Beam codebooks are a new feature of massive multiple-input multiple-output (M-MIMO) in 5G new radio (NR). Codebooks comprised of beamforming vectors are used to transmit reference signals and obtain limited channel state information (CSI) from receivers via the codeword index. This enables large arrays that cannot otherwise obtain sufficient CSI. The performance, however, is limited by the codebook design. In this paper, we show that machine learning can be used to train site-specific codebooks for initial access. We design a neural network based on an autoencoder architecture that uses a beamspace observation in combination with RF environment characteristics to improve the synchronization signal (SS) burst codebook. We test our algorithm using a flexible dataset of channels generated from QuaDRiGa. The results show that our model outperforms the industry standard (DFT beams) and approaches the optimal performance (perfect CSI and singular value decomposition (SVD)-based beamforming), using only a few bits of feedback.

Cellular networks continue to trend rapidly towards more bands and carrier frequencies, along with higher base station density, requiring complex decisions to be made when associating a mobile user with a band and cell. This paper develops a novel approach to optimizing frequency band and cell selection while taking into account user mobility and handovers. This is a complex problem because of the uncertain link failure events, handover related overheads, and the significant difference in the propagation characteristics between different frequency bands. The network dynamics due to user mobility are modeled as a Markov decision process, and we develop a recurrent Q-learning framework to exploit the relationship between user trajectories and the history of SINR measurements. The effective cell boundaries are therefore based on user trajectories and velocities rather than just position and signal strength. Detailed system-level simulations show that the proposed learning-based approach improves the throughput of the edge users by 54% and the median throughput by 34% compared to traditional SINR-based association and achieves a superior rate/coverage tradeoff (quantified as sum-log-rate) compared to SINR or signal-strength-based associations.

Wireless cellular networks have many parameters that are normally tuned upon deployment and re-tuned as the network changes. Many operational parameters affect reference signal received power (RSRP), reference signal received quality (RSRQ), signal-to-interference-plus-noise-ratio (SINR), and, ultimately, throughput. In this paper, we develop and compare two approaches for maximizing coverage and minimizing interference by jointly optimizing the transmit power and downtilt (elevation tilt) settings across sectors. To evaluate different parameter configurations offline, we construct a realistic simulation model that captures geographic correlations. Using this model, we evaluate two optimization methods: deep deterministic policy gradient (DDPG), a reinforcement learning (RL) algorithm, and multi-objective Bayesian optimization (BO). Our simulations show that both approaches significantly outperform random search and converge to comparable Pareto frontiers, but that BO converges with two orders of magnitude fewer evaluations than DDPG. Our results suggest that data-driven techniques can effectively self-optimize coverage and capacity in cellular networks.

In this article we introduce a novel solution called SuperCell, which can improve the return on investment (ROI) for rural area network coverage. SuperCell offers two key technical features: it uses tall towers with high-gain antennas for wide coverage and high-order sectorization for high capacity. We show that a solution encompassing a high-elevation platform in excess of 200 meters increases coverage by 5x. Combined with dense frequency reuse by using as many as 36 azimuthal sectors from a single location, our solution can adequately serve the rural coverage and capacity demands. We validate this through propagation analysis, modeling, and experiments. The article gives a design perspective using different classes of antennas: Luneburg lens, active/passive phased array, and spatial multiplexing solutions. For each class, the corresponding analytical model of the resulting signal-to-interference plus noise ratio (SINR) based range and capacity prediction is presented. The spatial multiplexing solution is also validated through field measurements and additional 3D ray-tracing simulation. Finally, in this article we also shed light on two recent SuperCell field trials performed using a Luneburg lens antenna system. The trials took place in rural New Mexico and Mississippi. In the trials, we quantified the coverage and capacity of SuperCell in barren land and in a densely forested location, respectively. In the article, we demonstrate the results obtained in the trials and share the lessons learned regarding green-field and brown-field deployments.

Massive MIMO, a building block of future 5G systems, is attractive for wireless information and energy transfer. This is largely due to its ability to focus energy towards desired spatial locations. In this paper, the overall energy efficiency of a wirelessly powered massive MIMO system is investigated where a multi-antenna base-station uses wireless energy transfer to charge single- antenna energy harvesting users on the downlink. The users exploit the harvested energy to transmit information to the base-station on the uplink. Using a scalable model for the circuit power consumption at the base-station, the energy efficiency performance (measured in bits/joule) of the overall system is characterized. A closed-form expression is derived for the energy efficiency- optimal downlink transmit power in terms of the key system parameters such as the number of base- station antennas and the number of users. Simulation results suggest that it is energy efficient to operate the system in the massive MIMO regime. As the number of antennas becomes large, increasing the transmit power as well as serving more users help improve the energy efficiency for moderate to large number of antennas.

We consider a downlink massive multiple-input multiple output system employing regularized zero-forcing precoding. We derive the asymptotic signal-to-leakage-plus-noise ratio (SLNR) as both the number of antennas and the number of users go to infinity at a fixed ratio. Focusing on spatially uncorrelated channels with homogeneous large scale fading gains, we show that the SLNR is asymptotically equal to signal-tointerference-plus-noise ratio, which allows us to optimize the user loading for spectral efficiency. The results show that the optimal user loading varies depending on the channel signal-tonoise ratio (SNR). As the SNR increases, the optimal user loading decreases at low SNR, but increases at high SNR.

Increasing access to telecommunication services in rural parts of the world has the potential to alleviate the digital divide felt by the people of these regions. Recently, both public and private sector entities have shown more interest in tackling this problem on a global scale. In the United States, for example, the Federal Communications Commission (FCC) subsidizes large telecommunication companies to offer affordable services in unserved areas through the Connect America Fund. Facebook's Internet. org [1] and Google's Project Loon [2] are some examples of large-scale rural connectivity efforts in the private sector. Deployment of connectivity solutions in rural communities, however, faces many practical challenges. Lack of availability and access to reliable electricity sources is one of the major hindrances for rural connectivity, particularly in underdeveloped countries. As such, wireless connectivity solutions for such applications must focus on low-power hardware operations and high energy efficiency (EE).

We propose a new hybrid precoding technique for massive multi-input multi-output (MIMO) systems using spatial channel covariance matrices in the analog precoder design. Applying a regularized zero-forcing precoder for the baseband precoding matrix, we find an unconstrained analog precoder that maximizes signal-to-leakage-plus-noise ratio (SLNR) while ignoring analog phase shifter constraints. Subsequently, we develop a technique to design a constrained analog precoder that mimics the obtained unconstrained analog precoder under phase shifter constraints. The main idea is to adopt an additional baseband precoding matrix, which we call a compensation matrix. We analyze the SLNR loss due to the proposed hybrid precoding compared to fully digital precoding, and determine which factors have a significant impact on this loss. In the simulations, we show that if the channel is spatially correlated and the number of users is smaller than the number of RF chains, the SLNR loss becomes negligible compared to fully digital precoding. The main benefit of our method stems from the use of spatial channel matrices in such a way that not only is each user's desired signal considered, but also the inter-user interference is incorporated in the analog precoder design.

We propose a mixed analog-to-digital converter ADC (mixed-ADC) structure for a cloud-RAN system, where a single-antenna user terminal communicates with a multi-antenna remote radio head (RRH). In the proposed structure, the RRH is equipped with a mixed-ADC pool that includes multiple ADC units with various resolutions. In this pool, the RRH selects the appropriate ADCs and connects the selected ADCs to each antenna to quantize the received signals; thereby each antenna can have a different resolution ADC. The fronthaul capacity is limited, so that the sum of the bits produced in the selected ADCs is also limited. To maximize the spectral efficiency or the energy efficiency of such a system, we propose algorithms for ADC resolution selection based on an approximation of the generalized mutual information in the low signal-to-noise regime. In the proposed algorithms, we show that for spectral efficiency, using high-resolution ADC on the strong channels is beneficial. The results for energy efficiency maximization are similar, though the largest resolutions are reduced to save power. The simulations show that the proposed method provides significant performance improvement.

Hybrid analog/digital precoding in the downlink of multi-user massive MIMO systems can reduce the number of RF chains hence reducing total cost and improving power efficiency. Having few RF chains, however, makes it difficult for a base station to acquire instantaneous channel state information across all antennas. We develop a hybrid technique that uses only long-term (slowly changing) channel statistics in computing the analog precoding matrix. The proposed analog precoder is designed to maximize signal-to-leakage-plus-noise ratio (SLNR) when combined with a baseband precoder. We also propose a constrained precoder design that reduces the effect of a hardware constraint where the analog precoders are realized with phase shifters.

We propose a new mixed-analog-to-digital convertor (mixed-ADC) architecture for cloud-RAN (C-RAN) systems. The RRH is equipped with a mixed-ADC pool that includes multiple ADC units with various resolutions. In this pool, the RRH selects the appropriate ADCs and connects the selected ADCs to each antenna to quantize the received signals, thereby each antenna can have a different resolution ADC. The quantized signals are sent to a centralized baseband unit (BBU) via a capacity limited fronthaul pipe. To maximize the spectral efficiency of the considered system in a single-user uplink phase, we formulate an optimization problem for ADC selection by exploiting an approximation of the generalized mutual information (GMI). Subsequently we propose a solution. The simulations show the improvement in the GMI by using the proposed ADC selection. Our major findings are: i) In a C-RAN with limited fronthaul capacity, the proposed mixed-ADC makes more efficient use of the fronthaul capacity. ii) In selecting ADCs, assigning a high resolution ADC to a strong channel is beneficial.