AI-Driven Terahertz Communication for 6G Networks: Challenges, Opportunities and Future Directions

Abstract

The large bandwidth capacity makes terahertz (THz) communication an important technology for the sixth generation (6G) wireless networks. It allows for transmission speeds of 1 terabit per second (Tbps) and facilitates data-intensive applications, including Holographic Communications, Extended Reality (XR) and Massive-scale Sensing. The implementation of a THz system is very challenging, however, because of high path loss, substantial molecular absorption, beam misalignment, limited coverage, hardware limitation and high dynamism of the channel condition. These issues are being solved by the use of Artificial Intelligence (AI) and Machine Learning (ML). They provide intelligent, adaptive and data-centric optimizations at multiple layers of communication stack. This article gives an overview of the AI based THz communication for 6G networks. We systematically explore the use of AI/ML techniques with key enabling technologies such as Reconfigurable Intelligent Surfaces (Reconfigurable Intelligent Surfaces), massive Multiple-Input Multiple-Output (MIMO), Non-Orthogonal Multiple Access (NOMA), and federated learning. The survey includes basic THz propagation properties, and cutting-edge AI solutions for beam management, channel estimation, resource allocation, energy efficiency and physical layer security. Studies have recently been reported that provide up to 120% improvement in secrecy rate and $2.5 \times$ in harvested energy. Limited amounts of data, real-time implementation, model interpretability and hardware compatibility are also extensively discussed. Opportunities for integrated sensing and communication (ISAC), on-chip wireless links, secure satellite communication and sustainable 6G architectures are highlighted. Last but not least, fully intelligent, adaptive and secure AI-native THz systems in 6G and beyond are highlighted by open research problems and future directions.