Low Complexity Receiver Algorithms For Massive MIMO

MIMO has become the key technology in the LTE and LTE-Advanced system withits potentially huge advantage to achieve high spectral efficiency by Vertical BellLaboratory Layered Space Time technology. The maximum number of antennas that aresupported by the current standard is eight and the spectral efficiency can be achieved isonly10bps/Hz. Very large MIMO, also known as Massive MIMO，is an extension of theexisting MIMO technology. It has tens to hundreds of antennas in both base stations andterminals. It is possible to increase diversity orders and multiplexing gain, as well asimproving channel capacity and spectral efficiency dozens of times. However, in theMassive MIMO system with so many antennas, the reception complexity has becomethe bottleneck of its hardware implementation. And so this paper mainly focused on theperformance and complexity of different Massive MIMO detection methods to explorehigh-performance, low-complexity detection algorithms.Conventional signal receiving methods are difficult to simultaneously satisfy thehigh-performance and low-complexity requirement of Massive MIMO. The researchreported in this thesis is concerned with Multistage Likelihood Ascent Search (M-LAS),Tabu Search (TS) and Markov chain Monte Carlo (MCMC) technique, which are basedon local neighborhood search and Gibbs Sampling algorithms. This paper has analysedthe performance and complexity of different detection algorithms by simulation, usingdifferent antenna numbers, iterations and parameters. Finally, we use repeatedly restartsand different stopping criteria to enhance the detection performance.In the first chapter of this paper, we generally introduce the research backgroundand the status of Massive MIMO. Then we briefly introduce the research work, whichhave been done to make the direction clarified. In the second chapter, we have brieflyanalysed the MIMO system model, common MIMO detection methods and errordistributions of Maximum Likelihood detection. In chapter three, we study the theoryand complexity of M-LAS algorithm, which can be used with QR decomposition toachieve joint detection or executed with different initial values in parallel to improve thedetection performance. But it is difficult to decide the optimal restart number. In order to solve this problem, we propose a low complexity detection method based on M-LASwith restarts (M-LASR). In the fourth chapter, we introduce the detection principle andimplementation steps of TS and Layered TS (LTS) algorithms, analysis the influence ofdifferent parameters on TS detection. We present the sorted LTS (SLTS) and itssimplified algorithms by utilize the advantage of sorted QR decomposition and analysisthe performance of TS with restarts (TSR) by simulation. In the fifth chapter, we aremainly to solve the poor performance in high signal to noise ratio (SNR) of MCMCdetection algorithm based on Gibbs sampling. Using escape mechanism and restarts toimprove the detection performance, and we finally simulate the performance ofrandomized MCMC (RMCMC)，Layered RMCMC (LRMCMC) and RMCMC withrestarts (RMCMCR) algorithms in16QAM mode.The simulation results show that the three detection algorithms have an averagecomplexity of (N2)per bit in Massive MIMO system with antenna number N. Theyhave the characteristics of more antennas, better detection performances, which is calledMassive MIMO characteristic. In a64×64V-BLAST system with4QAM, they achievean un-coded BER of103at an SNR of less than1dB away from SISO-AWGNperformance.