Transceiver Design for Multiuser MIMO Systems

Abstract

The use of multiple antennas at both base stations and mobile stations (multipleinput-multiple-output (MIMO)) increases the spectral efficiency and reliability in wireless communications. Multiuser MIMO communication, especially the point-tomultipoint downlink transmission is substantially more complicated compared with a single-user communication environment. The focus of this thesis is the joint design of linear transmitters and receivers (transceivers) for point-to-multipoint transmission in multiuser MIMO systems. Various optimization problems with respect to minimum mean square error (MMSE) and rate will be investigated. The non-convexity of these problems makes the derivation of globally optimal algorithms very difficult. We propose a new duality-based approach for downlink optimization under a total power constraint. Instead of optimizing directly, we set up equivalent uplink/downlink channels and show that under a total power constraint, any MSE point which is achievable in the uplink can be achieved in the downlink as well. As a direct consequence of this MSE duality, the complicated downlink optimization can be carried out efficiently by focusing on the equivalent uplink problem. Additionally, a heuristic alternating optimization strategy is proposed. Since the MSE duality ensures that the same MSEs can be achieved in both links, it is possible to optimize the powers and filters in an alternating manner by switching between the equivalent uplink channel and the downlink channel. This alternating approach is not only applicable to MMSE optimization, but also to rate optimization. To derive an algorithm from this framework, the essential issue is to specify the uplink power allocation according to the concrete optimization problem. New optimal power allocation strategies are derived for the problems considered here. Besides these duality-based approaches, we propose another framework, which performs optimization in the downlink directly. This framework exploits the convexity of certain MMSE-type transmitter optimization problems. We show that these problems can be reformulated as second order cone programs (SOCPs). The optimal linear receiver is known as the linear MMSE filter. We prove that the algorithms derived from this framework converge monotonically. Application examples are studied for the scenarios of multicasting and base station cooperation in network MIMO systems. However, the proposed optimization framework is very general and not restricted to the application examples. Any convex constraints can be included without changing the mathematical structure, e.g., different power constraints, such as a