Precoding Methods Based On Compressive Sensing In Massive MIMO

As the demand for high rate data and larger system capacity becoming more and more urgency, Massive MIMO becomes a study interest. Besides the plenty of advantages Massive MIMO takes, it also has a lot to be solved. When the number of antennas at the BS come to several hundreds, challenge turns to the complexity of compution. One BS serving tens of users at the same time and frequency resource also results in heavy interference. To take more advantages of Massive MIMO technology, BS should perform precoding to suppress interference.However, precoding needs the knowleage of the channel state information. When the number of BS antennas becomes hundreds, it’s very difficult to obtain the channel state information, thus, precoding is chanlleging. Many literature assume perfect channel state information at the BS when discussing precoding. However, this assumption is not practical. On the other hand, plenty of research point out that the Massive MIMO channel is sparse, and the sparsity can be used to reduce the overhead of channel state information estimation. So, this paper focus on the sparsity of Massive MIMO channel and design the method to estimate the essential channel state information for precoding based on compressive sensing.This paper first describes the sparsity property of Massive MIMO system channel in detail. This paper mainly describes the sparsity of time domain channel impulse response(CIR) and the frequency domain channel vectors in the virtual angel domain, and analyses the corresponding cause. In addition, this paper introduces the common support characteristics of Massive MIMO system channel under specific scenarios. More specifically, time-domain CIR of a plurality of different antennas pairs and the same antenna pair on different sub-carriers have the same sparse mode, that means they have the same support set. The introduction of these characteristics of Massive MIMO channel is for the later use.Followed, this paper introduces the compressive sensing technology, and describes the advantages of this technology and some common signal recovery methods, and points out that the greedy algorithms based on the 0 norm optimization have low computation complexity and good performance, thus they are more applicable in practice. Besides, the performance of several greedy algorithms is simulated. The results show that OMP and CoSaMP are more suitable for the scenario of this paper.Next, this paper models the channel information estimation process of TDD and FDD systems into standard compressive sensing model, and describes the methods to recover the channel information based on SMV and MMV model. Besides, simulation of the recovery performance is also given.Finally, a brief introduction about several common precoding algorithms is given in this paper, which shows the channel information is the key information required for pre-coding. As the antenna number of Massive MIMO system is huge, appropriate methods with lower overhead of obtaining channel information is particularly necessary. Perfect channel information is not available in practice, thus performance of precoding assuming ideal channel information can only be treated as performance bound. The precoding performance of the actual system can only approach the performance bound but difficult to achieve. To evaluate the performance of precoding with channel information estimated using compressive sensing methods, this paper also simulated the corresponding performance. Results show that the performance is pretty close to the performance bound, which highlights the advantage of precoding with compressive sensing methods to get channel information.