On the use of multiple antennas in a downlink of wireless systems

The role of multiple antennas in wireless communications has evolved greatly in the recent years. For point-to-point communication, Multi-Input Multi-Output (MIMO) technology has emerged as an attractive solution to increase the capacity by providing multiplexing gain and/or to improve the link quality by diversity gain without channel state information at transmitter (CSIT). The first part of the thesis focuses on the use of multiple antennas to increase the reliability of wireless channels via space-time coding (STC). For single-carrier transmission over frequency selective multi-input single-output (MISO) fading channels, we propose a STC scheme that concatenates trellis-coded modulation (TCM) with time-reversal orthogonal space-time block coding (TR-STBC). The decoder is based on reduced-state joint equalization and decoding that operates directly on the TCM trellis without trellis state expansion. We show that, in the limit of large block length, the proposed scheme can achieve the full diversity offered by the MISO multipath channel. Numerical examples show that our proposed scheme offers similar/superior performance at significantly lower complexity with respect to the best schemes previously proposed. The second and third parts of the thesis investigate the use of multiple transmit antennas in a point-to-multipoint downlink system (MIMO broadcast channel) with CSIT. In a multiuser scenario, the MIMO technique can offer even more significant performance gain than in a single user setting by exploiting CSIT. In the second part, we compare STC (transmit diversity) and random ``opportunistic'' beamforming by restricting ourselves to a simple rate feedback. We take into account the following fundamental and realistic aspects ignored in the existing works: random packet arrivals, correlated block-fading channels and non-perfect CSIT. We derive an adaptive policy that stabilizes the transmit queues whenever the arrival rates are in the system stability region. The realistic assumption of non-perfect CSIT yields a non-trivial tradeoff between multiuser diversity achieved by opportunistic beamforming and transmit diversity achieved by STC. In particular, the ability of accurately predicting CSIT emerges as a key factor of multiuser diversity based scheduling. In the third part, we consider a more informative feedback that enables a smarter use of the multiple antennas at the transmitter. In particular, we consider a distortion-wise optimal ``analog feedback" in which each user terminal sends back its estimated channel vector without quantizing and coding over a feedback link. We propose a novel method for user selection and linear beamforming that maximizes a weighted sum rate over a beamforming matrix. Under perfect CSIT, the proposed scheme achieves near dirty-paper-coding performance by exploiting multiplexing gain, power gain, and multiuser diversity. Over a time-varying channel with a given feedback delay, our scheme achieves a smaller average delay than the previously considered STC and opportunistic beamforming for any channel Doppler bandwidth. Dans la troisième partie, nous considérons un retour plus riche qui permet à l?émetteur d?utiliser les antennes multiples d?une façon plus intelligente. En particulier, nous considérons ``le retour analogique?? qui est optimal au sens distorsion dans lequel chaque utilisateur retransmet son vecteur du canal estimé directement sans discrétiser ni coder. Nous proposons une nouvelle méthode pour la sélection des utilisateurs et beamforming linéaire qui maximise la somme pondérée des taux. Avec la connaissance parfaite du canal, le schéma proposé atteint une performance proche de celle de Dirty-Paper coding en exploitant le gain de multiplexage et la diversité multi utilisateurs. Sur un canal corrélé temporellement, notre schéma atteint un délai moyen plus petit par rapport aux autres schémas considérés pour toutes les vitesses de mobile.