Multiuser multi-antenna systems with limited feedback

Kountouris, Marios

The use of multiple antennas has been recognized as a key technology to significantly improve the spectral efficiency of next-generation, multiuser wireless communication networks. In multiuser multiple-input multiple-output (MIMO) networks, the spatial degrees of freedom offered by multiple antennas can be advantageously exploited to enhance the system capacity, by scheduling multiple users simultaneously by means of spatial division multiple access (SDMA). A linear increase in throughput, proportional to the number of transmit antennas, can be achieved even by using linear precoding strategies if combined with efficiently designed scheduling protocols. However, these promising gains come under the often unrealistic assumption of close-to-perfect channel state information at the transmitter (CSIT). Therefore, at the heart of the downlink resource allocation problem lies that of feedback acquisition. In this thesis, we focus on linear beamforming techniques relying on low-rate partial CSIT. Several methods that allow the base station (BS) to live well even with coarse, limited channel knowledge are identified. One first key idea is based on splitting the design between the scheduling and the final beam design stages, thus taking profit from the fact the number of users to be served at each scheduling slot is much smaller than the total number of active users. This two-stage approach is applied to a scenario in which random beamforming (RBF) is exploited to identify good, spatially separable, users in the first stage. In the second stage, several refinement strategies, including beam power control and beam selection, are proposed, offering various feedback reduction and significant sum rate gains, even in sparse network settings (low to moderate number of users). In channels that exhibit some form of correlation, either in temporal or in spatial domain, we point out that significant useful information for the SDMA scheduler lies hidden in the channel structure. We show how memory-based RBF can exploit channel redundancy in order to achieve throughput close to that of optimum unitary beamforming with full CSIT for slow time-varying channels. In spatially correlated channels, long-term statistical CSIT, which can be easily obtained with negligible per-slot or no feedback overhead, reveals information about the mean spatial separability of users. A maximum likelihood (ML) channel estimation framework is proposed, which effectively combines slowly varying statistical CSIT with instantaneous low-rate channel quality information (CQI). User selection and beamforming techniques suitable for such settings are also proposed. It is demonstrated that in systems with reasonably limited angle spread at the BS, feeding back a single scalar CQI parameter per user is sufficient to perform SDMA scheduling and beamforming with near optimum performance. Limited feedback strategies utilizing vector quantization codebooks are also investigated. In particular, the problem of efficient, sum-rate maximizing CQI design is addressed and several scalar feedback metrics are proposed. These metrics are built upon inter-user interference bounds and can be interpreted as reliable estimates of the received signal-to-interference-plus-noise ratio (SINR) at the receiver side. It is shown that scalar CQI feedback combined with channel directional information (CDI), zero-forcing beamforming, and greedy user selection algorithms can achieve a significant fraction of the capacity of the full CSIT case by exploiting multiuser diversity. An efficient technique that provides the BS the flexibility to switch from multiuser (SDMA) to single-user (TDMA) transmission is provided, exhibiting linear sum-rate growth at any range of signal-to-noise ratio (SNR). Further feedback compression can be achieved if the CSIT information utilized by the scheduler is represented by ranking-based feedback. We show that an integer value is often sufficient in order to identify users with favorable channel conditions. In parallel, it equalizes the channel access probability in networks where users' channels are not necessarily identically distributed and mobile terminals experience unequal average SNRs due to different distances from the BS and the corresponding different path losses (near-far effects).

Communication systems
Eurecom Ref:
© TELECOM ParisTech. Personal use of this material is permitted. The definitive version of this paper was published in Thesis and is available at :
See also: