Research On Efficient Precoding Techniques For Massive MIMO Downlink Systems

As one of the key technologies for 5-th Generation(5G)mobile systems,Massive Multiple-input Multiple-output(MIMO)technology have recently attracted considerable attention.By increasing the number of antennas in conventional MIMO systems,massive MIMO technology can achieve remarkable improvement of system capacity and spectral efficiency.However,since supporting too many data streams,the interference in the massive MIMO downlink systems is quite tremendous.And restricted by weak signal processing ability of mobile devices,it seems not feasible to employ signal detection technique at the user side.Therefore,precoding at the base station(BS)becomes the main approach to suppress the interference in the downlink.Based on the above background,this thesis focuses on the researches on the precoding technique in Massive MIMO downlink systems,including several conventional linear and non-linear precoding techniques.Meanwhile,the thesis puts the emphasis on an efficient non-linear precoding technique,i.e.,constant envelope(CE)precoding.The main content and contribution of the thesis are as follows.Firstly,an introduction and detailed analysis to several conventional linear precoding techniques are given in the thesis,including zero-forcing(ZF)precoding,precoding based on minimum mean square error(MMSE),maximum ratio transmission and precoding based on singular value decomposition(SVD),where the ergodic sum rate of each precoding is discussed.In addition,two non-linear precoding,i.e.,vector perturbation(VP)precoding and a variation on VP precoding that introduces the norm descent search(NDS)algorithm,are discussed.Secondly,a low-complexity non-linear precoding technique—constant envelope(CE)precoding is emphatically studied.Specifically,two fundamental challenges in CE precoding are discussed.The first challenge is the characterization of the region of all possible noise-free receive signals generated by CE precoding,and the second is the algorithm for phase recovery of CE precoding.The thesis also analyzes the criterion of choosing symbol energies,and it is pointed out that the optimal choice of the symbol energies depends on three parameters: normalized transmit power,the number of transmit antennas and the number of users.Finally,to meet the need of 5G,the performance of CE precoding in high mobility scenarios is analyzed.The impact of channel estimation error due to high mobility on CE precoded massive MIMO systems is discussed theoretically.As a result,a closed-form expression for the achievable ergodic information rate is derived.The system performance is evaluated in terms of both achievable ergodic per-user information rate and bit-error-rate(BER)by numerical simulations,which shows that an extra Gaussian-like noise term introduced by high mobility degrades the system performance.This constant noise term is independent of channel matrix and transmitted signals,and it is equal to the variance of channel estimation error in value.Therefore,by simply optimizing the channel estimate algorithm,the impact of high mobility on CE precoding can be reduced.