Multiuser Detection In Space-Time Coded MIMO/ SDMA Systems

Recently, Multiple-Input Multiple-Output (MIMO) assisted Space-Division Multiple-Access(SDMA) systems (MIMO/SDMA) have drawn wide research interests, in order for meetingthe demands of high spectral efficiency and for providing high speed wireless transmissionservices. In this thesis, the performance of MIMO/SDMA systems using Space-Time BlockCoding (STBC) is investigated, when the MIMO/SDMA signals are transmitted over Rayleighfading channels. In the considered MIMO/SDMA systems, each mobile user employs two orfour transmit antennas and transmits symbols with the aid of their corresponding Space-TimeBlock Codes (STBCs), while the base-station (BS) receiver is equipped with multiple antennasfor achieving the objectives of spatial diversity, multipath fading mitigation, and/or supportingmultiple-access communications. For the purpose of efficiently suppressing the multiple accessinterference (MAI), which may result in severe near-far problem in multiple-access systems in-cluding the considered MIMO/SDMA, in this thesis the received MIMO/SDMA signals at theBS are detected with the aid of various multiuser detectors (MUDs) associated with advancedspace-time processing algorithms.In the context of the detection in the considered MIMO/SDMA systems, in this thesis a rangeof sub-optimal MUDs are investigated, since the sub-optimum MUDs have relatively lowercomplexity in comparison to the optimal MUD, which is originally derived for code-divisionmultiple-access (CDMA) systems based on the maximum likelihood principles. Specifically,both linear MUDs and decision-feedback MUDs (DFD) are considered. The linear MUDsinclude the linear decorrelating MUD and the linear minimum-mean squared-error (MMSE)MUD. By contrast, the DFDs considered include the decorrelating DFD, MMSE-DFD, modi-fied versions of decorrelating DFD as well as multistage DFD. Furthermore, the direct matrixinversion (DMI) MMSE MUD is proposed for the case, where the a-priori channel knowledgeis unavailable at the receiver.In this thesis, a range of adaptive MUDs are also investigated with emphasis on the adaptiveMMSE-assisted MUDs. This is because the MMSE-assisted MUD facilitates the implementa-tion using adaptive techniques. Specifically, we first consider the adaptive MMSE MUDs basedon gradient techniques. Then, three types of reduced-rank MMSE MUDs are investigated in con-junction with the MIMO/SDMA systems for the purpose of achieving low detection complexity.From our study and results, it can be shown that the reduced-rank MMSE MUDs constitute arange of low-complexity MUDs with high-efficiency in comparison with the full-rank MMSEMUDs.In this thesis, a range of simulation results are provided in order to characterise the perfor-mance of the MIMO/SDMA systems, which may employ various detection schemes. Our studyand simulation results show that the considered sub-optimal MUDs conventionally employed inCDMA systems are near-far resistant and potentially capable of achieving significant improve-ment of BER performance over the correlation-based single-user detector in the MIMO/SDMAsystems. The considered MUDs are capable of facilitating joint space-time processing and re-ceiver diversity combining, while simultaneously suppressing the multiuser interfering signals.In the context of the adaptive MUDs, our simulation results demonstrate that the adaptive MUDassisted by the steepest-descent algorithm is capable of approaching to the ideal MMSE solution.When the BS has no a-priori knowledge about the communication channels, the MIMO/SDMAsystems can achieve satisfactory BER performance by employing the DMI-assisted MMSEMUD or the adaptive MMSE MUDs based on the LMS family of algorithms. Our results showthat the adaptive MMSE MUD based on sliding-window LMS outperforms the adaptive MMSEMUD that is based on the conventional LMS algorithm, at the expense of an increase in com-putational complexity. Furthermore, our simulation results demonstrate that, in time-varyingMIMO channels, there exists an optimal window-length in DMI-based MMSE MUD, whichresults in that the corresponding DMI-based MMSE MUD achieves the best BER performance.This observation is very different from that observed in conventional CDMA systems, where theperformance of the DMI-based MMSE MUD becomes better when increasing the length of thesliding-window. Finally, our study and results show that the MIMO/SDMA systems utilising thereduced-rank MMSE detection can be capable of achieving a satisfactory trade-off between theaffordable detection complexity and the attained BER performance.