INTRODUCTION Multiple-input multiple-output (MIMO) cellular systems are cooperative wireless systems by design. This is true whether one is considering single-user MIMO (SU-MIMO) transmission with respect to a single mobile having all user- based antennas, or a multi-user MIMO (MU- MIMO) system in which such antennas are spread across multiple users. This is true whether MIMO is applied across collocated base station antennas or across many such antennas spread over multiple locations. In all cases the informa- tion streams and the wireless signals that carry them must cooperate at transmission and/or at reception and decoding. To appreciate the interest in the application of MU-MIMO to various architectures, note that downlink MU-MIMO systems rely mainly on cooperation on the infrastructure (base station) side. This allows the designer to consider improvements in system performance through infrastructure enhancements. 1 Such enhance- ments can consider various types and levels of cooperation, ranging from intracell cooperation in traditional cellular, to intra- and intercell cooperation through network MIMO to form a cooperative cellular network (CCN) [1–3], 2 to flexible intra- and intercell cooperation of inter- twined CCN systems. In all cases the mathemati- cal formulations of MU-MIMO are very similar, albeit for obvious properties such as the nature of fading and path loss between pairs of anten- nas. MU-MIMO can also be considered jointly with user scheduling over transmission resources. Hence, MU-MIMO, together with architectures, opens up novel possibilities in better sharing of resources among users. Efficient use of MU-MIMO, however, pre- sents some nontrivial challenges. Certainly dif- ferences in the nature of cooperation and the cost of such a deployment can make “fair” comparisons between architectures complicated even when done with respect to a basic charac- teristic such as a given cell density. At an even more fundamental level, however, MU-MIMO has an inherent (unavoidable) cost in physical layer (PHY) resources that depends on the type and degree of cooperation. Specifically, for many MU-MIMO techniques channel state information (CSI) must be known at the trans- mitter with sufficient accuracy in order to enable efficient cooperative transmission [4, 5]. As we show, this “required accuracy” is inti- mately tied to both user location and the archi- tecture itself, and can result in less complex architectures outperforming those that exploit more cooperation [6]. It is therefore clear that the consideration of coordinated MU-MIMO architectures has many aspects beyond simply the rates supported for IEEE Communications Magazine • May 2011 70 0163-6804/11/$25.00 © 2011 IEEE 1 User terminals can oper- ate with a single antenna or few antennas and maintain low decoding or transmission complexity. 2 The case of applying MU-MIMO across sta- tions is also an option in (CoMP) coordinated multipoint transmission and reception in Long Term Evolution (LTE). ABSTRACT Multi-user multiple-input multiple-output can be viewed as an interference control technique relying on cooperative transmission and/or recep- tion over multiple antennas. A cooperating set of antennas can be collocated at a common cell site or distributed across multiple sites. As such, MU- MIMO can be applied to a wide variety of archi- tectures ranging from traditional cellular to more elaborate intertwined cooperative designs. We consider examples of such MU-MIMO architec- tures, and study the impact the scheduling criteri- on, cell density, and coordination can have on the average and cell edge user rates across differ- ent designs. Importantly, MU-MIMO has inher- ent physical layer overheads, which, as the examples illustrate, depend on the degree of coordination and the architecture. Such overhead considerations are very important in assessing net rates and, depending on the scenario, can moti- vate designs using less cooperation. ADVANCES IN COOPERATIVE WIRELESS NETWORKING Sean A. Ramprashad and Haralabos C. Papadopoulos, NTT DOCOMO Communications Labs USA, Inc. Anass Benjebbour and Yoshihisa Kishiyama, Radio Access Network Development, NTT DOCOMO Inc. Nihar Jindal, University of Minnesota Giuseppe Caire, University of Southern California Cooperative Cellular Networks Using Multi-User MIMO: Trade-offs, Overheads, and Interference Control across Architectures