Optimizing K-Means for the Deployment of UAV-Assisted Wireless Networks: A Parametric Analysis

Presently, wireless communications have become a very popular and technologically advanced sector in the realm of telecommunications. Due to their comparatively lower cost and higher mobility, unmanned aerial vehicles (UAV s) are now being considered as substitutes for traditional terrestrial base stations to provide coverage in areas without or lacking communication infrastructure. However, to make the use of UAV s as “flying base stations” a reality, a lot of challenges ahead to handle efficiently. One of the challenges is the positioning of multiple UAVs. The k-means clustering algorithm is exploited for UAV deployment. However, the impact of various parameters of k-means algorithm on the performance of UAV-assisted networks is yet to be investigated. In this paper, we investigate the impact considering various distance calculation methods, initial centroid determination methods, and so on. Through simulations, we find that the network performance varies significantly with the parameters. For a larger number of UAVs (12 or more), the impact of distance calculation and initial centroid position is significant. Furthermore, user throughput and satisfaction ratio are also affected. The k-means++ and sample method for initial centroid selection outperform other methods for average user throughput and user satisfaction ratio, respectively.