<i>Learning to Fly</i>: A Distributed Deep Reinforcement Learning Framework for Software-Defined UAV Network Control

Control and performance optimization of wireless networks Unmanned Aerial Vehicles (UAVs) require scalable approaches that go beyond architectures based on centralized network controllers. At the same time, model-based is often limited by accuracy approximations relaxations necessary to solve UAV control problem through convex or similar techniques, channel models used. To address these challenges, this article introduces a new architectural framework optimize Deep Reinforcement Learning (DRL). Furthermore, it proposes virtualized, `ready-to-fly' emulation environment generate extensive data traces train DRL algorithms, which are notoriously hard collect battery-powered networks. The training integrates previously developed protocol stacks for UAVs into CORE/EMANE tool. Our virtual guarantees collection high-fidelity can be used agents. proposed architecture enables distributed data-driven (with up 3.7 × throughput improvement 0.2 latency reduction in reported experiments), facilitates reconfiguration, provides solution large