Research On Deep Iterative Network Based Channel Estimation For Reconfigurable Intelligent Surface-aided Communication Systems

Reconfigurable Intelligent Surface(RIS)is a promising candidate technology in future 6G wireless systems to cost-effectively implement Smart Radio Environment(SRE).In order to obtain the benefits brought by RIS,the RIS-assisted wireless communication system needs accurate channel state information(CSI)to control the wireless channel.However,it is difficult to accurately estimate the state information of the RIS auxiliary channel due to the following two reasons: 1)Since the RIS array is composed of passive devices,it does not have the ability to send,receive and process pilot signals;2)Adding RIS between user and base station,leads to a drastic increase in the dimension of the channel,which leads to large training overhead and high computational complexity.In order to achieve good channel estimation performance in general traditional algorithms,it is usually at the expense of higher computational complexity.Due to the powerful nonlinear mapping ability of deep learning technology,this paper builds a deep learning model to accurately estimate the channel with a lower complexity.The main work is as follows:RIS-aided wireless channel estimation in reflection mode.First,this paper models the RIS-assisted wireless communication system,then gradually embeds the domain knowledge and optimization algorithm of channel estimation into the neural network,and proposes a channel estimation method based on post-processing,a channel estimation method based on supervised network and a method based on self-supervised network channel estimation method.It is worth noting that self-supervised networks do not require real label channels in training,and can achieve nearly the same performance as supervised networks.Simultaneously Transmitting and Reflecting RIS-assisted wireless channel estimation.Compared with RIS with a single reflection mode,STAR-RIS can realize highly flexible full-space SRE.Therefore,this paper further studies the channel estimation problem of STAR-RIS assisted wireless communication system.Firstly,the communication system model assisted by STAR-RIS is constructed.Then the channel estimation problem is transformed into an optimization problem to be solved.Finally,the physical model and iterative optimization algorithm are integrated into the deep iterative network,and two deep iterative channel estimation methods based on GD-Net and deep iterative channel estimation methods based on ALISTA-Net are proposed.Simulation results show that,compared with traditional algorithms,the channel estimation method based on deep iterative network proposed in this paper not only has good generalization and robustness,but also achieves better estimation performance with lower pilot overhead.