SBIR Phase I: Software-Defined Networking and Resource Virtualization in Low Earth Orbit (LEO) Satellite Constellations

The broader impact/commercial potential of this Small Business Innovation Research (SBIR) Phase I project is the extension of state-of-the-art terrestrial 5G technologies to non-terrestrial satellite networks. Non-terrestrial networks (NTNs) are the ideal hosts for many commercial applications involving monitoring, reconnaissance, and remote sensing such as agricultural planting, remote factory operations, and industrial automation.

These applications will demand significant computation and bandwidth resources from NTNs. Virtualization technologies like software-defined networking (SDN) and network slicing are key enablers for similar applications in terrestrial networks, but satellite networks pose a unique set of challenges to existing 5G technologies. NTNs have complex and highly dynamic topologies caused by both the predictable movement of satellites as well as unpredictable weather events interfering with satellite-to-ground links.

Extending 5G algorithms to NTNs enables new applications and unlocks additional network capacity without launching any new satellites. 5G virtualization also simplifies access to satellite constellations. Customers can interact with multiple satellite networks via the same virtual network interface, eliminating the need to learn different interfaces for different constellations.

This Phase I project will explore more efficient networking solutions that satellite network operators can use to supply the growing demand for satellite internet.

This SBIR Phase I project explores methods of extending SDN and virtual network slicing technologies into space given the challenges posed by NTN topologies. First, it proposes a "NextG" framework for non-terrestrial SDN using existing 5G technologies whose core orchestrator can create end-to-end slices for seamless communication over terrestrial and NTN.

In SDN, one or multiple software entities called controllers are responsible for the control of the network. The number and locations of controllers in the network and how often these controllers communicate can be selected to balance latency in the NTN and the overhead costs from sending synchronization messages. This project explores the use of deep reinforcement learning algorithms to learn optimal controller placement and synchronization strategies.

Second, network slices are independent virtual networks that share a common infrastructure of network resources. Offline algorithms are used to allocate network resources to slices based on the requests of the virtual network operator. Subsequently, online algorithms dynamically reprovision network resources in real-time depending on the slices' actual resource usages.

The team will investigate integer programming techniques for virtual network embedding that scale to larger NTN topologies and deep reinforcement learning agents for online resource scaling.

This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.