Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Radar-Based Odometry for Low-Speed Driving

Luis Diener, Jens Kalkkuhl, Markus Enzweiler

TL;DR

The work introduces a radar-inertial SLAM framework optimized for low-speed automotive odometry and parking, leveraging a tightly coupled EKF that fuses inertial data with 4D radar measurements (range, bearing, elevation, and Doppler). By representing features on the unit sphere and integrating Doppler updates directly with feature states, the method improves bearing accuracy and information content, enabling centimeter-level localization without vehicle-specific calibration. An information-based feature pruning strategy and a multi-radar extension further enhance robustness and data association across sensors. Experimental results on proprietary and public datasets show strong performance at low speeds, with competitive results at higher speeds, and demonstrate the practical viability for automated parking where calibration-free operation and radar reliability are advantages.

Abstract

We address automotive odometry for low-speed driving and parking, where centimeter-level accuracy is required due to tight spaces and nearby obstacles. Traditional methods using inertial-measurement units and wheel encoders require vehicle-specific calibration, making them costly for consumer-grade vehicles. To overcome this, we propose a radar-based simultaneous localization and mapping (SLAM) approach that fuses inertial and 4D radar measurements. Our approach tightly couples feature positions and Doppler velocities for accurate localization and robust data association. Key contributions include a tightly coupled radar-Doppler extended Kalman filter, multi-radar support and an information-based feature-pruning strategy. Experiments using both proprietary and public datasets demonstrate high-accuracy localization during low-speed driving.

Radar-Based Odometry for Low-Speed Driving

TL;DR

The work introduces a radar-inertial SLAM framework optimized for low-speed automotive odometry and parking, leveraging a tightly coupled EKF that fuses inertial data with 4D radar measurements (range, bearing, elevation, and Doppler). By representing features on the unit sphere and integrating Doppler updates directly with feature states, the method improves bearing accuracy and information content, enabling centimeter-level localization without vehicle-specific calibration. An information-based feature pruning strategy and a multi-radar extension further enhance robustness and data association across sensors. Experimental results on proprietary and public datasets show strong performance at low speeds, with competitive results at higher speeds, and demonstrate the practical viability for automated parking where calibration-free operation and radar reliability are advantages.

Abstract

We address automotive odometry for low-speed driving and parking, where centimeter-level accuracy is required due to tight spaces and nearby obstacles. Traditional methods using inertial-measurement units and wheel encoders require vehicle-specific calibration, making them costly for consumer-grade vehicles. To overcome this, we propose a radar-based simultaneous localization and mapping (SLAM) approach that fuses inertial and 4D radar measurements. Our approach tightly couples feature positions and Doppler velocities for accurate localization and robust data association. Key contributions include a tightly coupled radar-Doppler extended Kalman filter, multi-radar support and an information-based feature-pruning strategy. Experiments using both proprietary and public datasets demonstrate high-accuracy localization during low-speed driving.

Paper Structure

This paper contains 18 sections, 31 equations, 4 figures, 4 tables.

Table of Contents

  1. Introduction
  2. Related Work
  3. Radar Odometry
  4. SLAM
  5. Notation
  6. System Description
  7. Motion Model
  8. Relative Feature Motion
  9. Radar Measurements
  10. Filter Framework
  11. Feature Management
  12. Extension to Multi-Radar Systems
  13. Experimental Results
  14. Experimental Setup
  15. Proprietary Datasets
  16. ...and 3 more sections

Figures (4)

  • Figure 1: Our proposed radar SLAM algorithm on a parking space. Depicted are the solution of our proposed SLAM algorithm (blue), the ground-truth trajectory (red), and the radar point map.
  • Figure 2: Overview of the proposed method: numerical integration of both the IMU motion (Sec. \ref{['sec:motion']}) and relative feature motion (Sec. \ref{['sec:feature']}), tightly coupled with 4D radar measurements (Sec. \ref{['sec:radar']}) and used for states estimation in an EKF framework (Sec. \ref{['sec:filter']}) with prior feature management (Sec. \ref{['sec:management']}).
  • Figure 3: The measured Doppler velocity (red) is applied to the estimated feature bearing (green), which exhibits lower uncertainty, thereby improving its contribution to the filter update.
  • Figure 4: Localization for two perpendicular parking maneuvers. The upper plots show the trajectories, while the plots below depict the error between the reference trajectory and the SLAM trajectory.