Design And Performance Analysis Of Downlink Pre-coding In Massive MIMO System

Massive multiple input multiple output(MIMO)is one of the key technologies for the 5th generation(5G)mobile communication system.Compared to the traditional MIMO,tens of antennas are equipped in base stations in massive MIMO system in order to achieve increase in spectral and energy efficiency.However,the base station requires to deal with a large amount of information due to the large antenna array.As a result,pre-coding techniques may involve high complexity computations.In order to address this issue,new pre-coding strategies are designed in this thesis.These strategies can achieve good performance with a low complexity.The main contributions in this thesis are as follows.1)Zero forcing(ZF)pre-coding is involved with the inversion of high-order Gram matrix in massive MIMO system.In order to address this issue,the Neumann series is introduced to approximate the matrix inversion by the sum of matrix polynomials.In this way,the computation complexity of matrix inversion is reduced.For convergence acceleration,the Neumann series is further combined with Newton iteration method.Furthermore,a high probability convergence condition is derived for the joint algorithm.Simulation results demonstrate that the Neumann-Newton iteration method converges faster than the existing Neumann series.2)In order to reduce the complexity of ZF pre-coding,a new GS(Gauss—Seidel)-based matrix inversion approximation(GSBMIA)approach is proposed to approximate the Gram matrix inversion directly.Furthermore,a joint algorithm based on the GSBMIA and Newton iteration method is proposed for convergence acceleration.The GSBMIA approach is employed to provide an efficient searching direction for the following Newton iterations.As a result,the performance of Newton iteration method can be improved at the beginning of the iteration,thus accelerating convergence.Simulation results demonstrate that the joint algorithm has a faster convergence speed than the Neumann series and Newton iteration method.Moreover,the joint algorithm still has a good bit error rate(BER)performance when the ratio between base station antenna number and user number becomes relatively low.3)The joint pre-coding and power control problem in multi-cell circumstances is studied in order to obtain the optimal pre-coding and power control strategy.First,for the reduced optimization problem,a necessary condition of the optimal solution is derived based on the user's signal-to-interference-plus-noise ratio(SINR).With the help of the condition,the reduced optimization problem is transformed to a Perron-Frobenius eigensystem.Then,for the further optimization with respect to the pre-coding matrix,the uplink and downlink duality and the property of the non-negative irreducible matrix are then leveraged.Finally,the global optimal pre-coding and power control policy is obtained via the alternative optimization.Simulation results demonstrate that the algorithm converges to the optimal value very fast.Compared with the matched filter(MF)pre-coding and ZF pre-coding,the optimal pre-coding and power control policy has an obvious advantage in performance.4)The joint pre-coding and power control problem in multi-cell circumstances is studied via asymptotic analysis.Firstly,the asymptotic expression of the dual uplink SINR is derived via the random matrix theory.The optimization problem is then reconstructed in the asymptotic field.Furthermore,the optimal solution i.e.,the asymptotic dual uplink power,is derived via the standard power control algorithm and the nonlinear Perron-Frobenius theory.As the asymptotic dual uplink power only depends on the second order statistic,it does not require updating along with the fast fading channel instantaneously.Hence,the frequency of data sharing between cooperative base stations decreases drastically.Simulation results demonstrate that the asymptotic dual uplink power can reflect the trend of the optimal dual uplink power,and thus can achieve the near-optimal performance.