Akademik Veri Yönetim Sistemi
IEEE Transactions on Cognitive Communications and Networking, cilt.12, ss.7926-7941, 2026 (SCI-Expanded, Scopus)
This paper presents intelligent blockage detection and mitigation techniques for near-field XL-MIMO systems, where the spatial non-stationarity and spot-like beam behavior exacerbate the impact of environmental blockages. By leveraging the power delay profile (PDP) shaped by spherical wavefronts across the array, we propose two complementary methods for blockage detection: a model-based approach using belief propagation within a Markov Random Field (MRF), and a learning-based approach using a tailored 3D CNN. Both methods operate without external sensing hardware. Based on the estimated blockage map, two self-healing beamforming strategies are explored: (i) a greedy antenna selection method that maximizes post-blockage SNR, and (ii) a codebook-based beam realignment approach. These techniques redistribute transmission energy away from blocked elements, enabling robust communication. Simulation results show that the model-based approach is robust under low-SNR conditions, while the CNN-based model excels at higher SNRs. Importantly, both approaches restore the spectral efficiency to near pre-blockage levels through adaptive beamforming, demonstrating their effectiveness in maintaining high-throughput links under dynamic blockage scenarios.