[Paper Review] Radar SLAM: A Robust SLAM System for All Weather Conditions

Ziyang Hong, Yvan Pétillot+2 more|arXiv (Cornell University)|Apr 12, 2021

Robotics and Sensor-Based Localization51 references19 citations

TL;DR

This paper proposes a robust FMCW radar-based SLAM system that enables accurate, real-time localization and mapping in all-weather conditions, including heavy snow and dense fog. By leveraging radar geometry for feature tracking, jointly optimizing motion compensation and pose estimation, and using a structural loop closure detection method, the system achieves state-of-the-art accuracy and reliability across diverse outdoor environments with minimal parameter tuning.

ABSTRACT

A Simultaneous Localization and Mapping (SLAM) system must be robust to support long-term mobile vehicle and robot applications. However, camera and LiDAR based SLAM systems can be fragile when facing challenging illumination or weather conditions which degrade their imagery and point cloud data. Radar, whose operating electromagnetic spectrum is less affected by environmental changes, is promising although its distinct sensing geometry and noise characteristics bring open challenges when being exploited for SLAM. % However, there are still open challenges since most existing visual and LiDAR SLAM systems do not operate in bad weathers. This paper studies the use of a Frequency Modulated Continuous Wave radar for SLAM in large-scale outdoor environments. We propose a full radar SLAM system, including a novel radar motion tracking algorithm that leverages radar geometry for reliable feature tracking. It also optimally compensates motion distortion and estimates pose by joint optimization. Its loop closure component is designed to be simple yet efficient for radar imagery by capturing and exploiting structural information of the surrounding environment. % while a scheme to reject ambiguous loop closure candidates is also designed specifically for radar. Extensive experiments on three public radar datasets, ranging from city streets and residential areas to countryside and highways, show competitive accuracy and reliability performance of the proposed radar SLAM system compared to the state-of-the-art LiDAR, vision and radar methods. The results show that our system is technically viable in achieving reliable SLAM in extreme weather conditions, e.g. heavy snow and dense fog, demonstrating the promising potential of using radar for all-weather localization and mapping.

Motivation & Objective

• Address the limitations of camera and LiDAR-based SLAM systems in adverse weather conditions such as heavy snow, fog, and low visibility.
• Overcome the challenges posed by radar's distinct sensing geometry, high noise levels, and lack of elevation data for SLAM applications.
• Develop a fully autonomous radar SLAM system that operates reliably across diverse outdoor environments without relying on external sensors.
• Achieve robust loop closure detection and motion compensation using only FMCW radar data, ensuring long-term consistency and accuracy.
• Demonstrate that a single parameter set can be used across multiple datasets and weather conditions, minimizing tuning effort.

Proposed method

• Proposes a novel radar motion tracking algorithm that leverages the geometric structure of radar returns to improve feature association and reject outliers.
• Introduces a joint optimization framework that simultaneously estimates pose and compensates for motion distortion caused by low radar scan rates.
• Designs a loop closure detection scheme that exploits structural patterns in FMCW radar intensity images to identify repeated locations efficiently.
• Employs a keypoint detection method based on Hessian thresholding and PCA-based outlier rejection to enhance feature stability.
• Uses a pose graph optimization framework with loop closure constraints to correct drift and improve long-term trajectory accuracy.
• Implements a keyframe-based system with distance and rotation thresholds to manage computational load and maintain map consistency.

Experimental results

Research questions

• RQ1 Can a radar-based SLAM system achieve comparable accuracy to state-of-the-art LiDAR and vision-based SLAM systems in normal weather conditions?
• RQ2 How effective is a radar SLAM system in maintaining localization and mapping performance under extreme weather conditions such as heavy snow and dense fog?
• RQ3 Can a single, fixed parameter set be used across diverse environments and weather conditions without significant tuning?
• RQ4 How does the proposed motion compensation model reduce trajectory drift caused by low scan rates in radar SLAM?
• RQ5 Can structural features in FMCW radar intensity images be effectively used for robust loop closure detection without relying on 3D geometry or external sensors?

Key findings

• The proposed radar SLAM system achieved 100% completion rate on all three public radar datasets—Oxford, MulRan, and RADIATE—demonstrating superior robustness compared to vision and LiDAR-based methods.
• On the RADIATE dataset, which includes heavy snow and nighttime sequences, ORB-SLAM2 failed to initialize due to camera occlusion and blur, while the radar SLAM system remained fully operational.
• The system maintained consistent performance across all test sequences, including high-speed motion in snow, with no tracking loss and successful loop closure in all cases.
• The average completion percentage was 100% for the proposed radar SLAM across all datasets, outperforming SuMa (20–72%) and baseline SLAM (100% only on some datasets).
• The system runs at 8 Hz on a standard laptop (Intel i7, 16GB RAM), exceeding the 4 Hz radar frame rate, ensuring real-time performance with independent loop closure and optimization threads.
• The same parameter set was used across all experiments, confirming the system’s minimal tuning requirement and high adaptability to varying radar resolutions, ranges, and environmental conditions.