Segmentation-based Perception and NMPC Control for Autonomous UAV Flight in Unstructured Environments

Rapid advancements in the UAV (Unmanned Aerial Vehicle) industry have enabled UAVs to operate in complex and inaccessible environments across various fields. However, most commercial UAVs rely on manual control owing to safety concerns, making them vulnerable to visual limitations and communication disruptions. Autonomous navigation allows UAVs to perceive their surroundings, plan optimal paths, and avoid obstacles. However, traditional methods featuring LiDAR and GPS struggle in unstructured environments such as dense forests, where irregular terrain and sparse textures hinder accurate perception. To address these challenges, this study proposes a deep learning-based segmentation and reinforcement learning-driven optimal control approach to enhance UAV navigation in dynamic and unstructured environments, improving both adaptability and precision in real-world applications.