MACHINE LEARNING-BASED OPTIMIZATION APPROACH TO MITIGATE MULTIPATH IN COMPLEX WIRELESS NETWORK ENVIRONMENTS

Multipath propagation remains a major challenge in wireless communication systems, causing signal degradation, fading, and reduced transmission reliability. Traditional antenna diversity techniques, while effective to some extent, suffer from fixed configurations that lack adaptability to dynamic channel conditions. This paper presents a machine learning-assisted optimization framework for antenna diversity to address multipath propagation in complex wireless environments. Using Simulink, real-world data were simulated based on key wireless quality metrics Signal-to-Noise Ratio (SNR), Bit Error Rate (BER), and Throughput. A fuzzy logic-based rule system was developed to establish relationships among these parameters, which was then trained on an Artificial Neural Network (ANN) capable of dynamically selecting the optimal antenna configuration spatial, polarization, selection, or pattern diversity according to real-time channel variations. Simulation results across urban, industrial, and vehicular scenarios demonstrated significant performance gains over conventional diversity systems, achieving an average SNR improvement of 43.02%, throughput increase of 48.13%, and BER reduction of 16.63%. These results confirm that integrating machine learning enhances system adaptability, signal quality, and reliability in multipath-prone environments, establishing machine learning-based antenna diversity optimization as a robust and intelligent solution for next-generation 5G and IoT communication systems.