Automatic Modulation Classification in RIS-Assisted Wireless Communication Systems using Ensemble Learning Techniques

This study investigates the application of ensemble learning techniques for automatic modulation classification (AMC) in a reflective intelligent surface (RIS)-aided wireless communication system. The transmitter is considered to utilize five possible phase-shift keying and quadrature-amplitude modulation schemes for data transmission. The receiver extracts cumulant-based and spectral-based features from the received data, and employs three ensemble classifiers, namely XGBoost, LightGBM, and Random Forest for AMC. Furthermore, the important features for each classifier are identified, and their performance is computed and compared with the classifiers using all the features. Numerical results show that the LightGBM classifier performs the best in terms of AMC for the considered system and effectively classifies the modulation schemes at low signal-to-noise ratio values.