Quantum-Driven Reinforcement Learning for Spectral Energy Optimization in Massive MIMO Hybrid Beamforming for 6G

Abstract The evolution of 6G wireless networks demands highly efficient beamforming strategies to optimize spectral and energy efficiency in massive MIMO systems. This study introduces a Quantum-Driven Reinforcement Learning (QDRL) framework for Spectral Energy Optimization in Massive MIMO Hybrid Beamforming for 6G, leveraging Quantum Deep Q-Networks (Q-DQN), Quantum Policy Gradient (QPG), and Quantum Approximate Optimization Algorithm (QAOA). The framework integrates mruby-based lightweight scripting for efficient deployment in edge-AI environments, enhancing computational flexibility and resource efficiency. Performance evaluations demonstrate that the Hybrid Quantum Model achieves 11.21 bps/Hz spectral efficiency, 97% resource utilization efficiency, and reduces energy consumption to 0.50 Joules/bit, outperforming classical models. The Bit Error Rate (BER) is minimized to 0.0025, and the convergence time is 48.7 s, significantly improving computational efficiency. Comparative analysis with conventional Deep Reinforcement Learning (DRL) techniques shows that the proposed quantum-enhanced model provides a 32% improvement in energy efficiency and a 21% reduction in computational complexity. The integration of mruby enhances the adaptability of the system in low-power and embedded environments, making it a viable solution for real-time 6G hybrid beamforming. This research highlights the transformative potential of quantum-assisted AI frameworks for scalable, high-speed, and energy-efficient wireless communication.