MIMO-OFDM Receiver Design Based On Factor Graph Transformation And Message Passing

As the core technology of the fourth generation and the fifth generation of mobile communication,OFDM(Orthogonal Frequency Division Multiplexing)can exploit narrowband orthogonal carrier to deal with the problem of frequency selective fading,while MIMO(MultipleInput Multiple-Output)can increase channel capacity and improve diversity gain by taking use of space resources of antennas in different positions.The combination of the above two technologies would have complementary advantages and improve the performance of the system.In recent years,massive MIMO system becomes the research focus in the field of communication due to its high system capacity,spectrum and energy efficiency.However,the issues of multiuser detection,channel estimation and complexity caused by the larger antenna size require more advanced receiving algorithm design.This thesis studies the MIMO and OFDM system for the next generation mobile communication,proposed a joint channel estimation and detection algorithm based on factor graph and message passing,solved the problem of incompatibility between performance and complexity,of receivers.The main contents of the thesis are as follows:1.Considering OFDM system,we design an iterative receiving algorithm for sparse channel environment.Firstly,we propose a combined mean field and expectation propagation receiving algorithm via the transformation of the existing factor graph.Using the sparsity of physical channels,we also apply sparse Bayesian learning model in channel estimation,resulting in the reduction of pilot overhead and increasing of frequency efficiency.To decrease the complexity,we approximate the proposed iterative receivers by the law of large numbers and find the inner link with the generalized approximate message passing,and propose a low-complexity combined receiver.Based on that,we also study several message scheduling methods,provides reference for the design of iterative algorithm.Comparing with existing algorithms,the proposed receiver can perform better at the cost of same complexity.2.Considering MIMO-OFDM,we design an iterative receiver based on factor graph transformation and cooperative message passing.Firstly,we divide the factor graph for MIMO multi-user detection and use different message passing rules in different subgraphs and proposed a combined iterative receiver algorithm.Secondly,to reduce the complexity of the proposed algorithm,we propose a low-complexity receiver based on cooperative rule to realize the collaborative update of belief propagation and mean field rule,avoiding their shortcomings.We then apply cooperative algorithm in block-fading channel case and find the connection between bilinear generalized approximate message passing,which shows the rationality of cooperative rules.Finally,based on the channel state information from different users,we proposed a more practical partial Gaussian approximate algorithm by joint the cooperative and combining algorithms.The simulation results show that comparing with the existing algorithms,the proposed cooperative algorithm can achieve better performance gain in same complexity cost;the proposed combining receiver can improve the performance gain significantly at the cost of slight increasing of complexity,and proposed partical Gaussian approximate algorithm can reach the good tradeoff between complexity and performance.3.Considering massive MIMO-OFDM system with spatial correlation,we design a receiver by exploiting the clustering characteristics of antenna array.Firstly,the Dirichlet processing from the field of machine learning is introduced as the priori of sparse Bayesian learning structure to set up DP-SBL model.We pass and schedule message based on the factor graph transformation and combined message passing algorithm,resulting in the clustering estimation algorithm for the problem of multiple measurement vectors.Then we design a probabilistic massive MIMO channel model and construct channels with different spatial correlation degree,where channel propagation environments are simulated by adjusting probability parameter and mutation parameters.Simulation results demonstrate that the proposed receiver can fully exploit space resources of massive MIMO systems and deal with the clustering problem of antennas which have the behavior of sparse common support.Comparing with the existing algorithms,the proposed receiver shows obvious performance gain and strong robustness.