Layered Gibbs Sampling Algorithm for Near-Optimal Detection in Large-MIMO Systems

Abstract

In this paper, we propose a Markov chain Monte Carlo (MCMC) based layered Gibbs sampling (GS) algorithm for low-complexity symbol vector detection in multiple-input multiple-output (MIMO) systems with large (tens to hundreds) number of antennas i.e. large-MIMO systems. The underlying idea in the proposed work is to use the MCMC based GS technique in a layered fashion where, in each layer the GS is used to make decision about a symbol probabilistically by sampling from a distribution. The proposed algorithm successfully alleviates the stalling problem encountered at high signal to noise ratios in the conventional GS based detection methods. Since, layered detection is prone to error propagation, therefore, two different sorting techniques namely, signal to noise ratio based sorting and log-likelihood ratio based sorting of the detection sequence are used for mitigating the error propagation. Simulation results validate superiority of the proposed algorithm over the existing GS based detection methods, and also, the bit error rate performance of the proposed algorithm approach near-optimal performance with low-complexity.