Robust navigation in off-road terrain using radar

Thumbnail Image

Files

master_Klinger_Nils_2025.pdf (6.18 MB)

URL

Journal Title

Journal ISSN

Volume Title

School of Electrical Engineering | Master's thesis
Unless otherwise stated, all rights belong to the author. You may download, display and print this publication for Your own personal use. Commercial use is prohibited.

Authors

Date

2025-07-30

Department

Major/Subject

Space Technology

Mcode

Degree programme

Master's Programme in Space Science and Technology

Language

en

Pages

94

Series

Abstract

Harsh and unstructured environments such as underground tunnels, forests, or extraterrestrial terrains pose significant challenges for autonomous vehicle navigation, especially where traditional positioning systems like GNSS are unreliable and vision or LiDAR-based sensors fail due to poor visibility conditions. In such scenarios, radar emerges as a robust alternative, capable of operating in darkness, fog, or dust. However, radar-based simultaneous localisation and mapping (SLAM) faces unique limitations, including its primarily 2D data output and therefore limited vertical resolution and usability. This thesis addresses these challenges by developing a robust radar-based SLAM framework that enables autonomous navigation in off-road conditions. The approach leverages a Navtech RAS3 radar in combination with other sensors, and commercial off-the-shelf hardware components. A novel image processing method is proposed to create a "2.5-dimensional stacked terrain model" using 2D radar data, exploiting a technique referred to as “terrain-induced beam steering”. The framework integrates sensor fusion algorithms implemented in the robot operating system (ROS), combined with an odometry estimation via the “Conservative Filtering for Efficient and Accurate Radar” (CFEAR) algorithm. Preliminary test results, including static, tunnel, and off-road experiments, demonstrate the technical feasibility of the integrated sensor suite and the implemented radar-centric SLAM framework. While full consistency in odometry and high-quality mapping were not fully achieved, the system successfully performs sensor fusion, angle dependent radar image correction, and initial odometry estimation. This work lays the foundation for more advanced and fully autonomous ra-dar-based navigation systems, particularly in visibility-impaired or GNSS-denied environments. The developed framework serves as a robust baseline for future iterations focusing on long-term autonomy, improved loop closure, and higher map fidelity in complex off-road scenarios.

Description

Supervisor

Kucner, Tomasz

Thesis advisor

Tsirvoulis, Georgios

Keywords

SLAM, radar, image processing, sensor fusion, navigation, robotics

Other note

Citation

Permanent link to this item

https://urn.fi/URN:NBN:fi:aalto-202508186223

Collections

[dipl] Sähkötekniikan korkeakoulu / ELEC