Study on the path optimization method of autonomous navigation of unmanned aerial vehicles integrating multi-sensor data

AI-driven and sensor-fusion methods for autonomous systems, our work addresses the critical challenge of enhancing UAV path planning through multi-sensor integration. Prior approaches—such as traditional GNSS-inertial fusion or single-camera vision—suffer from limitations in GPS-denied or dynamic environments, often lacking robustness, adaptability, or real-time performance amid obstacles and sensor noise citeturn0search1turn1search3.We propose a novel optimization framework that deeply integrates data from multiple onboard sensors—including LiDAR, vision, and inertial units—within a deep reinforcement learning architecture informed by our method section. By leveraging sensor fusion, our method dynamically adjusts flight trajectories with heightened spatial awareness, improving obstacle avoidance and significantly reducing path deviations. Experimental results demonstrate that our approach outperforms baseline single-sensor and non-fused systems, achieving smoother, safer routes with lower cumulative error and real-time responsiveness. This integration highlights the advantages of multi-sensor fusion inUAVnavigation, aligning strongly with the special issue’s theme of advanced data-driven autonomous flight.