Research On Adaptive Resource Allocation For OFDMA Systems With Imperfect Channel State Information

With the development of wireless cellular networks from first generation toLTE(Long Term Evolution) systems which have been partially put into thecommercial application，requirements of quality of service(QoS) supported bynetworks for users is increasing. With limited spectrum resources，how to provideusers with more high-quality services while maximizing system spectral efficiencyis major technical challenge faced by future wireless broadband cellular networks.OFDMA can achieve flexible and efficient resources allocation and has theadvantages of high spectral efficiency，so it is especially suitable for downlinktransmission with low requirement of peak to average power ratio(PAPR).Worldwide interoperability for Microwave Access and LTE have chosen OFDMA asits downlink transmission scheme on physical layer. Single carrier frequencydivision multiple access (SC-FDMA), can be considered as normal OFDMA withthe DFT precoding. Its PAPR is much lower than OFDMA. So, SC-FDMA is verysuitable for uplink transmission. LTE system has chosen SC-FDMA as its uplinktransmission scheme on physical layer. In this dissertation, the single-cellsingle-input single-output cellular systems with the OFDMA downlink andSC-FDMA uplink is called OFDMA system, the research work of this dissertation isundertaken based on this architecture.The instantaneous signal to noise ratio (SNR) on subchannels is used asChannel State Information (CSI) to indicate the channel state of subchannels.Precise CSI is very important to adaptive resource allocation. In the actual system,because of the restrictions of pilot overhead and so on，there must be a channelestimation error. There is a certain delay from the instant of channel estimation tothe instant of data transmission with resource allocation scheme based on the CSIand this delay is called CSI delay. The change of channel state during CSI delay willlead to inaccuracy of CSI. In FDD systems, CSI obtained through the channelestimation by users is digitized and feed back to base station (BS) via a feedbackchannel, so there are a quantization error and CSI feedback errors. In summary，theCSI obtained by BS is always inaccurate. The estimated SNR on subchannels foruser is called imperfect channel state information (ICSI) in this dissertation. Inactual systems, application of adaptive resource allocation can cause degradation ofperformance of systems when assuming that the CSI is accurate. ICSI eatimationaccuracy has a great influence on the performance of systems. Therefore，studying how to design resource allocation algorithm adaptive to ICSI estimation accuracy toensure users’s QoS while maximizing the system spectral efficiency is verynecessary.For the problem mentioned above，with the statistical characteristics of wirelessfading channel serving as entry point，the problems of resource allocation fornon-real time service is studied for OFDMA-TDD downlink systems，OFDMA-FDDdownlink systems and OFDMA uplink systems，respectively in this dissertation.Some practically valuable methods are obtained. In this dissertation, the following3aspects are studied:Firstly, for TDD downlink system，for the problem to maximize spectralefficiency of downlink while meeting bandwidth demands and the average bit errorrate(BER) constraints of users，a resource allocation algorithm based on bandwidthdemands of users for TDD downlink system is proposed. The user weights isintroduced and subchannel allocation strategy to meet bandwidth demands of usersis proposed. Based this strategy，the closed expression of average BER for user withchannel estimation error and CSI delay is achieved，and through this expression theabove problem is divided into the subchannel allocation problem and the problem ofmodulation mode switching threshold optimization subject to average BERconstraint. This J1dimensional problem of modulation mode switching thresholdoptimization is transformed into one-dimensional optimization problem throughLagrange multiplier method, and it is proved that the optimal solution to thisone-dimensional optimization problem must be obtained on the border of theaverage BER constraints so that the optimal solution is achieved. This algorithm cannot only satisfy bandwidth demands of users，but also adapt to the change of ICSIaccuracy so as to provide the corresponding optimal vector of modulation modeswitching threshold and maximize spectral efficiency while meeting the averageBER constraints for users. As the closed expression of average BER for user isachieved, this algorithm is without iteration and can be easily applied to the actualTDD system.Secondly, for FDD downlink system，for the resource allocaton problem(sameas above)，a resource allocation algorithm based on bandwidth demands of users forFDD downlink system is proposed. Subchannel allocation strategy to meetbandwidth demands of users is proposed. Based this strategy，the closed expressionof average BER for user with channel estimation error，CSI delay，the number ofquantization bits and feedback BER is achieved, and the above problem is dividedinto the subchannel allocation problem and modulation scheme optimization problem subject to average BER constraint through this expression. For modulationscheme optimization problem subject to a average BER constraint, the feasiblesearch algorithm for optimal modulation scheme is proposed. The resourceallocation algorithm based on bandwidth demands of users for FDD downlinksystem can not only meet bandwidth demands of users，but also adapt to the changeof ICSI estimation accuracy so as to provide the corresponding optimal modulationscheme and maximize spectral efficiency while meeting the average BER constraintfor users. In addition, because of achieving the closed expression of average BERwith the parameters of ICSI accuracy for user, so this algorithm can be easilyapplied to the actual FDD system.Thirdly, for the uplink, BSs in both FDD systems and TDD systems obtain ICSIin the same manner, the proposed algorithm is applicable for both systems. For theproblem of maximizing goodput of users, a rate allocation algorithm on subchannelsis proposed. The conditional probability density function (PDF) of the average ofSNR on subchannels occupied by user is achieved. Based on the conditional PDF，the expression of goodput of user is obtained，which is related to estimation errorsand CSI delay and the optimal rate allocation scheme on subchannels is achieved.This algorithm can adapt to the change of ICSI estimation accuracy so as to providethe corresponding optimal rate allocation scheme on subchannels and maximizegoodput of users. Because of obtaining the expression of goodput about channelestimation errors and CSI delay for user, so this algorithm can be easily applied tothe actual uplink systems.