Topographic Estimation, Online Trajectory Rollout, and Experimental Platforms for Autonomous Forest Machines
Time: Tue 2025-05-20 09.00
Location: Gladan, Brinellvägen 85, Stockholm
Video link: https://kth-se.zoom.us/j/63090566219
Language: English
Subject area: Machine Design
Doctoral student: Gustav Sten , Mekatronik och inbyggda styrsystem, MMK, ITM, KTH
Opponent: Associate Professor Grzegorz Orzechowski, LUT University
Supervisor: Docent Lei Feng, KTH-centrum inom inbyggda system, ICES, Mekatronik och inbyggda styrsystem; Dr Björn Möller, Mekatronik och inbyggda styrsystem
Abstract
This thesis presents a comprehensive framework for advancing autonomous operations in unstructured terrains, focusing on the forestry industry. The research addresses critical challenges in autonomous systems development by integrating cutting-edge navigation, perception, and control technologies. As automation demand in forestry grows, current systems struggle in unpredictable off-road terrains. Unlike structured road autonomy, forest machines must navigate rough, obstacle-filled landscapes without predefined paths, yet existing solutions lack the needed adaptability. Consequently, forestry remains reliant on manual labor, especially in planting and site preparation, where automation is limited. Tackling these challenges requires smarter autonomous navigation, enhanced machine-terrain interaction, and sustainable automation strategies that boost productivity while reducing environmental impact. Key contributions of this thesis include (1) a novel roll-out path planning algorithm leveraging terrain-vehicle interaction to optimize navigation in rough terrains, validated through simulations and real-world deployments, (2) a sensor fusion method combining LIDAR and stereo camera data to enhance topographic estimation with a good balance between accuracy and coverage, (3) a modular, reconfigurable test platform offering a scalable and cost-effective solution for evaluating autonomous system components, bridging the gap between simulation and real-world testing, and (4) a demonstration prototype system for autonomous plant regeneration, demonstrating the feasibility of fully autonomous forestry operations, including site preparation and planting, reducing environmental impacts, and improving efficiency. By addressing sustainability challenges and introducing robust methodologies for autonomous systems, this work contributes to the broader application of intelligent machinery in forestry and beyond.