AI-Augmented Metasurface-Aided THz Communication: A Comprehensive Survey and Future Research Directions

Abstract

Terahertz (THz) communication is becoming an enabling technology in beyond-5G and 6G networks with capabilities of ultra-high data-rates, sub-millisecond latency, and connecting devices in a massive number. This is disadvantaged by harsh propagation losses, lack of substantial scattering, and the requirement of the urgent beam orientation. Dynamic and reconfigurable wireless environments A powerful solution to those problems will be the use of intelligent metasurfaces engineered surfaces that are able to dynamically manipulate electromagnetic waves, in combination with artificial intelligence (AI). In this survey, a detailed analysis of AI-augmented metasurface-aided THz communication systems is to be given with a special concern on the linkage between electromagnetic design and data-driven intelligence. We categorize the available literature into architectures, AI, optimizations in the system and deployment. The question of the application of supervised, unsupervised as well as reinforcement learning to tasks that include the beamforming, channel estimation, and real-time metasurface control is put into focus. We select some of the key performance indicators and benchmark datasets and point out the limitations of the systems and the standards that are emerging. The paper ends with further research directions that need to be explored in the future, such as hardware-in-the-loop experimentation, federated AI control and physics-informed learning models to optimize metasurfaces. Our results indicate that the metasurfaces enabled by AI use are not only possible in technical perspective but essential to enable adaptable, scalable, and high-performance THz systems within next generation wireless networks.