The Theory Analysis And Application Of Detection And Pre-Coding Algorithm In Large-Scale DAS System

With the rapid development of mobile communication technology and widely deployment of 4G system, the social economy has been promoting economic development greatly, but the wireless transmission rate and communication service experience is still facing higher and higher demands and challenges. As the two main driving force(Mobile Internet and IOT(Internet Of Thing)) for future wireless communication system is challenging the system to cope with the explosive data growth, so the fifth generation mobile communication technology emerged. Large-scale MIMO, as one of the core technologies in 5G MIMO, has been widely concerned and researched for significantly improve transmission rate and spectrum efficiency. However, the traditional detection and pre-coding algorithm(like ZF and MMSE algorithms) for MIMO system will turn to be infeasible computational complexity as the number of antennas turning very large. So the low complexity detection and pre-coding algorithm turns to be one of the key research topics in large-scale MIMO system, which is also the key research point in the paper. In this paper, we study the low complexity detection and pre-coding algorithm for large-scale DAS system base on the sparsity of the systems, and proposed a practicable solution.The paper firstly introduces the research background, the development history of MIMO technology and topic related current research status.Wireless communication environment is the basis of studying communication problems. With full con-sideration of large-scale DAS's channel characteristics, the system model of this paper is introduced. Under this system model, the way to sparsity of the channel model is studied. What's more, we store the set of the index of none-zero elements in channel matrix for the calculate of low complexity detection and pre-coding algorithm. Above all, we also calculate the capacity of the channel after sparsity.The detection at the uplink is to demodulate the constellation point which minimize the detection error from the constellation set, so we can use BP(belief propagation) algorithm to demodulate the constellation point. In this paper, we compared the complexity between the BP based on markov random fields(MRF) and BP based on factor graph(FG) and realized that the FG-BP algorithm perform better. Then we implement the FG-BP in large-scale DAS system, and integrate with the large sparsity of channel matrix, we propose the low complexity BP detection algorithm for uplink of large-scale DAS system. Further more, the overall computational cost at BPU end is decomposed into many smaller computation tasks at RAUs that will also improve the system's overall performance greatly.The per-coding at the downlink is to calculate the result from continuous domain, so we cannot use MRF-BP or FG-BP algorithm like in the uplink. We take advantage of the characteristics of Gaussian variable, then only update the mean and variance of unknown variable among the iteration to get the final result of the unknown variable. In this paper, firstly we introduce the theory of GaBP(Gaussian BP) and how to use GaBP to calculate the RZF(regularize zero forcing). Secondly, we implement the GaBP-RZF algorithm for large-scale DAS system, and integrate with the large sparsity of channel matrix to reduce the computational complexity greatly. What's more, we also decompose the overall computational cost at BPU end to many smaller computation tasks at RAUs to improve the system's overall performance greatly.Finally, we simulate the performance of my low complexity detection and pre-coding algorithm, and the results shows that we can reduce the complexity of computation and improve the system's overall performance greatly within acceptable capacity loss when we set different sparsity threshold value.