EM-AMP-Based Joint Active User Detection and Channel Estimation in Cell-Free System

This article considers the massive access problem in an uplink cell-free multiple-input–multiple-output (MIMO) system where massive potential users communicate with a set of access points (APs), while all APs are connected to a central processing unit (CPU). By exploiting the sporadic traffic of users, we use expectation maximization-approximate message passing (EM-AMP) algorithm to implement channel estimation based on the aggregate received signal of all APs at the CPU. Then, the active user detection (AUD) can be accomplished with the corresponding posterior support probabilities. It should be noted that the a priori distribution of channel coefficients plays an important role in EM-AMP, which will make a significant difference to the system performance. Moreover, provided with different prior information about the users' location, the CPU will obtain different a priori distribution of the channel coefficients. Inspired by this, instead of using a priori distribution with hypothesised parameters as the original EM-AMP does, we assume that the CPU has some prior information about the users' location. Then, we propose a scheme that can efficiently obtain the a priori distribution of channel coefficients by fully exploiting this information. After that, we propose a modified algorithm based on the derived a priori distribution. Simulation results reveal that the proposed algorithm has a noticeable performance gain than the original EM-AMP when the pilot length is relatively small. More importantly, with our proposed scheme, given rough prior information, we can achieve almost the same system performance as when the users' location is precisely known.