Heterogeneous Multi-Sensor Fusion for UAV Autonomous Obstacle Avoidance and Dynamic Target Tracking
Abstract
In complex scenarios such as power inspection, tunnel exploration, and logistics distribution, a single sensor is insufficient to meet the safety and high-precision requirements for autonomous obstacle avoidance and target tracking of unmanned aerial vehicles. Multi-sensor fusion has become the core technical path. This paper reviews representative literature and focuses on the multi-sensor fusion architecture, obstacle avoidance and tracking methods of unmanned aerial vehicles. The research shows that the heterogeneous configuration of lidar + vision + IMU can make up for the shortcomings of a single sensor. The lightweight deep learning and multi-level fusion strategies work together to significantly improve the detection accuracy of small obstacles and the stability of dynamic target tracking of unmanned aerial vehicles. Its obstacle avoidance success rate, positioning accuracy, and environmental adaptability are all superior to the single-sensor solution. This conclusion provides a reference for the design and deployment of unmanned aerial vehicle autonomous systems and helps to promote the intelligent application of unmanned aerial vehicles in low-altitude economy and special operations.