Enhancing Dynamic Connectivity: Employing Machine Learning for Cognitive Link Adaptation in Narrowband and Wideband Networks

Abstract

The demand for faster data transfer rates rises along with the number of mobile devices, and the radio spectrum becomes more crowded. The forthcoming 5G wireless commu nication technology seeks to significantly enhance data speeds and spectrum efficiency by dynamically adjusting to fluctuating channel conditions. This research introduces a hierarchical multimode, multiband, multi-scheme machine learning methodology for the classification of several NBWF modulation schemes, including 2-PSK, 4-PSK, 8 PSK, 2-FSK, 4-FSK, 8-FSK, and CPM, with parameters such as modulation order (M =[2, 4, 8]), spectral efficiency (h = [1/2, 1/4, 1/8, 1/16]), and overlap factor (L = [1, 2, 3]), across diverse channel conditions. In addition, a Multimode Multiband Wideband Waveform Framework (WBWF) suitable for system communications is proposed. The framework utilizes the prototype filter design method to suppress interference between symbols. A self-optimizing hierarchial classifier is introduced for adaptive modulation and parameter setting which guarantees the best channel mode selection. The study also considers FBMC for WBWF scheme analyzing its respective parameters, including modulation order (M = [2, 4, 8]), subcarrier count (N = [128, 256, 512]), code rate (C = [1, 1/2, 2/3]), and overlap factor of Phydyas prototype filter (K = [2, 3, 4]) across sim ilar channel conditions as utilized in NBWF. Using a thorough MATLAB simulation, we modulated data with various parameters (M, h, and L) for NBWF, and (M, N, C, and K) for WBWF, transmitted them through AWGN and Stanford University Interim (SUI) channels, and used a variety of frequency ranges for demodulation. The bit error rate (BER) and data rate were assessed over various signal-to-noise ratio (SNR) values. Hierarchical classifiers were constructed to identify the modulation type and determine the individual parameters affecting throughput under different SNR situations. The suggested method achieved an accuracy of about 87.11% for NBWF and 78.47% for WBWF, which shows that it could be useful for cognitive radio systems and flexible NB/WB communication networks.