Design of RHS assisted Wireless Communication Systems

Abstract

Next-generation 6G networks support ultra-reliable low-latency communications (URLLC) with short packet transmission, while operating at millimeter-wave and terahertz (THz) frequencies under stringent energy and computational constraints. These requirements challenge classical long-blocklength assumptions and motivate new designs based on intelligent surfaces and robust multiple-access strategies. This thesis develops a unified framework for intelligent-surface-assisted wireless communications, integrating energy-efficient optimization, finite-blocklength information theory, and learning-based real-time control.

First, energy-efficient downlink transmission is studied for IRS-assisted THz systems employing rate-splitting multiple access (RSMA). An energy-efficiency maximization problem is formulated under realistic propagation models, hardware power consumption, Quality-of-Service constraints, and quantized IRS phase control. Successive convex approximation and a bio-inspired salp swarm algorithm are investigated, with complexity and convergence analysis. Results show that IRS-assisted RSMA consistently outperforms SDMA and NOMA in interference-limited and QoS-constrained regimes.

Second, finite-blocklength information-theoretic limits are established for reconfigurable holographic surfaces (RHS) with joint amplitude–phase control. Using angularly sparse channel models and Mellin-transform techniques, tractable expressions for mutual information and information dispersion are derived, enabling Berry–Esseen based achievability and converse bounds on the coding rate. The results quantify the impact of SNR, blocklength, and target reliability, and demonstrate that amplitude–phase holography reduces dispersion compared with phase-only surfaces.

Finally, RHS configuration is formulated as a graph learning problem. A reliability-aware graph attention network is proposed, achieving near-optimal performance with low inference latency suitable for real-time URLLC deployment.