A Comparison of Clustering vs YOLO for Drone Swarm Centroid Detection

The rapid advancement of drones has led to the emergence of drone swarm applications in various domains, including surveillance, search and rescue, and package delivery. Efficient coordination and formation control of drone swarms is crucial for accomplishing complex tasks. The research presented in this paper proposes a novel approach for detecting individual drones in drone swarm formations through the utilisation of the K-means clustering algorithm. The algorithm assigns drones to the nearest centroids, creating cohesive subgroups and optimizing formation quality. To assess its efficacy, a comparative analysis is conducted between the K-means clustering algorithm and you only look once (YOLO) based computer vision detection algorithm. Through extensive simulation experiments, it is found that the K-means clustering algorithm outperforms the YOLO-based detection algorithm in terms of formation quality and computational efficiency. It consistently achieves more accurate and stable swarm formations making it suitable for real-time swarm control applications. The results of this presented research open the way for the effective use of drone swarms in a wide range of real-world applications and contribute to the development of cutting-edge swarm control approaches.