A hybrid deep reinforcement learning and genetic algorithm approach for optimized resource allocation and reconfigurable intelligent surface deployment in 6G holographic communication

Abstract

The advent of 6G holographic communication demands extremely high data rates and ultra-low latency, creating unprecedented challenges for resource allocation and Reconfigurable Intelligent Surface (RIS) configuration. To address these requirements, we propose a novel hybrid framework that integrates deep neural networks (DNN), genetic algorithms (GA), and deep reinforcement learning (DRL) to jointly optimize resource allocation and RIS deployment. The DNN first predicts future channel states and holographic traffic requirements, enabling the GA to perform proactive multi-objective optimization. The DRL agent then refines these solutions in real time, adapting to dynamic channel and traffic variations. Simulation results demonstrate the superiority of the proposed method over benchmark approaches. The hybrid DNN-GA-DRL achieves an average throughput of 6.55 Gbps, a latency of 0.1 ms, energy efficiency of 9.5 × 10 8 bits/Joule, and spectral efficiency of 6.0 bps/Hz, while maintaining near-perfect QoS satisfaction (99.8% throughput, 99.9% latency). Compared to state-of-the-art GA, DRL, and hybrid baselines, our framework consistently delivers higher efficiency, lower delay, and superior adaptability. These findings confirm the effectiveness of combining predictive intelligence, evolutionary search, and real-time reinforcement learning for enabling seamless, immersive holographic communications in future 6G networks. • Novel Hybrid Optimization Framework: Proposes an original integration of Deep Learning (DL) for channel and traffic prediction, guiding a Genetic Algorithm (GA) and Deep Reinforcement Learning (DRL) for dynamic resource allocation and Reconfigurable Intelligent Surface (RIS) configuration in 6G holographic communication. • Addresses 6G Holographic Communication Challenges: Directly tackles the critical requirements of 6G holographic communication, specifically ultra-high data rates and ultra-low latency, by optimizing resource management in dynamic environments. • Demonstrated High Throughput and Efficiency: Simulations show the approach achieves peak throughputs of approximately 1.4 x 10ˆ10 bits/second, with high energy efficiency (up to 1.6 x 10ˆ9 bits/Joule) and spectral efficiency (up to 9 bps/Hz). • Significant Latency Reduction Achieved, Yet Challenges Remain: Average achieved latency is significantly reduced to 2.50 ms, validating the hybrid system's efficiency. However, this still frequently exceeds the sub-millisecond requirements for holographic applications, highlighting a persistent challenge. • Robustness in Dynamic Environments: The hybrid approach demonstrates adaptive capabilities to varying channel conditions and traffic demands, maintaining a 78.50% overall SLA compliance rate for holographic QoS requirements. • Insights into DNN Prediction Limitations: Reveals that while the DNN accurately predicts stable channel parameters, its performance is limited for highly stochastic elements like shadowing, fast fading, and fluctuating latency/jitter requirements.