Cell-free (CF) multiple-input multiple-output systems employ distributed access points (APs) to enhance their performance. In the downlink, CF systems resort to precoding techniques to overcome multiuser interference. However, network-wide precoders incur high computational and signaling costs that cannot be afforded by practical systems. Therefore, we develop user-centric precoders that use clusters of APs for downlink transmission. The clusters are formed by employing the large-scale fading coefficients between users and APs. An antenna selection policy is proposed to define the clusters of APs. Different from other techniques, we also employ a user selection matrix to restrict which users are taken into account in the computation of the precoder. Two major benefits emerge from the proposed approach. First, the computational cost is reduced because several channel estimates are unnecessary. Second, the computational complexity of the precoder is reduced by restricting the number of users, which leads to matrices with reduced size. Closed-form expressions to compute the sum-rate of the proposed precoders are derived. Numerical results demonstrate that our approach can attain more than 90% of the network-wide precoders performance, while offering lower costs.

暂无