Predictive Routing for NB-IoT Uplink Traffic over a Very Low Earth Orbit (VLEO) Satellite Constellation

Abstract

Very low Earth orbit (VLEO) satellite constellations are a promising platform for uplink-dominated narrowband internet of things (NB-IoT) services. However, rapid topology variations and short satellite-gateway and satellite-user contact durations introduce significant routing challenges. This paper proposes a predictive routing framework that jointly accounts for future connectivity, heterogeneous delay requirements, and instantaneous congestion information. The routing problem is formulated over a finite prediction horizon with the objective of maximizing packet delivery ratio (PDR) and is addressed via a two-stage heuristic approach: a backward dynamic programming procedure is used to identify deadline-feasible forwarding options, while congestion-aware onboard decisions determine next-hop selection. Simulation results show that the proposed method improves PDR by up to 8% compared to contact graph routing and up to 33% compared to shortest-path routing, while effectively prioritizing delay-sensitive traffic.