AN AI-DRIVEN RECONFIGURABLE METAMATERIAL ANTENNA FRAMEWORK FOR INTELLIGENT BEAMFORMING, ENHANCED SIGNAL COVERAGE, AND ENERGY-EFFICIENT 5G/6G WIRELESS COMMUNICATION SYSTEMS

Abstract

This study proposes an artificial intelligence (AI)-driven reconfigurable metamaterial antenna framework for intelligent beamforming, enhanced signal coverage, and energy-efficient fifth-generation/sixth-generation (5G/6G) wireless communication systems. The framework integrates a compact metamaterial antenna array, electronically controllable unit cells, channel-state sensing, and a deep reinforcement learning controller that dynamically selects radiation patterns, phase states, and transmit-power levels under changing propagation conditions. A simulation dataset containing 240,000 channel realizations was generated across urban-macrocell, urban-microcell, indoor-office, and high-mobility scenarios. Each record incorporated operating frequency, user location, angle of arrival, signal-to-noise ratio, path loss, interference level, mobility, channel-gain coefficients, beam direction, power consumption, and achievable data rate. The dataset covered 28, 39, 60, and 100 GHz frequency bands and was divided chronologically into 70% training, 15% validation, and 15% testing subsets. Data normalization, outlier screening, channel augmentation, and leakage-free partitioning were applied before model development. The proposed controller was compared with conventional phased-array steering, particle-swarm optimization, and a non-reconfigurable antenna baseline. Electromagnetic and link-level simulations indicated an impedance bandwidth of 27.8%, peak realized gain of 14.6 dBi, radiation efficiency of 91.3%, and beam-steering range of −60° to +60°. On the independent test subset, the AI controller achieved 96.4% beam-selection accuracy, reduced mean beam-alignment error to 2.1°, and improved average received signal strength by 8.7 dB relative to the conventional baseline. Signal-coverage probability increased from 78.6% to 94.2%, spectral efficiency rose by 31.5%, average throughput increased by 28.9%, and outage probability declined by 43.7%. Adaptive power control reduced antenna-system energy consumption by 24.8% and improved energy efficiency by 36.2%, while maintaining a median decision latency of 1.8 ms. Statistical significance testing confirmed that the principal gains were consistent across all evaluated environments, while ablation experiments demonstrated that combining metamaterial reconfiguration with learning-based power allocation outperformed either mechanism when it was applied independently alone. These findings demonstrate that jointly optimizing antenna reconfiguration, beam direction, and transmission power can provide reliable, low-latency, and sustainable connectivity in complex 5G/6G environments. The framework offers a scalable foundation for intelligent base stations, vehicular networks, massive machine-type communications, and future integrated sensing-and-communication platforms.