Search

Physical layer network coding (PNC) significantly improves the throughput of the two-way relay channel (TWRC). However, most of current work on PNC just focuses on one pair users, and there are limit work on multi-pair PNC, in which the main bottleneck on system performance is the interference caused by the remaining communicating pairs. In this paper, we consider a scenario where multiple pairs communicate with each other through one relay node. For a purpose of comparison, we present the analytical capacities of the multi-pair version of the existing TWRC schemes in closed forms. Considering more practical issues such as the direction distribution of incident signals, we propose a spatial multi-pair physical layer network coding scheme. The novelty includes: in the multiple access phase, the log likelihood ratio combination scheme is proposed, which is robust in bad communication conditions; in the broadcast phase, two adaptive group broadcasting schemes are proposed, which can efficiently overcome the near interference problem. Simulation results validate the superiority of the proposed algorithms. [ABSTRACT FROM AUTHOR]

The IEEE 802.11ax standard defined a set of new specifications to improve spectrum efficiency, power efficiency, and reliability of future wireless local area networks (WLANs). Among these specifications, the uplink multi-user multiple-input, multiple-output (MU-MIMO) remedies the uplink shortcomings of existing WLANs and enables high-efficiency uplink transmissions. This paper considers the problem of adaptive modulation and coding (AMC) for uplink MU-single-input, multiple-output (MU-SIMO) in the upcoming WLANs. Our target is to select the appropriate modulation and coding scheme (MCS) for the existing users so as to maximize the throughput and at the same time guarantee that all users can satisfy a frame error constraint. We adopt a deep learning approach to tackle this problem. We propose to let the access point (AP) leverage the estimated channel state information (CSI) of all users along with the estimated noise standard deviation as the input features to a deep convolutional neural network (DCNN) that is trained to perform AMC. Simulation results reveal that the proposed DCNN for the uplink MU-SIMO AMC outperforms rival machine learning techniques such as fully-connected deep neural network (DNN), k-nearest neighbors (KNN), support vector machines (SVM), Gaussian Naive Bayes (GNB), and random forest (RF) in terms of the throughput and the frame error rate (FER). [ABSTRACT FROM AUTHOR]