Resource allocation problems in wireless networks have attracted tremendous attention in the past few decades. Proper resource allocation can significantly improve the network performance and save the precious wireless resources. In this thesis, we consider two practical resource allocation problems in multi-antenna wireless networks.

First, we consider the multi-cell cooperative wireless network. We propose a new backhaul cost metric that considers the number of active directional cooperation links, which gives a first order measurement of the backhaul loading required in asymmetric Multiple-Input-Multiple-Output (MIMO) cooperation. We focus on a downlink scenario for multi-antenna base stations and single-antenna mobile stations. The design problem is minimizing the number of...[ Read more ]

Resource allocation problems in wireless networks have attracted tremendous attention in the past few decades. Proper resource allocation can significantly improve the network performance and save the precious wireless resources. In this thesis, we consider two practical resource allocation problems in multi-antenna wireless networks.

First, we consider the multi-cell cooperative wireless network. We propose a new backhaul cost metric that considers the number of active directional cooperation links, which gives a first order measurement of the backhaul loading required in asymmetric Multiple-Input-Multiple-Output (MIMO) cooperation. We focus on a downlink scenario for multi-antenna base stations and single-antenna mobile stations. The design problem is minimizing the number of active directional cooperation links and jointly optimizing the beamforming vectors among the cooperative BSs subject to signal-to-interference-and-noise-ratio (SINR) constraints at the mobile stations. This problem is non-convex and solving it requires combinatorial search. A practical algorithm based on smooth approximation and semidefinite relaxation is proposed to solve the combinatorial problem efficiently. We show that semidefinite relaxation is tight with probability 1 in our algorithm and stationary convergence is guaranteed. Simulation results show the saving of backhaul cost and power consumption is notable compared with several baseline schemes and its effectiveness is demonstrated.

Second, we consider the multi-antenna base station-assisted D2D network. We propose a dynamic resource allocation scheme to exploit the mixed timescale CSI knowledge structure in a multi-antenna BS-assisted device-to-device (D2D) network. The short-term multi-antenna beamforming control at each transmit device is adaptive to the local real-time CSI. The long-term routing and flow control is adaptive to the global topology and the long-term global CSI statistics of the D2D network. The design objective is to maximize a network utility function subject to the average transmit power constraints, the flow balance constraints as well as the instantaneous physical layer capacity constraints. The mixed timescale problem can be decomposed into a short-term beamforming control problem and a long-term flow and routing control problem. The short-term problem is solved using SDR. The long-term problem is non-convex and we use a convex approximation approach to tackle this challenge. Using the stochastic cutting plane (SCP), we propose a low complexity, self-learning algorithm, which asymptotically converges to the global optimal solution without explicit knowledge of the channel statistics. Simulation illustrated huge performance gains with several reference baselines.