Localization, Mapping, and Navigation
• Machine Learning: Hardware-aware ML
• Planning, Routing, and Scheduling: Activity and Plan Recognition

Application Category

• Systems and Infrastructure for Web, Mobile and WoT: Applied ML and AI for Web-based mobile applications
• Search and Retrieval-Augmented AI: Multilingual and cross-lingual Web search
• User Modeling, Personalization and Recommendation: On-Device user modeling, personalization, and recommendation

Abstract

Place recognition plays an important role in achieving robust long-term autonomy. Real-world robots face a wide range of weather conditions (e.g. overcast, heavy rain, and snowing) and most sensors (i.e. camera, LiDAR) essentially functioning within or near-visible electromagnetic waves are sensitive to adverse weather conditions, making reliable localization difficult. In contrast, radar is gaining traction due to long electromagnetic waves, which are less affected by environmental changes and weather independence. In this work, we propose a radar-based lightweight and robust place recognition. We achieve rotational invariance and lightweight by selecting a one-dimensional ring-shaped description and robustness by mitigating the impact of false detection utilizing opposite noise characteristics between free space and feature. In addition, the initial heading can be estimated, which can assist in building a SLAM pipeline that combines odometry and registration, which takes into account onboard computing. The proposed method was tested for rigorous validation across various scenarios (i.e. single session, multi-session, and different weather conditions). In particular, we validate our descriptor achieving reliable place recognition performance through the results of extreme environments that lacked structural information such as an OORD dataset.

Problem

• Achieve robust place recognition in adverse weather conditions
• Develop radar-based lightweight rotational-invariant localization
• Mitigate false detection using free space and feature noise

Innovation

• Radar-based lightweight localization using ring-shaped descriptors
• Robustness via opposite noise characteristics in features
• Heading estimation aiding SLAM with odometry and registration