Research On Adaptive Transmission Schemes For MIMO Downlink System

With the development of growing number of mobile users and growing demands of diverse multimedia services, frequency resources have been becoming more important. Muliple-input multiple-out (MIMO), which is considered as one of key techniques for B3G/4G system, has been proven its protential to offer high sprectral efficiency as well as link reliability without additional power and frequency expenses. To date, research has focused on the single-user point-to-point scenario where the transmitter and receiver each have arrays. More recently, attention has shifted to multiuser MIMO system especially for MIMO downlink system. On one hand, broadband multimedia services are always asymmetric, and downlink link is considered to be the bottleneck of data transmission. On the other hand, adaptive technique could adptively adjust transmission scheme and transmission parameters to match the dynamic changes of wireless channel so as to improve the performance. Therefore, reaseach on adaptive transmission schemes for MIMO downlink system becomes highly important. Supported by the National High Technology Research and Development Program of China under Grant (No. 2003AA12331005 2006AA01Z277) and National Science Foundation of China under Grant No. 60496310, our research concentrates on several representative problems for MIMO downlink system based on MIMO Gaussian broadcast channel information theory. Our research is mainly focused on several aspects as follows: adaptive modulaton, multiuser scheduling, multitraffic transmission and adaptive transmission scheme under imperfect CSI.Firstly, we explore adpaptive modulation (AM) schemes for MMO downlink system based on TDMA, zero-forcing (ZF) precoding and zero-forcing dirty-paper coding (ZF-DPC) respectively. For TDMA, only the user who has the highest rate is supported at one time. For the other two schemes, several users could be simultaneously transmitted. To date, reseach on adaptive modulation schemes are mainly focused on point-point system. In this paper, we investigate AM for MIMO downlink system. For all these different transmission schemes, adaptive power and bit allocations are conducted under the constraint of QoS (demand of bit error rate) and transmit power to maximize the immediate sum-rate. The optimal bit allocation acheme is extremely complex when the number of users becomes huge, due to the fact that the degree of complexity is exponential to the number of users. Therefore several low-complexity scehmes but with vanishing loses of sum-rate are proposed for different transmission schemes respectively, which have linear degree of compleixy with the number of users. Additionally, considering the effect of impectly CSI, AM based on Error Channel Model for TDMA and ZF precoding are proposed under the instantaneous error CSI and statistics of error CSI at the transmitter. And the low-complexity suboptimal approach is given too. It is shown that AM yields severely degraded sum-rate when the imperfectness of CSI is remarkable. Therefore, space-time coding with diversity gain should be the transmission scheme insteadly under such circumstances.Secondely, multiuser diversity scheduling scheme for MIMO downlink system is analyzed. As the capacity-achieving dirty paper coding (DPC) approach is difficult to implement due to its high complexity, in this paper, we focus our attention on downlink linear precoding which has been shown asymptotically optimal in the sum-rate sense as DPC based on the multiuser diversity. Because scheduling scheme depending only on the CSI results in unfairness among users, we first propose a multiuser scheduling scheme with guaranteed fairness. Block diagonalization precoding is employed to decompose the broadcast channel into non-inferencing sub-channels in physical layer (PHY), and the optimal user set is chosen based on proportional fairness scheduling in link layer. System resources such as power and rate are adaptively allocated so as to guarantee fairness at the same time sum-rate is improved. Because the degree of complexity of the optimal scheduling scheme is exponential to the number of users, a low complexity scheme of linear relationship is proposed. And to reduce the complexity imposed by water-fill power allocation, a method of allocating user powers in direct proportion to user weights is proposed when average SNR is asymptotically high. Furthermore, considering the queue sytem with random packet arrivals, a cross-layer scheduling stategy with guaranteed stability for MIMO multiuser downlink system is proposed to keep average queue length finite. Spatial division multiple access (SDMA) is exploited to support several users'simultaneous transmission and ZF precoding is used to remove interference among users in PHY layer. Both user's CSI and queue state information is considered in link layer. When user's packets are generated according to independent Poisson arrival processes, simulation results show that this scheme could achieve large system sum-rate while guaranteeing the stability of the system compared with those only considering CSI in PHY layer. Meanwhile a low complexity but with vanishing losses greedy method is proposed.One of distinct characteristics of next generation wireless system is its sustaining diversive multiuser services to cater for the increasing broadband demands. A novel multitraffic multiuser transmission scheme is proposed for MIMO downlink system. The key idea is that the power is firstly allocated to real-time (RT) users with high priority, then the optimal non RT (NRT) users is scheduled so as to maximize the system sum-rate at the same time guaranteeing RT user's QoS. This scheme could be conducted in two steps: firstly, the feasibility of demands of target rates of RT users is examined; secondly, the resources such as power, rate and scheduling slot are adaptively allocated. And a suboptimal greedy scheduling scheme is proposed with vanishing losses to reduce the complexity of optimal one. The content described above mainly assumes that the transmitter has perfect CSI, which may be infeasible in pratical environment due to the fact of much feedback information. In order to reduce feedback information, orthonormal random beamforming (ORBF) proposed by Sharif achieves the optimal sum-rate capacity of DPC for a large number of users based on asymptotic analysis. In this paper, we propose two adaptive transmission schemes to further increase the system performance based on ORBF. Firstly, ORBF is quickly degrading with decreasing of number of users, and the system tends to be interference-limited with growing average SNR. To overcome the shortcomings of ORBF, a novel multiuser multibeam selection scheme (MBS) is investigated. The key idea of MBS is to feedback more CSI for the user set indentified by ORBF, based on which the transmitter adaptively selects the optimal user set and beam set. The equal power allocation and adaptive power allocation among users are respectively analyzed. Furthermore, the correlation among transmitter antennas is explored for MBS. Simulation results show that the performance improvement is remarkable especially for large correlation factor, when compared with ORBF and whitening ORBF. Secondly, a feedback scheme based on signal to interference plus noise ratios (SINR) theshhold is proposed to further reduce the feedback of ORBF. The relationship between SINR theshhold and sum-rate is given too. Simulation results show that the proposed scheme achieves greatly reducd feedback with almost the same sum-rate compared with ORBF by the proper SINR threshold.