Researches On Performance Analysis And Transmission Design For Massive MIMO Systems

Currently,the amount of mobile communications service increases rapidly and the amount of mobile terminal and its type reaches the unprecedented scale.To cope with this situation,academia and industry have been planning the future-oriented mobile communication system,i.e.,the fifth generation mobile communication system(5G).In the setting of 5G standard,there has high requirements on transmission rate,transmission rate,density of connectivity,spectral efficiency,energy efficiency.In the current situation of wireless resource,to meet the requirements of 5G standard,a bunch of new wireless technologies which can efficiently develop potential of time-frequency resource are coming up.In these new technologies,massive MIMO(or large-scale MIMO)is very promising.By increasing the number of antennas equipped on base station(BS),massive MIMO is able to greatly promote the energy efficiency,spectral efficiency and transmission rate,without adding extra time-frequency resource.However,massive MIMO still has some problem needed to be solved.Therefore,researches on performance analysis and transmission design for massive MIMO systems are carried out in this paper.The main contents and contributions are as follows.Firstly,we conduct the non-asymptotic analysis of capacity of massive MIMO.Currently a large number of works on massive MIMO assume that massive MIMO systems are equippedwith infinite antennas.Nevertheless,a massive MIMO system cannot have infinite number of antennas in practice.Different from previous works resorting to asymptotic random matrix theory,concentration inequalities are used in our non-asymptotic analysis for massive MIMO systems with a finite number of antennas.We first derive deterministic bounds on the ergodic capacity for massive MIMO systems.Furthermore,we obtain statistical bounds within which the instantaneous capacity falls with a falling-in probability.As the number of antennas increases,the statistical bounds on the instantaneous capacity become tighter and the corresponding falling-in probability increases,which reveals an interesting phenomenon that the instantaneous capacity has fewer random fluctuations as the number of antennas becomes larger.Simulations show that these theoretical bounds match well with our experimental results.Secondly,we conduct the research on the minimum number of antennas required to satisfy outage probability in massive MIMO systems.We first apply the asymptotic tool to derive the asymptotic probability distribution function of instantaneous capacity of massive MIMO.Based on it,we obtain the minimum number of antennas required to satisfy outage probability in closed form.However,the asymptotic results is not very accurate,since the asymptotic assumption is not practical in practical systems.Thus,by exploiting the statistical bounds on instantaneous MIMO capacity,we derive the equivalent problem of determining the number of antennas needed to satisfy outage probability constraints.Through solving the equivalent problem,we obtain the minimum number of antennas required under pre-specified outage probability requirements.Analysis and simulation results demonstrate that,in massive MIMO systems,the required minimum number of antennas obtained by the proposed method is more accurate than that obtained by the asymptotic method.Thirdly,we study the optimal ratio of receive antennas and transmit antennas in full-duplex massive MIMO systems.Different to the previous works,we investigate the problem of antenna ratio under imperfect channel estimation.We consider a full-duplex systems where the BS exploits zero-forcing linear processing.We first derive a tight lower bound on the sum-rate for systems with large number of antennas.By resorting to the tight lower bound,we formulate a sum-rate maximization problem in terms of antenna ratio.Since the optimal problem to maximize sum-rate is a concave function,the optimal solution can be obtained easily.Fourthly,we research on the sum-rate performance of massive MIMO under dynamic TDD scenario.Dynamic TDD is plagued by the cross-link interference of adjacent cells.To suppress it,we consider dynamic TDD with massive multiple-input-multiple-output(MIMO).We derive tight deterministic equivalents of the uplink and downlink SINR.Exploiting these deterministic equivalents further,we analyze cross-link interference and then reveal that their effects both scale down as O(1/ M).Consequently,the uplink-downlink decoupling would be achieved in dynamic TDD systems by combining massive MIMO.Based on the analysis results,we also propose a power control scheme for maximizing the total sum-rate,which is based on geometric programming.Lastly,we research on low-complexity analog beam selection for hybrid beamforming.Based on machine leaning,we propose a data-driven method of analog beam selection to achieve a near-optimal sum-rate with low complexity.To be more specific,we take the beam selection problem as a multiclass-classification problem,where a large amount of samples of mm-wave channel are considered as training data.Based on these training data,we exploit support vector machine to obtain a statistical classification model in terms of maximizing sum-rate.During the real-time transmission,by using the derived classification model,we are able to select the optimal analog beam for each user with low complexity.Analysis and simulation results reveal that,as long as the training data is sufficient enough,the proposed data-driven method is able to achieve a near-optimal sum-rate performance,while the complexity would be reduced by several orders of magnitude,compared with exhaustive search.