Resource allocation for MIMO-OFDM systems with cooperative base stations

© 2014 Dr. Chih-yu HsuMIMO-OFDM is a promising wireless technology for current and future telecommunication standards due to its inherent advantages, including high spectral efficiency and data throughput. With knowledge of the channel condition available to base stations (BSs) and user terminals (UTs), dynamic resource allocation algorithms can be employed to optimize system performance subject to physical constraints such as transmit power and per-UT rate targets. This thesis focuses on proposing and analyzing novel resource allocation algorithms for MIMO-OFDM systems in fixed wireless applications, where channel conditions are assumed to be quasi-stationary. We consider the use of cooperative BSs employing coordinated multi-point transmission with joint processing to mitigate the effect of inter-cell interference. We assume that full channel-state information is available to both BSs and UTs, and multiple co-channel UTs are permitted to share the same OFDM subchannel. We first focus on a single-user MIMO-OFDM system where we seek to maximize the discrete sum-rate subject to per-antenna average power constraints across two cooperating BSs. The consideration of per-antenna power constraints limits the average transmit antenna power. This is a practical approach to indirectly control the resulting peak antenna powers such that it helps to limit the occurrences of exceeding the dynamic range of the high-power amplifiers, and therefore results in a higher transmit signal quality. The proposed algorithm utilizes an efficient look-up table for allocating discrete rates in each MIMO-OFDM spatial-subchannel. Furthermore, the proposed algorithm applies an ellipsoid method and analytic center cutting-plane method for finding the optimal solution in a multi-dimensional feasible region. We then extend to a multiuser MIMO-OFDM system where a non-convex continuous weighted sum-rate maximization problem subject to per-antenna average power constraints is considered. We apply a successive convex approximation (SCA) technique to transform the optimization problem into a convex one. This convex transformation results in a suboptimal power allocation algorithm where powers are allocated to a subgroup of co-channel UTs with the least amount of mutual interference. The proposed algorithm is compared with two other power allocation schemes: iterative waterfilling (IWF) and zero-forcing beamforming (ZFB). Numerical results show the proposed algorithm outperforms IWF and ZFB in both low and medium interference scenarios. As expected, ZFB has the best performance in high interference scenarios due to the interference cancellation of zero-forcing beamformers, whereas both SCA and IWF are limited by the severity of multiuser interference. Next, the practical aspect of achieving user fairness in multiuser MIMO-OFDM systems is also considered. We make use of the proposed SCA approach to solve the total transmit power minimization problem subject to both per-antenna average power constraints and per-UT rate targets. This results in a suboptimal power allocation algorithm which the per-UT rate target is satisfied regardless of the channel conditions of UTs. Numerical results demonstrate the performance difference between the proposed algorithm and IWF depends on the mutual interference between co-channel UTs. In general, a higher interference between UTs results in a larger difference in achievable per-UT rate targets for given per-antenna power constraints. Finally, we tackle the practical challenge of discrete sum-rate maximization for multiuser MIMO-OFDM systems. We propose a heuristic discrete algorithm that takes the continuous-rate SCA solution and round it to the nearest integer. The corresponding powers for given discrete rates are then computed by solving a system of linear equations per OFDM-subchannel, followed by redistributing the resulting antenna powers in a greedy approach in order to satisfy per-antenna average power constraints. A discrete sum-rate solution can be obtained with a reasonable number of iterations and computational complexity