5G-Enabled UAV System: Implementation, Deployment and Analysis

With the growing adoption of Unmanned Aerial Vehicles (UAVs) across various industries, the need for reliable, real-time control of these systems has become more critical than ever. UAVs are now utilized in various use-cases ranging from civilian to military, many of which demand low-latency highly reliable communication between the drone and its control system. However, conventional communication networks like 4G LTE and earlier technologies often fall short of meeting the demanding needs of these applications. In this context, a key enabling technology is edge computing: in the edge computing paradigm, computation and data storage are brought closer to the location where they are needed, typically just outside the network core. This has the main benefit of reducing the latency of communication towards the User Equipment (UE) compared to using cloud servers and allows for real-time processing. Concerning UAV operations, edge computing also enables offloading resource-intensive tasks, such as trajectory calculation and image processing, to local servers, thereby reducing the computational load on the UAV itself and extending its battery life. This thesis seeks to demonstrate the deployment and performance of a 5G-enabled UAV system. The study aims to explore the challenges involved in its implementation and assess its capacity to deliver reliable and real-time control in UAV operations. The first aspect that is investigated is the deployment of an OpenAirInterface 5G network leveraging Ettus USRP B210 software-defined radios. After the correct operation of the network is validated, the thesis moves on to examine a second critical component: the PX4 Autopilot framework, an open-source flight control software providing a simulation environment and integration with Robot Operating System 2 (ROS2). Having laid the theoretical background, the complete deployment of the 5G-enabled UAV testbed is documented, detailing the development of the ROS2 UAV control application and proposing solutions to the issues that arise, especially regarding the ROS2 Middleware for communications. The experimental UAV testbed is then analyzed in terms of trajectory error, occupied network bandwidth and ROS2 application round-trip time (RTT). The study proposes a method to measure RTT based on talker/listener ROS2 nodes and researches the causes of the high variability of RTT over time by gathering and analyzing the behavior of RAN KPIs. Finally, trajectory error between the desired and actual flight paths is evaluated, in particular when varying angular velocity over the circular trajectory and control publishing frequency. The analysis identifies a critical angular velocity over which the UAV is not controlled effectively and suggests a value for the control publishing frequency. Since the thesis makes use of a simulated UAV, future work could concentrate on repeating the performance evaluation with an actual UAV, as well as applying a URRLC network slice to the UE and employing more advanced edge-computing frameworks (such as Multi-access Edge-Computing) to further improve trajectory error.