Deep Reinforcement Learning for RIS-Aided Full-Duplex Systems: Advances and Challenges

Deep reinforcement learning (DRL) has gained significant attention in recent years as a powerful approach for solving complex optimization problems. One of the promising applications of DRL in wireless communication is full-duplex (FD) reconfigurable intelligent surface (RIS)-assisted wireless systems, which has emerged as a potential solution for the next-generation wireless communication networks. FD-RIS-assisted systems can simultaneously transmit and receive data using the same frequency band, which can significantly improve the system capacity and spectral efficiency. This article provides an overview of the DRL background and its applications in FD-RIS-assisted communication systems. It discusses recent research advances in various scenarios, including resource allocation, sum-rate optimization, and secure communications. Furthermore, it investigates the DRL performance in optimizing large-scale FD-RIS-assisted systems. Major challenges and shortcomings of DRL in FD-RIS-assisted wireless systems are presented and supported through numerical simulations. Based on this discussion, the article highlights prospective use cases that can bring the FD-RIS-assisted systems into practice.