The interactions between the areas of mobile networking and drone robotics are currently attracting significant attention from both the robotics and the telecommunications engineering communities. The key scenarios where such interactions manifest themselves include the so-called flying base station or flying radio access network (FRAN) on the one hand, and the drone as a flying terminal on the other hand.
The use of drones and unmanned aerial vehicles (UAV) as FRAN nodes is rapidly emerging as a powerful tool to complement traditional fixed terrestrial deployments. The advantage of using UAVs will be particularly felt in those use cases where being able to quickly deploy a network where and when it matters is critical. However, the success of so-called FRANs hinges on the ability of the UAVs to place themselves spatially in an efficient and autonomous manner.
Having this in mind, the first part of this thesis aims to investigate current works and technologies of UAV-assisted wireless communications and develops novel methods for both the placement and trajectory design of a UAV as a flying RAN in the wireless networks for both mobile broadband coverage scenarios and IoT data harvesting scenarios. We highlight how the exploitation of city 3D maps can bring about substantial benefits for the reliable self-placement of flying radios. A suite of methods are presented that lie at the cross-road between machine learning and traditional communication theoretic network design.
Regardless of the placement or trajectory design, all the algorithms operate on the basis of an array of side-information such as node GPS location, the 3D map of the city, and terrain-dependent propagation parameters allowing the prediction of radio signal strengths. While such data may be collected via the network beforehand allowing placement or the trajectory to be optimized before the actual UAV flight,
part or all of the information may also have to be discovered or learned by the UAV. In this regard, a part of this thesis is devoted to discussing how to learn and estimate such information just from the UAV-borne measurements.
Assuming the availability of safe cellular connectivity beyond visual line of sight, UAVs are becoming appealing solutions for a wide range of applications in the areas of transportation, goods delivery, and system monitoring. All these use cases pertain to the UAV as an aerial terminal scenario, the availability of a reliable radio link is essential to make sure the drone can be guided effectively towards completion of its mission. The main challenge however in these domains is the design of trajectories which indeed can guarantee reliable and seamless cellular connectivity all along the path while allowing the completion of the UAV mission which is today a lacking feature of existing technologies. Therefore, in the second part of this thesis, we propose a novel approach for optimal trajectory design between a pre-determined initial location and a given destination point by leveraging on a coverage map. The coverage map can be obtained with a combination of a 3D map of the environment and radio propagation models. We establish a graph theory-based framework to evaluate the feasibility of the problem and obtain a high-quality approximate solution to the optimal UAV trajectory design problem.
Lastly, we come back to the FRAN scenario and discuss the experimental verification of the placement algorithm of a UAV relay in LTE networks and we present the design of the Rebot (Relay Robot) prototype. The Rebot functions both as an outdoor LTE relay between ground users and a fixed base station, as well as an autonomous robot capable of positioning itself at a throughput maximizing location.
