Abstract-In this letter, we propose a novel modularity-based dynamic clustering relying on modified Louvain method for UAVs aided mobile communications. Our aim is to save the transmit power of the mobile devices, by locating the UAVs vertically projected on the centroids of the user clusters. We further propose two types of operation for the modularitybased dynamic clustering, namely the recurring operation and the differential operation. We show that the proposed method requires substantially lower transmit power of the mobile devises and lower energy consumption of the UAVs than that required by the K-means based solution. We also show that the differential operation is more suitable for networks with lower proportion of moving users, since it consumes significantly less energy than that required by the recurring operation at the cost of requiring slightly higher transmit power of mobile devices.
I. INTRODUCTIONRecently, unmanned aerial vehicles (UAVs) aided mobile communications has drawn great attentions, thanks to its capability in providing better line-of-sight (LoS) connections with adjustable flying positions. The use cases of the UAVs aided mobile communications mainly cover emergency-responding services for both public and military areas [1]. Recent studies have investigated various air-to-ground channel models [2], flying altitude versus coverage trade-offs [3], energy efficient UAV transmission schemes [4], etc. Indeed, it is highly important to save the transmit power of mobile devices so that to prolong their usage in emergency scenarios. One promising approach is to locate the UAVs closer to the mobile devices for establishing shorter radio links. To elaborate, [5] showed that the transmit power of mobile devices can be substantially reduced by adapting the UAVs' positions based on the mobile devices' locations.In UAVs aided mobile communications, each UAV serves a cluster of mobile devices, where the clustering is typically based on the de facto K-means criteria. However, it is known from network science that modularity is the most used and best measure of the quality of clustering performance. Indeed, it has been widely studied in sociology, biology and computer science in terms of community detection [6]. Hence, we propose a novel modularity-based dynamic clustering for energy efficient UAVs aided mobile communications, relying on modified Louvain method in both recurring and differential operation to construct clusters. Specifically, after forming dynamic clusters, the UAVs are relocated to the positions vertically projected on the centroids of clusters.