Collision Avoidance for Autonomous Unmanned Aerial Vehicles with Dynamic and Stationary Obstacles
Elif Ece Elmas, Mustafa Alkan · Journal of Polytechnic · 2025
The progress of UAV autonomous navigation is steadily progressing at a much faster pace, especially in the field of real-time collision avoidance. In this study, an all-in-one solution proposed that uses a LiDAR data analysis for voxel-based environmental model creation and an optical flow (OF) for predicting moving object trajectories. By fusing LiDAR’s high-resolution spatial data with the relative motion detection capability of the OF, proposed system enables detection and avoidance of both static and dynamic objects on-the-fly. In this regard, based on a holistic perception system, the UAV can adapt to any changes in the environment, and its navigational autonomy can be enhanced. Due to these multi-tiered components, this collision avoidance algorithm is able to efficiently provide safety for UAV operations in a vast number of various conditions. Experiment and simulation results indicate that the system is capable to keep the UAV’s flight path steady through various situations where low-light or high-speed components are involved.