Beam Allocation based on Spatial Compatibility for Hybrid Beamforming C-RAN Networks L´ aszlon R. Costa *† , Yuri C. B. Silva † , F. Rafael M. Lima † , and Anja Klein * * Communications Engineering Lab, Technische Universit¨ at Darmstadt, Merckstrasse 25, 64283 Darmstadt, Germany † Wireless Telecom Research Group (GTEL), Federal University of Cear´ a, Fortaleza, Brazil {l.rodrigues, a.klein}@nt.tu-darmstadt.de, {laszlon, yuri, rafaelm}@gtel.ufc.br Abstract—In this work, we consider a beam allocation problem in hybrid beamforming (HBF) cloud radio access networks (C-RANs) to maximize the sum rate. The problem considers a codebook-based analog beamforming performed at the remote radio head (RRH) and digital beamforming at the baseband unit (BBU). Differently from previous works, we assume that a given user equipment (UE) can be served by multiple beams, without being specifically associated to a given RRH. Due to the relation between the analog beam allocation and the digital beamforming, we consider a new metric based on channel correlation and channel attenuation for the analog beamforming solution. The metric measures the spatial compatibility in order to improve the spectral efficiency before the digital beamforming solution. In order to evaluate the proposed metric we present a low-complexity greedy algorithm, which is shown to provide a reasonable performance/complexity trade-off. Index Terms—hybrid beamforming, Cloud-RAN, Beam alloca- tion I. I NTRODUCTION In order to attend to the increasing demands for high data rates, low latency and massive connectivity in upcoming 5G wireless networks [1], some key technologies will be required. Among these we can mention millimeter wave transmission [2], massive multiple-input multiple-output (MIMO) [3] and cloud radio access network (C-RAN) [4]. The use of mmWave massive MIMO leads to increased bandwidth and high data rates, by efficiently using the spatial and frequency domains. In order to reduce the complexity and energy consumption of massive MIMO, hybrid beamforming (HBF) techniques [5]–[7] split the precoding design into two parts, with the analog processing being carried out at the radio frequency (RF) stage and the digital part at the baseband. The deployment of C-RAN, on the other hand, allows the signal processing functions to be split among the baseband unit (BBU) and the remote radio heads (RRHs). By using C-RAN solutions, it is possible to improve the effectiveness of interference management and cooperative techniques. The implementation of mmWave massive MIMO in a C-RAN context may lead to benefits in terms of cost and capacity, as shown in [8]. In such scenario, the analog beamforming can be implemented at the RRHs and the digital part at the BBU. Considering a fast enough backhaul among the BBU and RRHs, it becomes possible to implement joint transmission techniques [9] for C-RAN scenarios, as consid- ered in [10], [11]. In this paper, we consider the problem of beam allocation in HBF C-RAN networks. The problem is formulated for the joint precoding scenario, assuming a codebook-based analog beamforming at the RRHs and digital beamforming at the BBU. Differently from previous works, such as [10]–[12], we assume that a given user equipment (UE) can be served by multiple beams, without being specifically associated to a given RRH. The formulated problem is non-convex and composed of integer and continuous variables, which makes the optimal solution impractical for realistic scenarios. Therefore, we propose to separate the analog beamforming allocation from the digital precoder solutions as two steps performed sequentially. The first step is the analog beam allocation and the second step is the digital precoder calculated considering the analog beamforming solution. In order to improve the system capacity, we propose an analog beam allocation method based on a spatial compatibility metric, which considers the channel correlation and attenuation. The solution considering this metric will create an equivalent channel with low spatial correlation, therefore, the interfering signals can be easily isolated by the digital beamforming. It is worth to mention that the components of the proposed metric, i.e., channel correlation and attenuation, can be acquired from channel statistics, thus avoiding the need for instantaneous channel state information (CSI). In order to evaluate our beam allocation metric, we propose a low-complexity greedy algorithm. The paper is organized as follows. In section II, the con- sidered system model is presented. The considered beam allocation problem and the proposed suboptimal solutions are detailed in section III. Next, simulation results are presented and discussed in section IV, while conclusions and next steps are provided in section V. II. SYSTEM MODEL We consider M base stations (BSs), each one corresponding to an RRH equipped with N antennas and B RF chains, simultaneously serving K single-antenna UEs. Each BS has a fully connected hybrid beamforming architecture, which means that all antennas from one BS are connected to all B RF chains. Therefore, each BS is capable to design B analog beams using N antennas, where B<N . For simplicity, let us consider MB ≥ K, which means that it is possible to point at least one analog beam to each UE considering the overall Workshop on Smart Antennas WSA, April 2019, Vienna, Austria