Hybrid precoding design is a high-complexity problem due to the coupling of analog and digital precoders as well as the constant modulus constraint for the analog precoder. Fortunately, the deep learning based hybrid precoding methods can significantly reduce the complexity, but the performance remains limited. In this paper, inspired by the attention mechanism recently developed for machine learning, we propose an attention-based hybrid precoding scheme for millimeter-wave (mmWave) MIMO systems with improved performance and low complexity. The key idea is to design each user’s beam pattern according to its attention weights to other users’. Specifically, the proposed attention-based hybrid precoding scheme consists of two parts, i.e., the attention layer and the convolutional neural network (CNN) layer. The attention layer is used to identify the features of inter-user interferences. Then, these features are processed by the CNN layer for the analog precoder design to maximize the achievable sum-rate. Simulation results demonstrate that the attention layer could mitigate the inter-user interferences, and the proposed attention-based hybrid precoding with low complexity can achieve higher achievable sum-rate than the existing deep learning based method.

In this paper, a new multi-cell MMSE detector is proposed for massive MIMO systems. Let K and B denote the number of users in each cell and the number of available pilot sequences in the network, respectively, with B = βK, where β ≥ 1 is called the pilot reuse factor. The novelty of the multi-cell MMSE detector is that it utilizes all B channel directions that can be estimated locally at a base station, so that intra-cell interference, parts of the inter-cell interference and the noise can all be actively suppressed, while conventional detectors only use the K intra-cell channels. Furthermore, in the large- system limit, a deterministic equivalent expression of the uplink SINR for the proposed multi-cell MMSE is derived. The expression is easy to compute and accounts for power control for the pilot and payload, imperfect channel estimation and arbitrary pilot allocation. Numerical results show that significant sum spectral efficiency gains can be obtained by the multi-cell MMSE over the conventional single-cell MMSE and the recent multi-cell ZF, and the gains become more significant as β and/or K increases. Furthermore, the deterministic equivalent is shown to be very accurate even for relatively small system dimensions.

In this paper, a new multi-cell MMSE precoder is proposed for massive MIMO systems. We consider a multi-cell network where each cell has K users and B orthogonal pilot sequences are available, with B = βK and β ≥ 1 being the pilot reuse factor over the network. In comparison with conventional single-cell precoding which only uses the K intra-cell channel estimates, the proposed multi-cell MMSE precoder utilizes all B channel directions that can be estimated locally at a base station, so that the transmission is designed spatially to suppress both parts of the inter-cell and intra-cell interference. To evaluate the performance, a large-scale approximation of the downlink SINR for the proposed multi-cell MMSE precoder is derived and the approximation is tight in the large-system limit. Power control for the pilot and payload, imperfect channel estimation and arbitrary pilot allocation are accounted for in our precoder. Numerical results show that the proposed multi-cell MMSE precoder achieves a significant sum spectral efficiency gain over the classical single-cell MMSE precoder and the gain increases as K or β grows. Compared with the recent M-ZF precoder, whose performance degrades drastically for a large K, our M-MMSE can always guarantee a high and stable performance. Moreover, the large-scale approximation is easy to compute and shown to be accurate even for small system dimensions.

Channel measurements show that significant spatially non-wide-sense-stationary characteristics rise in massive MIMO channels. Notable parameter variations are experienced along the base station array, such as the average received energy at each antenna, and the directions of arrival of signals impinging on different parts of the array. In this paper, a new channel model is proposed to describe this spatial non-stationarity in massive MIMO channels by incorporating the concepts of partially visible clusters and wholly visible clusters. Furthermore, a closed-form expression of an upper bound on the ergodic sum capacity is derived for the new model, and the influence of the spatial non-stationarity on the sum capacity is analyzed. Analysis shows that for non-identically-and-independent-distributed (i.i.d.) Rayleigh fading channels, the non-stationarity benefits the sum capacity by bringing a more even spread of channel eigenvalues. Specifically, more partially visible clusters, smaller cluster visibility regions, and a larger antenna array can all help to yield a well-conditioned channel, and benefit the sum capacity. This shows the advantage of using a large antenna array in a non-i.i.d. channel: the sum capacity benefits not only from a higher array gain, but also from a more spatially non-stationary channel. Numerical results demonstrate our analysis and the tightness of the upper bound.

Spatial Modulation is a newly developed concept for multi-antenna systems, which could greatly reduce the signal processing complexity and hardware implementation burden faced by these systems today. To improve the spectral efficiency of spatial modulation system, a novel precoding scheme for downlink multi-user system is proposed in this paper. The precoding scheme not only cancels the inter-user interference, but also preserves the information embedded both in constellation symbols as well as in transmit antenna indexes, decomposing the system into independent single-user spatial modulation systems. Simulation results show that with a high transmission rate, the proposed scheme offers a smaller bit error rate than multi-user Zero Forcing precoding in practical SNR region. With the same base station antenna number and single-antenna user number, it also outperforms Zero Forcing precoding in terms of achievable sum rate. Furthermore, systems using our scheme could achieve a multi-user gain over single-user spatial modulation system.