Research On The Spectral And Energy Efficiency Of Full-duplex Massive MIMO Systems

With the improvement of processor and popularization of the embedded oper-ating systems,more flexible and powerful terminals are widely used in human life and business,and new applications are changing our living styles day by day.There are much more requirements of information exchange between people and things in now days than the past tens of years.With the attraction of the huge market,the 5th generation communication appeals to more and more researchers' interests.The key techniques of 5G include the network and wireless key techniques.Mas-sive MIMO and full-duplex(FD)are hot spots in the wireless key techniques.Both the massive MIMO and full-duplex can improve system spectral efficiency(SE).Massive MIMO can suppress interference and noise with its large number of anten-nas,this is because more parallel subchannels can be provided to transmit data,and much narrower beams can be formed to enhance the signal strength and avoid the interference.In addition,adopting simple linear precoding and detection methods,such as MRT/MRC and ZFR/ZFT,we can acquire very good performance.Massive MIMO also can proved high reliability,because more independent transceivers are adopted.if one or two transceivers do not work,the impaction to the system is small,no system outage happens,thus it's more robust than SISO and traditional MIMO systems.The FD system can transmit and receive data at the same frequency and time slot,.Since the FD transceivers can receive its own and others transceivers signals,it will be impacted by the so called "self interference"(SI)or "loop interfer-ence"(LI),so it is very important to the suppression of SI or LI in the FD system.Generally,the transmit.antenna and receive antenna are isolated with the physical isolation method to reduce the SI.However,the residual SI is still as strong as the desired signals.Fortunately,the transmit data have been known by the receiver,so we can use signal cancellation or other suppression methods to suppress it until its strength is as small as the noise,and thus the FD special efficiency outperforms the half-duplex system at last.In this paper,we study the full-duplex massive MIMO system.We first discuss the special efficiency of the FD massive MIMO system,then the SI suppression under the imperfect channel state information(CSI),and we also study the combined antenna FD massive MIMO system.At last we propose a multipair two-way full-duplex massive MIMO relay network,and analyse the power allocation in the users.Our main contributions are listed as follows:1.We propose a shared-antenna full-duplex massive multiuser MIMO system,it has the merits of both the massive MIMO and full-duplex,i.e.,the higher spectral efficiency and channel reciprocity.It also can save the costs of RF hardware.With the MRT/MRC and ZFT/ZFR linear processing,we prove that the SI can still be suppressed even thought the uplink and downlink have complete dependency.Unlike the separate-antenna system.the suppression ability is proportional to the square root of the antennas number in the array.We also derive the lower bounds of the uplink/downlink achievable rates,the bounds are very close to the Monte Carlo simulation.Then we compare the FD systerm and TDD half-duplex(HD)system,it is shown that the FD outperforms the HD system in terms of the special efficiency.(The contributions correspond to Chapter 2 and the first paper listed at the end of the dissertation.)2.We study the uplink SE of the shared-antenna full-duplex massive MIMO system.Unlike the separate-antenna system,we suppose there exist line-of-sight(LOS)SI.We prove that the SI can be suppressed by the linear processing method,and the Ricean interference channel can be converted into a Rayleigh interference channel.We also proposed an approximation method for the uplink achievable rate,with this approximation,a very succinct closed-form expression can be acquired,this result only rely on the channel distributed information(CDI),rather than the CSI.(The contributions correspond to Chapter 3 and the second paper listed at the end of the dissertation.)3.We study the SE and energy efficiency(EE)with the channel estimation and proved that the SI at the BS can still be suppressed by the linear processing method if imperfect CSI rather that the perfect CSI can be acquired.We propose a SI cancellation method,namely,with the estimation of the SI channel,the estimated SI can be subtracted before we used the large scale antenna array linear signal processing(LALP)suppression,thus the SI can be mostly removed.The simulation results show that.,extra 36 dB suppression ability can be achieved by using the combination method.We also discuss the SI channel precoding methods for the channel estimation,this precoding can decrease the estimation dimension of the SI channel.The lower bounds of the uplink channel achievable rate are also studied,simulation shows that the lower bounds are very close to the real values.At last,we optimize the channel training time and the uplink transmit power based on the approximation.(The contributions correspond to Chapter 4 and the third paper listed at the end of the dissertation.)4.We propose a combined antenna FD massive MIMO system,with the MRC/MRT and ZFR/ZRT linear processing,we derive the lower and upper bounds of the uplink/downlink achievable rates.This model allow users to work in both the FD and HD mode.Since the uplink and downlink can be interfered by its own transmitter,the system load changes with the fluctuations of the traffic.We study the impaction to the system SE with selection of the duplex mode at light load and heavy load.(The contributions correspond to Chapter 5 and the 4th paper listed at the end of the dissertation.)5.We propose a multipair two-way FD massive MIMO relay system.The amplify-and-forward(AF)protocol and MRC/MRT and ZFR/ZFT linear processing methods are used.And we prove that the LI can be treated as the i.i.d.AWGN when the antenna array scale is large enough.We also derive the asymptotic results of the relay transmit power and the SINR at the users.The succinct closed-form expressions of the SINR for the none power-scaling case are also derived.The optimal power allocation is given by using the Lagrangian multiplier method at the case of the identical user channel large scale fading coefficients.With the none identical large-scale fading coefficients,a GA method is developed to acquire the optimal results of the power allocation.(The contributions correspond to Chapter 6 and the 5th paper listed at the end of the dissertation.)...