Unmanned aerial vehicles (UAVs) are expected to be deployed in a variety of applications in future mobile networks due to several advantages they bring over the deployment of ground base stations. However, despite the recent interest in UAVs in mobile networks, some issues still remain, such as determining the placement of multiple UAVs in different scenarios. In this paper we propose a solution to determine the optimal 3D position of multiple UAVs in a capacity enhancement use-case, or in other words, when the ground network cannot cope with the user traffic demand. For this scenario, real data from the city of Milan, provided by Telecom Italia is utilized to simulate an event. Based on that, a solution based on k-means, a machine learning technique, to position multiple UAVs is proposed and it is compared with two other baseline methods. Results demonstrate that the proposed solution is able to significantly outperform other methods in terms of users covered and quality of service.

This work proposes an algorithm to optimize the positioning and the transmit power of Drone Small Cells (DSCs) based on Q‐learning, a technique where the agents learn to maximize a given reward. We consider two different rewards in this work, the first focusing on coverage, while the second maximizes the lifetime. Then, the Q‐learning solution determines the best positioning of the DSC in the 3D space, as well as the optimal transmit power. Results show that the optimization of the transmit power is of paramount importance to reduce the outage probability. In addition, we show that the second reward can considerably increase the network lifetime with a small penalty to the coverage.