Events Calendar

Stability Optimal Resource Allocation in Gaussian MIMO-BC: Theory and Practical Approaches

Speaker: Dr. Mari Kobayashi, Technical University of CatalunyaAbstract: In this talk, we address the resource allocation problem in the Gaussian MIMO broadcast channels (BC) where transmission buffer queues destined to each user are located in the transmitter.Under the random packet arrival assumption, stabilizing the transmission buffers is the single most important criterion for "fairness''. We review the well-known results on MIMO-BC and MIMO-Multiple Access Channel (MAC), stability region, and an adaptive policy achieving the largest stability policy (referred to "max-stability policy"). In the first part, we assume perfect channel state information at transmitter (CSIT). We propose a simple iterative waterfilling algorithm to implement the max-stability policy based on the capacity achieving dirty-paper coding (DPC). As a more practical signaling strategy, we consider linear beamforming because it can be generalized to the non-perfect CSIT case. The comparison in terms of average delay shows that the linear beamforming scheme achieves near DPC performance especially for the case of two transmit antennas.In the second part, we shift to more realistic scenario where CSIT is assumed non-perfect due to the fact that the channels are time-varying and the feedback link is delayed. Under non-perfect CSIT, the rate allocation should take into account the outage event, i.e. the event that the transmission rate is above the rate supported by the channel. We compare two schemes based on linear beamforming with different types of CSI feedback : a novel linear beamforming scheme with "analog feedback" and improved opportunistic beamforming with SNR feedback. In "analog feedback", each user sends back its estimated channel vector without quantizing and coding.It is found that the proposed linear beamforming scheme outperforms improved opportunistic beamforming for any channel Doppler bandwidth with a similar amount of feedback. Specifically, for moderate to fast fading channels, the proposed linear beamforming scheme tends to allocate the whole power to one user by operating in a TDMA mode while improved opportunistic beamforming becomes strongly interference-limited and thus unstable.Bio: Mari Kobayashi received the B.E. degree in electrical engineering from Keio University, Yokohama, Japan, in 1999, the M.S. degree in Radio Communications, and the Ph.D. from Ecole Nationale Supérieure des Télécommunications, Paris, France, respectively in 2000 and 2005. Her current research interests include space-time coding, multiuser communication theory.Host: Dr. Giuseppe Caire, x.04683, caire@usc.edu

Location: Hughes Aircraft Electrical Engineering Center (EEB) - -539

Audiences: Everyone Is Invited

Contact: Mayumi Thrasher