Welcome to the Jungle: Traversability Estimation for Autonomous Ground Vehicles in Densely Vegetated Environments
Fabio Ruetz · Queensland University of Technology · 2025
This thesis focuses on autonomous ground vehicle navigation in densely vegetated environments by developing strategies that enable robots to push through vegetation and learn from interaction in real-time. It introduces a novel 3D probabilistic representation that fuses LiDAR scans, camera images, and robot experience for improved environmental modelling. Several methods, including ForestTrav, were proposed and published, enabling real-time terrain evaluation in dense vegetation. Additionally, an adaptive terrain assessment method allows robots to learn from interactions and quickly adapt to new environments. This work advances robotic navigation in unstructured, vegetated environments - to boldly go where no robot has gone before.