Time-Varying Channel Prediction For FDD Massive MIMO Systems

Massive MIMO(multi-input multiple-output)is a key technology in future wireless communication systems due to its higher spectral energy efficiency and link reliability.In order to take full advantage of the massive MIMO technology,accurate channel state information(CSI)is essential.Massive MIMO systems are generally assumed to operate in time division duplex(TDD)mode,due to the reciprocity between the uplink and downlink channels,avoiding the large pilot overhead required for downlink channel estimation.Applying massive MIMO technology to frequency division duplex(FDD)systems will achieve better transmission rates and reliability.However,in the FDD systems,the channel reciprocity is not applicable,so the huge pilot overhead required by the traditional downlink channel estimation scheme will seriously impact the performance of the systems.In addition,due to the movement of the user,the channel is usually time-varying,leading to the CSI obtained by the downlink feedback quickly out of date.In order to solve the problems sucn as large pilot overhead and high computational complexity of channel estimation for massive MIMO systems,this thesis reconstructs the uplink and downlink channel responses of massive MIMO systems by path parameters.The specific work content is as follows:Firstly,this thesis introduces the system model and common channel estimation algorithm for massive MIMO systems.It focuses on mapping multipath channels into three parameters: direction of arrival(DOA),time delay and fading coefficient,and extracts these three parameters in the uplink channel estimation.In addition,considering that the DOA and time delay are changing slower than the fading coefficient,this thesis assumes that the DOA and time delay remain constant for a period of time,and proposes a first-order Taylor expansion model for the fading coefficient,further proposing a RLS fading coefficient tracking algorithm with optimal forgetting factor.Finally,based on the first-order Taylor expansion model,this thesis proposes a fading coefficient prediction algorithm,and derives the interval of effective prediction.Then according to the physical path reciprocity between the uplink and downlink channels,construct the downlink channel response with the channel parameters.Decomposing the channel into three parameters effectively reduces the computational complexity of channel tracking and prediction,and avoids the huge pilot overhead required by downlink channel estimation for massive MIMO systems.