Robust Localization for Mobile Robots in Forest Environments

This book presents a novel method for robust and accurate localization of mobile robots in forests. The central idea of this approach is to use trees as natural landmarks and employ a simultaneous localization and mapping procedure to build a tree map. Tree detection and parameter determination is performed on environment data perceived by a 3D laser scanner. These detected trees are combined in a common map and the vehicle is localized within this map. To solve the association problem between trees in a scan and the ones in a map, a Kalman filter based ego motion estimator and an iterative search algorithm are utilized. Global map alignment and long-term accuracy of the solution are achieved by leasts squares adjustment, incorporating GNSS measurements. Within this thesis all performed steps and the complete system are described in detail. The results of multiple experiments, demonstrating the performance of single components and the combined system, are presented.