Study On Low-Complexity Multi-user Layered Space-Time Coding

As one of the key techniques for future wide-band wireless communication,MIMO can substantially increase the channel capacity and improve the communicationquality. At present, the channel capacity can be explored by using MIMO-based layeredspace-time (LST), while space-time block codes (STBCs) can be used to increase thediversity gain to overcome deep fading. As two important STBCs, orthogonal STBCsand quasi-orthogonal STBCs have low decoding complexity with a relatively low coderate. To achieve the diversity gain and high code rate simultaneously, LST codes andSTBCs can be combined to be multi-user LST codes. The transmit antennas are firstpartitioned into multiple groups (users), and then each group (user) is codedindependently with a space-time code, finally each group’s (user’s) transmit symbols aredecoded at the receiver after cancelling interferences from other groups. Because of itsadvantages, this coding scheme is adopted widely in current multi-user MIMO systems.If maximum likelihood (ML) decoding is used for its decoding, the high complexitymay restrict its applicationsPartial interference cancellation (PIC) group decoding, which groups the transmitsymbols and decodes them independently after interference cancellation, is optimum interms of symbol-error-rate (SER) among all the similar group decoding algorithms withproject matrix. For single-user MIMO systems, full diversity can be achieved byspace-time code design with low complexity PIC group decoding. However, formulti-user MIMO system, its diversity performance and space-time coding are stillunknown.In the thesis, multi-user LST coded system is considered. Perfomance analysis forthe system decoded with the PIC group decoding is done and the methods to improvethe decoding performance are proposed. Furthermore, full-diversity multi-userspace-time coding scheme based on successive interference cancellation (SIC) aidedPIC (PIC-SIC) group decoding are also studied. The achievements are as follows.①The method to obtain the equivalent channel models for multi-user LST codedsystems is proposed. Based on the linear expression of STBCs, the equivalent channelequation for multi-user LST coded systems is derived by linear matrix transformation.Next, it is proved that the projection matrix for the PIC group decoding algorithm canbe further simplified to lower the decoding complexity after interference cancellation. ②The diversity performance of multi-user layered space-time coding with themaximum likelihood (ML) decoding and PIC group decoding algorithm is analyzed andproved. Based on the rank criterion for single-user space-time code design, theML-decoded system diversity gain is proved to be the same as that of the single-userMIMO system with the same antenna configuration and STBC. It is further proved that,for a K-user layered MIMO system, each user equipped with M transmit antennasand N receive antennas and coded with Alamouti and quasi-orthogonal STBC,diversity order M (N-K+1)can be achieved using the PIC group decoding. Theobtained conclusion is a novel and important theoretical result for low-complexitymulti-user LST coded systems.③In order to improve the performance of multi-user LST codes with the PICgroup decoding, by analyzing and comparing the power gains of all possible groupingschemes after interference cancellation under any channel realization, a PIC groupdecoding algorithm with adaptive grouping is obtained. The grouping scheme withmaximum power gain can be adaptively selected in real-time, and thus better systemperformance can be achieved. Simulation results show that the adaptive groupingscheme can obtain about3dB bit-error-rate (BER) improvement compared to thetraditional grouping schemes.④For traditional multi-user LST coded systems, if using group decodingalgorithms including the PIC group decoding, the diversity order may decrease. Bychanging the coding scheme to strengthen the independence among the column vectorsin different groups’ equivalent channel matrices when using group decoding, a codingscheme is proposed and proved to be full diversity with the PIC-SIC group decodingbased on the analysis of symbol pair error probability. Simulation results show that thescheme can achieve better BER performance and higher diversity gain compared to thetraditional LST codes at the same spectrum efficiency.