A Sensing-Assisted Environment Classification Technique for CF-mMIMO Enabled Non-Terrestrial Networks

Abstract

This letter proposes a sensing-assisted user environment classification framework for cell-free massive multiple-input multiple-output (CF-mMIMO) enabled non-terrestrial networks (NTNs) to accurately distinguish user environments under different channel and deployment scenarios. Specifically, a swarm of unmanned aerial vehicles (UAVs), coordinated by a high-altitude platform (HAP), transmits positioning reference signals (PRSs) embedded with radar-like sensing features. Thereby, the UAVs are allowed to infer propagation characteristics without increasing user-side complexity. Following this, the proposed framework jointly leverages spatial diversity, frequency diversity, and coding diversity to enhance detection robustness under multipath conditions and vertical-level differentiation of user positions by utilizing a bit-error-rate (BER)-driven classifier. Simulation results under standardized NTN channel models show up to 95% improvement in environment classification accuracy and significantly reduced false indoor detections compared to single-domain aggregation methods.