Performance Analysis And Optimization Of Massive MIMO With Low-Resolution ADCs

According to the applications of the fifth generation(5G)system,three typical usage scenarios was defined,i.e.,enhanced mobile broadband(eMBB),massive ma-chine type communication(mMTC)and ultra-reliable and low latency communication(URLLC).To reach the requirements of the three scenarios,a series of performance indi-cators was proposed,such as,increasing throughput by 1000-fold and improving energy efficiency by 10-fold.Massive multi-input-multi-output(MIMO)equipped with low-resolution analog-to-digital converter is a potential technology to meet these objects.Therefore,it is of great importance to investigate the performance analysis and the pa-rameter optimization of massive MIMO with low-resolution analog-to-digital convert-ers(ADCs).This dissertation focuses on the capacity analysis and capacity-based parameter optimization of massive MIMO with mixed-ADC structure,and choose sum rate as the measurement of system capacity.Random matrix theory is applied to derive the sum rate of mixed-ADC structure and provide theoretical support for performance analysis and optimization.Large-scale optimization theory is utilized to design efficient algorithms to optimize the transmit power and resolutions of the ADCs.The main contributions of this dissertation are:· In narrow band system,a closed-form asymptotic equivalent of the sum rate is de-rived,based on random matrix theory when minimum mean square error(MMSE)detector and additive quantization noise model(AQNM)is considered.Based on this closed-form asymptotic equivalent,it is found that capacity ceiling effect still exists when MMSE detector is applied.Increasing transmit power to infinity can not entirely compensate for the sum rate loss brought by low-resolution ADCs.Numerical results verify the tightness of the closed-form asymptotic equivalent.It is also found that massive antennas are able to compensate for the loss of low-resolution quantization,especially for 1-bit quantization.· In narrow band system,the effect of correlation of quantization distortion is con-sidered.Bussgang decomposition is used for the quantization process.Then an expression of the sum rate is derived.Based on this expression,a resolution opti-mization problem with maximal sum power constraint is formulated.This prob-lem is a non-convex integer programming and the first-order gradient is difficult to calculate.This dissertation use barrier function method and coevolution coop-erative particle swarm optimization(CCPSO)to solve this problem and propose an efficient optimal solution algorithm.Numerical results show the effectiveness of our proposed algorithm,and also find that using AQNM may not bring correct result in practical.·In wideband system,the expression of sum rate is derived when AQNM is ap-plied.This expression implies that the non-linear quantization brings the inter-subcarrier interference in orthogonal frequency division multiplexing(OFDM),thus leads to a decrease of sum rate.Based on this,an uplink transmit power allocation problem is formulated to maximize the sum rate.This problem is a difference of convex programming with large problem scale.Dealing with these two features,this dissertation use successive convex programming,Lagrangian duality,and bundle-level method to propose efficient solution algorithms with and without quality of service(QoS)constraints,respectively.Our proposed al-gorithm outperforms conventional subgradient descent,both in speed and preci-sion,which shed lights on the practical application of large-scale optimization in future wireless systems.