Research On Key Technology Of Signal Detection For Massive MIMO Systems

Massive multiple-input multiple-output(MIMO)system is a critical technology for the next generation wireless communication systems due to its significant gains in spectral efficiency,energy efficiency,reliability,and coverage compared with the traditional MIMO system.However,the deployments of practical massive MIMO systems with large-scale antenna arrays result in huge challenges.The classical linear minimum mean square error(MMSE)signal detection algorithm exhibits high computational complexity in massive MIMO systems owing to the matrix inversion operations.The centralized baseband processing at the base station results in excessively high interconnect,chip input/output data rates,and detection computational complexity,while the recently proposed decentralized baseband processing(DBP)architecture can effectively alleviate the bandwidth bottleneck.Moreover,the multiple radio frequency chains at the transceiver bring expensive overhead and high complexity,and the recently proposed generalized spatial modulation(GSM)technique is able to reduce the RF overhead effectively.To solve the above-metioned issues,we explore the signal detection algorithms for massive MIMO systems.Low-complexity high-performance algorithms are designed based on the classical linear iterative algorithm and message passing algorithm,and the convergence as well as computational complexity analysis are also provided.The main contributions of this dissertation are shown as follows.1.We propose a low-complexity MMSE signal detection based on the accelerated overrelaxation(AOR)iterative algorithm.The theoretically optimal relaxation and acceleration parameters,convergence analysis,and computational complexity analysis are also provided.Simulation results illustrate that the proposed AOR algorithm outperforms the recently proposed iterative algorithms and approaches the MMSE detection.In addition,the proposed AOR algorithm exhibits approximate 78%computational complexity reduction compared with MMSE when the number of iterations and users are 3 and 32,respectively.2.We propose a decentralized Gaussian message passing(GMP)signal detection for massive MIMO systems with DBP architecture.A novel decentralized signal detection as well as message fusion rule are designed based on the factor graph and message passing rules.The state evolution and fix point are analyzed under the assumptions of large system limit and Gaussian sources.The computational complexity analysis is also provided.Simulation results illustrate that the nonuniform antenna cluster partition scheme exhibits higher BER performance than the uniform counterpart.Moreover,the proposed decentralized GMP algorithm outperforms the recently proposed decentralized algorithms.3.A joint signal detection based on expectation propagation(EP)algorithm is proposed for massive GSM-MIMO systems,which jointly detects the active transmit antenna indices and constellation symbol indices.To reduce the computational complexity of EP algorithm,we also propose a low-complexity two-stage(TS)-EP separate signal detection based on a pruned factor graph,which drastically simplifies the computational complexity of the symbol belief calculation.The multivariate complex Gaussian functions are replaced with the univariate counterparts.Numerical results show that the proposed EP algorithm outperforms the state-of-the-art counterparts.The proposed TS-EP algorithm achieves similar performance as that of the EP algorithm,and strikes a flexible performance-complexity tradeoff.To sum up,this dissertation designs low-complexity high-performance signal detection algorithms for massive MIMO system and its DBP architecture as well as GSM transmission scheme,and provides theoretical research basis of signal detection algorithms for the next generation wireless communication systems.