Capacity Analysis And Precoding Design In Massive MIMO Systems

To satisfy the requirement of 1000 times capacity in the fifth generation(5G)mobile communication systems,the radio transmit technology and network technology have been extensively researched.Due to the shortage of low frequency and the implementation difficulty of high frequency,massive multiple-input multiple-output(massive MIMO),which can improve the spectral efficiency of wireless communication systems,becomes an attractive topic in academia and industry.Massive MIMO,where each base station is equipped with massive antennas,can mitigate the interference by exploiting the huge degrees of freedom and schedule much more users to improve the system capacity and energy efficiency.However,the capacity under the favorable propagation channel model doe not curve the practical system performance and provide a valuable theoretical insight or optimization on precoding design.Thus,analyzing the capacity under practical scenario and design the efficient precoding schemes in massive MIMO systems deserve necessary research effort.In this dissertation,our research focus on the capacity analysis and precoding design in massive MIMO systems.By deriving the analytical capacities of point-to-point,multi-user(MU)and multi-cell massive MIMO system,some parameters optimization and theoretical elevation are exploited.Furthermore,to aim at the huge computational complexity,we design some efficient precoding schemes to satisfy the requirement of system rate.To achieve this challenging research,we have also learn and study a new mathematical tool——random matrix theory(RMT).Our main contributions are summarized as follows:(1)We derive a closed-form expression for the moments of finite-dimension non-central Wishart matrices in a point-to-point massive MIMO system.An exponential mean function is adopted to approximate the mean of capacity and an analytical expression of approximation function is obtained,which is based on the property of expectation of characteristic polynomials.The capacity analysis of massive MIMO system is established in the finite dimension regime via the concentration of spectral measure of random matrix.Two kinds of the significant probability measures are used to characterize the spectral distributions of random matrix.By constructing the objective function to satisfy the requirement of two spectral measures,the concentration of spectral measures and convergence probability are obtained.With the aid of concentration of spectral measure,both lower and upper bounds of capacity are analyzed are derived.The impact of parameters on capacity can be obtained based on this closed-form expression.(2)Considering the effects of spatial and temporal correlation jointly,the achievable rates of the uplink MU-MIMO and downlink multi-cell MIMO are derived.Firstly we construct the spatial-temporal model to describe the impact of spatial and temporal correlation on the received signal.In the single cell MIMO system,since signal-to-interference-plus-noise ratios of zero-forcing(ZF)receiver is too complex to analyze,both upper bound and lower bound of ZF receiver on achievable rate are obtained.Motivating by asymptotic equivalent analysis(AEA),we obtain AE expressions of these bounds and matched filter(MF)receiver.Depending on these AE expressions,the impact of spatial and temporal correlations are explored.Due to the characteristic feature of these two receivers in the sum rate and computation complexity,we propose a receive framework to balance the performance and computation complexity.By analyzing and transforming the characteristic feature of performance gap,we obtain an optimal analytical solution.Simulation results demonstrate the validity of our proposed scheme.On the basis of above work,AE expression of the downlink multi-cell massive MIMO is obtained.Different from single cell system,channel estimation error caused by pilot reuse is non-negligible.In time division duplexing,we construct a spatial-temporal model based on the minimum mean square error(VMSE)estimation and the received SINRs of MF,ZF and regularized zero-forcing(RZF)precoders are obtained under the spatial-temporal model.AEA can be applied directly to derive the AE rate with MF and RZF precoders.By analyzing these AE expressions,we find that AE SINR is independent with small-scale fading and dependent on large-scale fading and spatial-temporal correlation parameters.Simulation results shows that RZF and ZF precoders are more robust to the spatial correlation compared with MF precoder.(3)Considering the channel estimation error,precoding design of the multi-cell massive MIMO system is studied with the aide of AEA.Firstly,the transmission power minimization problem is formulated subject to the non-outage probability constraints,where Stochastic model is used to characterize the channel estimation error.In respect that the non-convex probability constraint makes the downlink precoding difficult to solve,Uplink-downlink duality algorithm(UDDA)is proposed to design precoding by scaling the non-convex probability constraint.To reduce the signaling overhead in massive MIMO system,a distributed algorithm based on large system analysis(DALSA)is proposed,which only needs the large-scale channel information.Since the Stochastic is such complex that it does not provide the intuitive insights,we also formulate a transmit power minimization problem subject to signal-to-interference-plus-noise ratio(SINR)constraints,where Gauss-Markov model is used to characterize the channel estimation error.An uplink-downlink duality scheme is employed to handle this problem.Then,we derive the AE SINRs of uplink and downlink system by employing AEA and propose a scheme to design the precoding matrix.Simulation results show that the proposed scheme performs well in reduce much information exchange,cumbersome computational complexity and possesses fast convergence rate.