Research On MIMO Precoding With Imperfect Channel State Information

Multi-input multiple-output(MIMO)technology is one of key technologies for mobile communication system.By use of precoding technology,the MIMO system can eliminate the co-channel interference among users and improve the channel capacity and spectrum utilization rate.Generally,the perfect channel state information(CSI)should be known to system if precoding is used.However,it is easy to obtain CSI errors in both frequency division duplex(FDD)system and time division duplex(TDD)system,which is denoted by imperfect CSI.Therefore,it is important to study MIMO system precoding technology with imperfect CSI.In this thesis,the study includes four aspects.Firstly,the MIMO system model is investigated to establish the system framework and theory foundation for the precoding research.Secondly,the MIMO system precoding is discussed with perfect CSI,which will be a foundation for studying precoding with imperfect CSI.Thirdly,the MIMO system precoding technique is studied with imperfect CSI.The linear precoding and nonlinear precoding matrices are derived based on minimum mean square error(MMSE)criterion with imperfect CSI model.Fourth,since the CSI has important effect on the performance of precoding,this thesis expands the study based on the previous research and combines geometric mean decomposition(GMD)technology for channel matrix into the precoding.If we decompose channel matrix by use of GMD,the system can achieve extra gain.Therefore,combing GMD with previous non-linear precoding could improve the system performance with imperfect CSI.As a result,there are two novel ideas in this thesis.One,we propose GMD-THP precoding method by combining GMD and THP algorithm with imperfect CSI.We first decompose channel estimation matrix by use of GMD,and then derive the THP transceiver matrices based on decomposed matrices and MMSE criterion.The first and second order statistics information is used in the derivation.Two,we propose GMD-VP precoding method by combining GMD and VP algorithm with imperfect CSI.The precoding matrix and perturbation vector are derived by use of the second order statistics information of channel error matrix.Since the first and second order statistics information of channel errors are used in the derivation process,the designs of GMD-THP and GMD-VP show robustness to system.Computer simulation also proves that the two schemes could reduce the bit error rate and improve the system performance compared with the non-robust design.