Clustering with Multiple Receiving Antennas in Downlink FDD CoMP Systems P. Baracca † , F. Boccardi † , and N. Benvenuto ∗ † Bell Laboratories, Alcatel-Lucent. Email: {paolo.baracca, federico.boccardi}@alcatel-lucent.com ∗ Department of Information Engineering, University of Padova, Italy. Email: nb@dei.unipd.it Abstract—Inter-cell interference in downlink cellular networks can be managed by coordination among the base stations (BSs). Constraints on the backhaul throughput make full coordination still challenging and typically clusters of BSs are organized to serve the user equipments (UEs). Joint precoding within each cluster is designed to deal with intra-cluster interference. Moreover, inter-cluster interference (ICI) can be reduced by implementing dynamic clustering, i.e., by changing BS clusters over time to provide more fairness among the UEs. In this work we assume that UEs are equipped with multiple antennas and use an interference rejection combiner to suppress the ICI not managed by precoders at transmitter side. In this framework, we develop an algorithm that dynamically organizes clusters and schedules UEs in each cluster by requiring a channel state information at the transmitter which is independent of the number of receiving antennas. Simulations provide two main results: a) a considerable improvement is achieved by adding antennas at the UE and b) the gain of dynamic clustering over static clustering sensibly decreases by equipping the UEs with more antennas. I. I NTRODUCTION In cellular downlink networks coordinated multi-point (CoMP) schemes provide a huge gain with respect to single cell processing (SCP) schemes: the performance degradation in SCP due to inter-cell interference can be perfectly nulled by allowing full sharing of user equipment (UE) data and channel state information (CSI) among the base stations (BSs), which, in turn, serve the scheduled UEs by using joint precoding [1]. This scheme is also known in literature as CoMP with joint processing (JP), opposed to CoMP with coordinated beamforming (CB) where only CSI is shared among the BSs and inter-cell interference is reduced by properly designing the precoder used by each BS while serving its own UEs [2]. Even if CoMP-JP is a very promising technique, many practical constraints make its implementation still challenging. First of all, a backhaul network able to support UE data sharing among the BSs is required. Typically, clusters of BSs are organized and joint precoding is implemented only within each cluster. However, UEs at the cluster border still suffer inter- cluster interference (ICI), which is managed either by dynamic clustering [3], where clusters change over time adapting to channel conditions, or by designing precoders in each cluster in order to limit the ICI caused to UEs in neighbouring clusters [4]. A second issue is that CSI at the BSs may be unreliable because of noise on channel estimation in time division duplex (TDD) systems or limited bandwidth available for feedback in frequency division duplex (FDD) systems. Most of the works on dynamic clustering [3], [5], [6], [7], assume that UEs are equipped with only one antenna, and propose different methods to optimize clusters [3], [5], also jointly either with feedback design [6] or UE selection [7]. However, the LTE-Advanced standard developed by the 3rd Generation Partnership Project (3GPP) assumes that UEs may be equipped with up to eight antennas [8]. Even if this number seems a bit optimistic for current mobile devices, the technological innovation may allow in the near-future manufacturing smartphones or tablets with numerous antennas. Thus, much more attention should be paid to the study of CoMP schemes when UEs may exploit the benefits of multiple antennas [9]. In this work we study the achievable performance in downlink FDD CoMP systems when UEs are equipped with multiple antennas. We start from the dynamic clustering and scheduling algorithm proposed in [10]. To simplify CSI at the transmitter side, we consider a feedback scheme that requires a feedback bandwidth independent of the number of UE antennas. Moreover, we also assume that BSs transmit only one stream of data to the scheduled UEs, which, in turn, implement an interference rejection combiner (IRC) [11] to suppress the residual interference. Then, we compare the proposed strategy against both static clustering, where clusters do not change over time, and single cell processing, where no cooperation among the BSs is allowed. Numerical results show that a) an important gain is achieved by equipping UEs with many antennas to suppress ICI and b) the gain of dynamic clustering over static clustering and SCP sensibly decreases as the number of UE antennas increases. Hence, the main finding of this paper is that the complex operation of dynamic clustering is not worthy when the number of UE antennas is large enough. Notation: We use (·) T to denote transpose and (·) H conjugate transpose. 0 N×M denotes the matrix of size N × M with all zero entries, I N the identity matrix of size N , tr(X) the trace of matrix X, [X] n,m the entry on row n and column m of X, and [X] ·,m the m-th column of X. II. SYSTEM MODEL We consider a system where a set J = {1, 2,...,J } of BSs, each equipped with M antennas, is serving a set K = {1, 2,...,K} of UEs, each equipped with N antennas, with K ≥ JM . As the overall number of transmitting antennas 978–1–4673–6190–3/13/$31.00 c 2013 IEEE