Research On Precoding And Interference Coordination For Multiuser MIMO Communication Systems

As an important means of improving the spectrum efficiency of wireless communication systems, point-to-point single-user multiple-input multiple-output (MIMO) technique has been extensively developed during the last decade. However, more complex point-to-multipoint and multipoint-to-multipoint MIMO communication models will inevitably be encountered in actual systems such as cellular and WLAN networks. Both of these MIMO systems contain multiple user receivers. Conventionally, the former is called MIMO broadcast channel, while the latter is called MIMO interference channel. Both MIMO models are of great importance for improving performance in MIMO wireless communication systems, and traditional single-user MIMO transmit strategy is no longer applicable. Therefore, the aforementioned two multiuser MIMO communication systems have drawn wide attention in recent years.This thesis mainly focuses on precoding for MIMO broadcast channels (also called multiuser MIMO downlinks/systems) and interference coordination for MIMO interference channels. The main contribution of our work lies in providing several new algorithms for obtaining simple and efficient transmit strategies, and giving theoretical results on the achievable sum rate.Firstly, a low complexity multiuser MIMO downlink block diagonalization precoding algorithm is proposed. The algorithm eliminates interference by performing Gram-Schmidt orthogonalization to zero forcing beams, and then obtains precoding matrices by maximizing single-user MIMO capacity. It has been proved that our algorithm also achieves the optimal sum rate under the no-interference constraint. Avoiding high complexity matrix SVD in the process of interference elimination, the computational complexity of our proposed algorithm has been greatly reduced compared with the traditional one, and it can be reduced by fifty percent.Secondly, we propose a low complexity multiuser MIMO downlink iterative zero forcing precoding algorithm applicable to a specific scenario where each user is allocated one data stream. In the algorithm, precoding vectors are obtained by zero forcing, receive filters are updated via maximum ratio combining, and parallel updating is applied to the iteration process, thus speeding up convergence. As the algorithm is completely based on linear computation, a great amount of computational complexity has been saved. And simulation results show that the proposed algorithm still performs very close to the iterative algorithm which has the best capacity performance.Finally two distributed interference coordination algorithms are proposed to maximize the sum rate of MIMO interference channels. We derive and define the concept of price function by performing first-order Taylor approximation to the sum rate. In the proposed iterative Taylor approximation (ITA) algorithm, users maximize the approximate sum rate one after another via necessary price exchange, thus achieving efficient tradeoff between maximizing his own rate and suppressing interference to others. It has been proved that the distributed ITA algorithm achieves the same sum rate performance with that of the centralized one. Furthermore, we construct a fictitious concave n-person game to analyze the more general asynchronous transmit strategy updating process, and prove the existence and uniqueness of a Nash equilibrium. Based on this game model, an iterative covariance-price updating (ICP) algorithm is proposed for more flexible updating, and simulation results show that the ICP algorithm achieves almost the same capacity performance with that of ITA, while it requires less price information exchange. One major contribution is that we generalize the price concept of single-input single-output (SISO) interference channels to the MIMO case without any additional restrictions, and provide a reasonable explanation from the Taylor approximation point of view. Compared with other distributed algorithms, both our proposed algorithms can be applied to the most general MIMO interference channels and the convergence of our algorithms can be proved.