ML-Optimized THz-Band Massive MIMO Architecture for Energy-Efficient 6G Communication

Terahertz-band massive multiple-input multiple-output (MIMO) systems, a fundamental enabling technology of the sixth-generation communication, have a tremendous bandwidth capacity, yet with serious path losses, molecular absorption, and high-power consumption. To overcome these difficulties, it shows an ML-optimization-based THz-band massive MIMO system which incorporates learning-based channel modeling, hybrid beamforming, and resource-controlling by reinforcement. The framework utilizes absorption-aware channel forecasting, adaptive beam selection, and energy-conscious RF chain activation to optimize spectral and power efficiency jointly. The proposed system achieves an effective beamforming gain of 41.8 dB, a spectral efficiency of 212.4 bps/Hz, and maintains a downlink throughput of 1.18 Tb/s. With 34% active RF chains, it is possible to achieve energy efficiency of 18.7 Gbit/J, and the beam alignment latency is minimized to 72 μs. The probability of a beam misalignment is 0.0018, and the probability of a link outage is lower than 10-6, which is very reliable. These findings prove the adequacy of ML-aided THz massive MIMO to scalable energy-efficient 6G communication.