Scientific Development Research

Aiming at the reliability and real-time challenges of autonomous navigation and obstacle avoidance for unmanned aerial vehicles(UAVs)in complex low-altitude scenarios, this study integrates multi-source sensor data including LiDAR, vision, and inertial measurement units to establish a multimodal environmental perception system, which compensates for the perception deficiencies of single sensors under extreme conditions such as illumination variations, rain and fog. Deep learning and reinforcement learning algorithms are introduced to achieve high-precision obstacle recognition, dynamic trajectory prediction and millisecond-level path replanning, forming a closed-loop system of perception–decision–control. This technology can significantly improve the autonomous flight capability of UAVs in complex low-altitude environments such as urban streets and forests, reduce collision risks, and adapt to operational requirements in multiple scenarios including logistics distribution, power inspection and emergency rescue.

Unmanned Aerial Vehicle(UAV); Low-altitude autonomous navigation; Multi-sensor fusion; Artificial Intelligence(AI); Obstacle avoidance technology