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.

Non-orthogonal multiple access (NOMA) has been recently considered in millimeter-wave (mmWave) massive MIMO systems to further enhance the spectrum efficiency. In addition, simultaneous wireless information and power transfer (SWIPT) is a promising solution to maximize the energy efficiency. In this paper, for the first time, we investigate the integration of SWIPT in mmWave massive MIMO-NOMA systems. As mmWave massive MIMO will likely use hybrid precoding (HP) to significantly reduce the number of required radio-frequency (RF) chains without an obvious performance loss, where the fully digital precoder is decomposed into a high-dimensional analog precoder and a low-dimensional digital precoder, we propose to apply SWIPT in HP-based MIMO-NOMA systems, where each user can extract both information and energy from the received RF signals by using a power splitting receiver. Specifically, the cluster-head selection algorithm is proposed to select one user for each beam at first, and then the analog precoding is designed according to the selected cluster heads for all beams. After that, user grouping is performed based on the correlation of users' equivalent channels. Then, the digital precoding is designed by selecting users with the strongest equivalent channel gain in each beam. Finally, the achievable sum rate is maximized by jointly optimizing power allocation for mmWave massive MIMO-NOMA and power splitting factors for SWIPT, and an iterative optimization algorithm is developed to solve the non-convex problem. Simulation results show that the proposed HP-based MIMO-NOMA with SWIPT can achieve higher spectrum and energy efficiency compared with HP-based MIMO-OMA with SWIPT.

Heterogeneous networks (HetNets), which involve densely deployed femtocells underlaid traditional macrocell network, is a promising solution to the extremely high data rate requirements of the future 5G communications. In this paper, we analyze the closed-form optimal deployment of femtocells in HetNets to maximize the network throughput under the outage constraints from both macrocells and femtocells. Specifically, we model the random distribution of macro cell users (MUEs) and femtocell base stations (FBSs) as Poisson Point Processes (PPPs). Then, the closed form expressions for outage probabilities in both uplink and downlink transmissions are derived. Further, we study the network throughput maximization problem under the outage probability constraints. Finally, With the help of convex optimization, the interval of FBS density, which contains the maximum network throughput is obtained in closed form. Simulation results validate the impact of the system parameters on the different optimal FBS density as well as the influence of interference to the maximum network throughput.

The recent concept of beamspace multiple-input multiple-output (MIMO) is capable of significantly reducing the number of radio-frequency (RF) chains required by millimeter-wave (mmWave) massive MIMO systems. However, the fundamental limit of the existing beamspace MIMO is that, the number of supported users cannot be higher than the number of RF chains using the same time-frequency resources. To break this limit, beamspace MIMO is integrated with non-orthogonal multiple access (NOMA) in the proposed MIMO-NOMA system in this paper, where the number of supported users can be higher than the number of RF chains. To reduce the inter-beam interference, a transmit precoding (TPC) scheme based on the principle of zero-forcing (ZF) is designed. Furthermore, a dynamic power allocation scheme is proposed for maximizing the achievable sum rate. Moreover, a low-complexity iterative optimization algorithm is conceived for dynamic power allocation. Simulation results show that the proposed beamspace MIMO-NOMA achieves a higher spectrum and energy efficiency than the existing beamspace MIMO for mmWave communications.

The recent concept of beamspace multiple input multiple output (MIMO) can significantly reduce the number of required radio frequency (RF) chains in millimeter-wave (mmWave) massive MIMO systems without obvious performance loss. However, the fundamental limit of existing beamspace MIMO is that the number of supported users cannot be larger than the number of RF chains at the same time-frequency resources. To break this fundamental limit, in this paper, we propose a new spectrum and energy-efficient mmWave transmission scheme that integrates the concept of non-orthogonal multiple access (NOMA) with beamspace MIMO, i.e., beamspace MIMO-NOMA. By using NOMA in beamspace MIMO systems, the number of supported users can be larger than the number of RF chains at the same time-frequency resources. In particular, the achievable sum rate of the proposed beamspace MIMO-NOMA in a typical mmWave channel model is analyzed, which shows an obvious performance gain compared with the existing beamspace MIMO. Then, a precoding scheme based on the principle of zero forcing is designed to reduce the inter-beam interferences in the beamspace MIMO-NOMA system. Furthermore, to maximize the achievable sum rate, a dynamic power allocation is proposed by solving the joint power optimization problem, which not only includes the intra-beam power optimization, but also considers the inter-beam power optimization. Finally, an iterative optimization algorithm with low complexity is developed to realize the dynamic power allocation. Simulation results show that the proposed beamspace MIMO-NOMA can achieve higher spectrum and energy efficiency compared with the existing beamspace MIMO.

Time domain synchronous orthogonal frequency division multiplexing (TDS-OFDM) has higher spectral efficiency than standard cyclic prefix OFDM (CP- OFDM), which is achieved by using a known pseudorandom noise (PN) sequence to replace the classical CP. However, due to the interference between the PN sequence and the data block, the performance of TDS-OFDM degrades severely over fast fading channels. To solve this problem, based on the distributed compressive sensing (DCS) theory, we propose an efficient way to realize simultaneous multi-channel reconstruction, which is achieved by using the inter-block-interference (IBI)-free region to reconstruct the high-dimensional sparse multipath channel. Specifically, we propose to utilize the temporal correlation of wireless channels as well as the channel property that path gains change much faster than path delays to simultaneously reconstruct multiple sparse channels. Then, we propose the parameterized channel estimation method based on simultaneous compressive sampling matching pursuit (S-CoSaMP) algorithm to achieve better channel estimation performance in fast time-varying channels. Simulation results demonstrate that the proposed scheme can achieve improved performance than conventional solutions.

To fully exploit advantages of massive MIMO, channel state information at the transmitter (CSIT) is essential to obtain the system performance gains. By far, both channel estimation and channel feedback have been proposed for FDD massive MIMO by exploiting the sparsity of CSI, but they are usually separately discussed, which may impair the CSIT acquisition performance and lead to unnecessary complex computation for users. In this paper, we propose the structured-CS based differential CSIT acquisition scheme for massive MIMO systems, where the downlink channel training and uplink channel feedback are jointly considered. Specifically, we first exploit the temporal correlation of time- varying channels to propose the differential CSIT acquisition scheme, which can reduce both the overhead for downlink training and uplink feedback. Then, we propose the structured compressive sampling matching pursuit (S-CoSaMP) algorithm to further reduce overhead by leveraging the structured sparsity of wireless MIMO channels. Moreover, the proposed differential operation and S-CoSaMP can also be used at users for better channel estimation performance if channel state information at the receiver is needed. Simulation results have demonstrated that the proposed scheme can achieve better CSIT acquisition performance than its counterparts.

Non-orthogonal multiple access (NOMA) has become one of the promising key technologies for future 5G wireless communications to improve spectral efficiency and support massive connectivity. However, in the uplink grant-free NOMA system, the current near-optimal multi-user detection (MUD) based on message passing algorithm (MPA) assumes that the user activity information is exactly known at the receiver, which is impractical yet challenging due to anyone of massive users can randomly enter or leave the system. In this paper, inspired by the observation of user sparsity, we jointly use compressive sensing (CS) and MPA to propose a CS-MPA detector to realize both user activity and data detection for uplink grant-free NOMA. Specifically, the MUD problem is firstly formulated under CS framework by exploiting user sparsity, and then user activity can be detected by sparse signal recovery algorithms in CS. Then, MPA can be performed to reliably detect active users' data. It is shown that the proposed CS-MPA detector with affordable complexity not only outperforms the conventional MPA detector without user activity information, but also achieves very close performance to the genie- knowledge MPA detector with exact knowledge of user activity, especially when the signal-to-noise ratio (SNR) is high.

Massive multiple-input multiple-output (MIMO) using a large number of antennas at the base station (BS) is a promising technique for the next-generation 5G wireless communications. It has been shown that linear precoding schemes can achieve near-optimal performance in massive MIMO systems. However, classical linear precoding schemes such as zero- forcing (ZF) precoding suffer from high complexity due to the fact they require the matrix inversion of a large size. In this paper, we propose a low-complexity precoding scheme based on the least square QR (LSQR) method to realize the near-optimal performance of ZF precoding without matrix inversion. We show that the proposed LSQR-based precoding can reduce the complexity of ZF precoding by about one order of magnitude. Simulation results verify that the proposed LSQR-based precoding can provide a better tradeoff between complexity and performance than the recently proposed Neumann-based precoding.