Autonomous Route Generation and Execution for UAV-Based Inspection of Railroad Infrastructure in Satellite-Denied Environments

The expansion of railroad construction has intensified the demand for efficient inspection of infrastructure, especially in complex environments such as the undersides of bridges and enclosed box girders where satellite signals are unavailable. Traditional inspection methods struggle to address these areas due to accessibility and safety concerns. This paper proposes an autonomous route generation and execution framework for UAV-based inspection of railroad facilities in satellite-denied scenarios. Leveraging SLAM-generated 3D raster maps, we identify surface points using a weighted PCA approach to estimate normal vectors, followed by clustering and merging to define coherent inspection surfaces. Hierarchical waypoints are then systematically generated based on UAV field-of-view constraints, ensuring safe distances and optimal coverage. These waypoints are connected into structured multi-layer routes, which are adaptively selected and executed using kinodynamic planning for obstacle avoidance and yaw control. Simulation results demonstrate that the proposed method can rapidly generate inspection routes immediately after environment mapping, enabling precise and comprehensive inspection of complex infrastructures. This approach provides a robust foundation for deploying autonomous UAV inspections in GNSS-denied railroad scenarios.