Research On Channel Estimation Algorithm In Intelligent Reflecting Surfaces Systems

Higher channel capacity and inter-connection of more than 100 billion devices could be obtained in 5G system.However,the high complexity of the MIMO system,the high hardware cost and the increased energy consumption are still the key problems of 5G.Intelligent reflecting surfaces(IRS),composed of a large number of low-cost passive reflectors,is a new technology with lower complexity intelligent wireless communication environment by software-controlled reflection.In order to develop the potential benefits of IRS,channel state information(CSI)is essential in the design of both base station(BS)and IRS precoding.There are a large number of reflection units in IRS and signals can be only reflected passively.It is difficult to estimate the channel between each reflection unit and BS and between the reflection unit and each user with low complexity.In this paper,channel estimation is studied for both quasi-static channel and rapid mobility channel in the IRS system.The main work and achievements are as follows:(1)In the quasi-static channel scenario,a new IRS channel estimation algorithm based on extreme learning machines(ELM)is proposed to improve the defects of traditional IRS channel estimation algorithm.ELM network parameters are trained by past pilot signal so that the output matrix is calculated for CSI estimation.In order to solve the problem of hidden layer nodes in ELM,particle swarm optimization(PSO)algorithm is proposed for the input weights optimization and hidden layer bias correction in ELM system.Better accuracy can be achieved and the regression performance can be greatly improved.Simulation results show that the proposed ELM algorithm can effectively suppress the noise interference compared to the traditional algorithm.When the noise increases,system performance can be improved much better with the proposed algorthm than with the traditional on/off algorithm.Furthermore,compared to ELM algorithm,the proposed PSO-ELM algorithm is robust to the initial parameters,and the performance can still be improved.(2)In the rapid mobility scenario,a novel algorithm is proposed to improved to solve the problem of too many parameters to be estimated in dual-time scale channel estimation algorithm.In the stage of both estimating IRS-UE channel and BS-UE channel,the appropriate BEM model is used for simulating the fast time-varying channel,which reduces the number of estimation parameters and reduces the estimation complexity.In order to improve the accuracy of channel estimation,recursive least squares(RLS)algorithm is adopted to improve the performance of channel estimatio.Simulation results show that the proposed BEM algorithm is better than the traditional dual-time scale algorithm and the On/Off algorithm,higher channel capacity can be achieved.