Federated Learning in Satellite Constellations
Bho Matthiesen, Nasrin Razmi, Israel Leyva‐Mayorga, Armin Dekorsy, Petar Popovski
Abstract
Empowered by their exceptional versatility and autonomy, unmanned vehicles (UxVs), including ground, aerial, surface and underwater vehicles, are emerging as promising tools to execute tasks ubiquitously. Due to the increasing complexity of the environment and tasks, cooperation among multiple UxVs of different types is often required, which significantly relies on high-performance networking among them. However, the mobility, autonomy and heterogeneity of UxVs pose enormous challenges for UxV network management. This article summarises the promising perspectives and difficulties of the current and future UxV networks and proposes a software-defined management architecture for UxV networks to achieve wider coverage, reduced latency, and higher delivery ratio. The feasibility and advantages of this innovative architecture are further discussed, with preliminary experiment results confirming its superiority and potentials. Finally, future challenges and opportunities of networking and deployment are summarised.