Intelligent Reflecting Surface Enhanced Wireless Communications With Deep-Learning-Based Channel Prediction

Over the past decade,the growth of data transmission rates and high demand have driven advances in wireless communications technology.MIMO is of outstanding significance for 5G technology and promises to enable high spectrum efficiency,while the deployment of large-scale antenna arrays typically results in high implementation costs and increased power consumption.IRS is a novel and disruptive technology that could enable spectrum and energy efficient wireless communications in the future at low cost.And recently,advances in deep learning techniques have attracted increasing attention,with possible implications in wireless communications.In this paper,a typical single-antenna BS,a single-antenna user is deployed for an IRS-enhanced wireless communication system.Channel prediction is introduced in order to reduce the channel estimation overhead.As the elements in IRS are usually arranged in two dimensions,this paper draws an analogy between channel prediction and computer vision.Furthermore,Unmanned Aerial Vehicles can effectively act as temporary base stations or aerial access points for data transmission/reception with ground terminals.However,the link between UAVs and ground users may be blocked by ground obstacles(e.g.,buildings in urban areas),resulting in poor data transmission or rate performance.To address this issue,reconfigurable intelligent surface serves as a promising technology to intelligently reflect the received signals between UAVs and ground users to significantly improve the communication quality.This paper further presents a theoretical framework to analyze the performance of an integrated IRS-UAV relay system,where IRS enables more flexible deployment of the system,combined with flexible deployment of fullduplex UAVs to enhance communication between ground nodes.We maximize the rate by optimizing reflection coefficients and trajectory through gradient descent optimization algorithm considering channel state information which is brought by channel prediction.We also investigate these two models effectively improve the performance of wireless communication through numerical results.IRS aided wireless communication provides a green and sustainable solution to increasing capacity and higher spectrum utilization requirements.