Elevation-Aware BLER-Centric Link Adaptation for Dynamic LEO 5G NTN Systems

Link adaptation in fifth-generation (5G) NonTerrestrial Networks (NTN) faces unique challenges compared to terrestrial systems, including long round-trip latencies, large Doppler shifts, stale channel state information (CSI), and limited on-board computational resources. Conventional Outer Loop Link Adaptation (OLLA), designed for terrestrial deployments, struggles under these conditions. This paper introduces a Block Error Rate (BLER)-targeted link adaptation framework tailored for Low Earth Orbit (LEO) satellite systems. Unlike OLLA, which reacts to error outcomes through stepsize adjustments, our approach anchors adaptation decisions directly to explicit BLER targets. The framework integrates baseline comparisons, machine learning (ML)-enhanced lookup tables (LUTs), and over-the-air (OTA) validation on a commercial off-the-shelf (COTS) 5G testbed using the open-source platforms srsRAN and Open5GS. Simulation and OTA results show that the BLER-centric algorithm converges more quickly and achieves higher throughput than conventional approaches, particularly as satellites move from low to high elevation angles. ML-based refinements further improve performance in specific signal-to-noise ratio (SNR) regimes, while OTA experiments confirm the stability-throughput trade-off central to NTN link adaptation. These findings demonstrate that explicit BLER anchoring, supported by ML and OTA validation, offers a practical pathway for improving 5G NTN performance.