Path planning algorithms for autonomous navigation of a non-holonomic robot in unstructured environments

Path planning is a crucial aspect of autonomous robot navigation, enabling robots to efficiently and safely navigate through complex environments. This thesis focuses on autonomous navigation for robots in dynamic and uncertain environments. In particular, the project aims to analyze the localization and path planning problems. A fundamental review of the existing literature on path planning algorithms has been carried on. Various factors affecting path planning, such as sensor data fusion, map representation, and motion constraints, are also analyzed. Thanks to the collaboration with E80 Group S.p.A., the project has been developed using ROS (Robot Operating System) on a Clearpath Dingo-O, an indoor mobile robot. To address the challenges posed by unstructured and dynamic environments, ROS follows a combined approach of using a global planner and a local planner. The global planner generates a high-level path, considering the overall environment, while the local planner handles real-time adjustments to avoid moving obstacles and optimize the trajectory. This thesis describes the role of the global planner in a ROS-framework. Performance benchmarking of traditional algorithms like Dijkstra and A*, as well as other techniques, is fundamental in order to understand the limits of these methods. In the end, the Hybrid A* algorithm is introduced as a promising approach for addressing the issues of unstructured environments for autonomous navigation of a non-holonomic robot. The core concepts and implementation details of the algorithm are discussed, emphasizing its ability to efficiently explore continuous state spaces and generate drivable paths.The effectiveness of the proposed path planning algorithms is evaluated through extensive simulations and real-world experiments using the mobile platform. Performance metrics such as path length, execution time, and collision avoidance are analyzed to assess the efficiency and reliability of the algorithms.