ScholarWorks@경상국립대학교

초록

A collision avoidance system is essential for the safe operation of multiple Unmanned Aerial Vehicles(UAVs). Previous studies utilizing sensors for UAV collision avoidance may exhibit performance variations depending on the sensor used, while reinforcement learning-based approaches face challenges related to hardware limitations and power consumption due to computational demands. This paper proposes a Digital Twin-based Collision Avoidance System(DT-CAS) that predicts potential collision points based on pre-determined flight paths in a digital twin environment and controls UAVs to avoid collisions. DT-CAS extracts flight path intersections of UAVs within the digital twin environment and controls the entry order to avoid collisions. DT-CAS can monitor UAVs for collision detection and avoidance with an average error of about 6.75cm. Also, DT-CAS was able to avoid collisions in scenarios with varying flight distances, speeds, and wind conditions, with only an about 5% increase in flight time and battery usage. © 2025 The Korean Society for Aeronautical and Space Sciences.

키워드