Low Complexity MIMO Detection based on Belief Propagation over Pair-wise Graphs

Abstract: This paper considers belief propagation algorithm over pair-wise graphical models to develop low complexity, iterative multiple-input multiple-output (MIMO) detectors. The pair-wise graphical model is a bipartite graph where a pair of variable nodes are related by an observation node represented by the bivariate Gaussian function obtained by marginalizing the posterior joint probability density under the Gaussian input assumption. Specifically, we consider two types of pair-wise models, the fully connected and ring-type. The pair-wise graphs are sparse, compared to the conventional graphical model in [18], insofar as the number of edges connected to an observation node (edge degree) is only two. Consequently the computations are much easier than those of maximum likelihood (ML) detection, which are similar to the belief propagation (BP) that is run over the fully connected bipartite graph. The link level performance for non-Gaussian input is evaluated via simulations, and the results show the validity of the proposed algorithms. We also customize the algorithm with Gaussian input assumption to obtain the Gaussian BP run over the two pair-wise graphical models and, for the ring-type, we prove its convergence in mean to the linear minimum mean square error (MMSE) estimates. Since the maximum a posterior (MAP) estimator for Gaussian input is equivalent to the linear MMSE estimator, it shows the optimality, in mean, of the scheme for Gaussian input.