Research on Active/Passive Hybrid IRS Assisted Communication Method Based on Deep Learning

With the rapid development of mobile communication technology, wireless communication systems have higher and higher requirements for spectral efficiency and energy efficiency. However, in the optimization process of many technologies, the wireless communication environment is still an uncontrollable factor, and is increasingly becoming a bottleneck for improving communication quality. Intelligent Reflecting Surface can reflect incident signals through a large number of low-cost reflective units, thereby improving the wireless communication environment. This paper introduces a solution based on deep learning method with the goal of maximizing the achievable rate of the system under the condition of intelligent reflecting surface hardware architecture of hybrid active/passive units. Firstly, the performance of communication system with intelligent reflecting surface and relay is compared. After that, the hardware architecture of intelligent reflecting surface with hybrid active/passive units is introduced, and the research scheme of intelligent reflecting surface based on deep learning is designed and simulated. The results show that the proposed scheme can approach the upper limit of the achievable rate of the system as much as possible while reducing the overhead of beam training.