Low-complexity Detection Algorithms In Massive MIMO Systems

Massive MIMO(multiple-input multiple-output) with tens or hundreds of antennas at the base station can significantly improve the system capacity and spectrum efficiency, enhance the reliability of the system as well, which is considered as one of the most key technologies to future wireless communication systems, especially for the fifth generation(5G) communications. However, the computational complexity will increase sharply with the increasement of the number of base station antennas. Therefore, a key requirement on the uplink(user terminals to base station link) in massive MIMO systems is to achieve reduced signal detection complexity at the base station receiver, while maintaining good performance.In massive MIMO systems, traditional linear signal detectors used in conventional MIMO systems can achieve nearly optimal performance when the number of base station antennas is far larger compared to the number of users. However, these linear detectors involve complicated matrix inversion. In the case when the number of users is large enough, even comparable to the number of base station antennas, traditional linear detection algorithms can't meet the demand, further study should be made on other low-complexity detection algorithms. This thesis puts emphasis on low-complexity signal detection algorithms in massive MIMO systems. Some typical algorithms have been studied, and two new detection algorithms have been proposed. The main research contents and achievements are listed as below.1. Some existing detection algorithms are introduced and studied. For the convenience of presentation, a multiuser massive MIMO system is presented at first. Then, a brief introduction about the traditional classical linear detection algorithms are given, including matched filter(MF), zero forcing(ZF) and minimum mean square error(MMSE) algorithms. In addition, three massive MIMO detection algorithms which come from neural networks and image processing are studied, and they are likelihood ascent search(LAS) algorithm, markov chain monte carlo(MCMC) algorithm and belief propagation(BP) algorithm, respectively. Simulation results about these three algorithms are also given.2. In the case when the number of base station antennas is far larger compared to the number users, this thesis proposes an improved MMSE algorithm, called relaxation iteration detection(RID) algorithm. The proposed RID algorithm exploits the channel characteristics occurring in massive MIMO channels and the relaxation iteration method to avoid the matrix inversion. A proper initial solution is given to accelerate the convergence speed, which can further reduce the computational complexity of the RID algorithm. In addition, considering the imperfect channel state information, we point out that a proper channel estimation scheme which is very appropriate for the RID algorithm. The influences of large-scale fading and channel estimation error about the RID algorithm are also analyzed.3. When the number of users is large, even comparable to the number of base station antennas, in order to further reduce the computational complexity of the message passing detection(MPD) algorithm, this thesis proposes an improved MPD(IMPD) algorithm. The improved MPD algorithm optimizes the original algorithm in initial vector selecting, iterative process, and convergence condition, respectively. The proposed RID algorithm accelerates the convergence speed of the MPD algorithm, which can reduce the computational complexity while ensuring the detection performance. Simulation results show that the proposed algorithm can always achieve the very close performance to that of the MPD with lower complexity.