Downlink scheduling optimization and performance analysis over fading channels in a CDMA network

Due to technological advances in mobile communications, together with the explosive growth of internet access, transmitting multimedia applications over wireless channels is no longer a remote concept. In third generation (3G) multimedia CDMA networks, a variety of techniques will be used to meet the quality of service (QoS) requirements for various types of traffic. These include adaptive modulation and coding (AMC) which improves performance by adapting the employed modulation and coding scheme (MCS) to changing channel conditions and multicode transmission which provides higher bit rates to mobile stations (MS's). The problem of allocating radio resources in the downlink of a CDMA network is first studied. In particular, the modulation and coding schemes, numbers of multicodes, and transmit powers used for all MS's are jointly chosen so as to maximize the total transmission bit rate, subject to certain constraints. In addition, a scheduler which uses knowledge of MS traffic loads is also proposed and shown to yield a significant improvement in throughput. The proposed multiuser schedulers are optimal in terms of system throughput. However, the implementation complexity can be high. Consequently, a suboptimal scheme is proposed, in which resources are allocated sequentially on a per-MS basis. Essentially, the sequential scheme reduces the problem of multiuser resource optimization to a single user problem, thereby greatly reducing complexity. Numerical results show that the performance of the sequential scheme is generally close to optimal if the MS's are ordered optimally. The thesis also addresses the problem of downlink single user scheduling in the context of AMC and multicodes with imperfect channel estimation. Since the selection of MCS and number of multicodes requires an estimation of the downlink channel quality, it is important to assess the performance degradation due to estimation errors. It is shown that the system throughput is quite sensitive to channel estimation errors, and methods are proposed to reduce the resulting degradations. In multiuser scheduling, the aim is to select users with good channel conditions so as to improve the system performance. The selection process is a classical problem in the theory of order statistics. Since users are generally located in different parts of a cell, their respective channels can often be assumed independent, but their fading statistics are not necessarily identical. In this thesis, some useful general results assuming independent and non-identically distributed (i.n.d.) order statistics over the Nakagami and Weibull fading channels are derived. The thesis also proposes a new statistical distribution for the CDMA downlink signal-to-interference ratio given that the simultaneously transmitted interfering and desired signals from the same base station undergo the same fading process. Finally, a simple method for approximating complex statistical distributions which arise in the study of wireless communications is investigated. The resulting relatively simple approximations are shown to be quite accurate.Applied Science, Faculty ofElectrical and Computer Engineering, Department ofGraduat