| Summary: | 博士 === 國立臺灣大學 === 電子工程學研究所 === 99 === In this thesis, we propose a new concept that saves the average computational cost of multiple-input multiple-output (MIMO) detection. The detection methods are adapted on-the-fly according to the estimated channel state information subject to just-acceptable error rate (JAER) performance specified by standards, rather than striving to optimal error rates. For example, it is natural to employ only low-complexity MIMO detection solutions, such as ZF
detection or OSIC, under "good" channel scenarios as they achieve acceptable performance with less computational complexity. On the other hand, under "poor" channel conditions, the search-based MIMO detectors seem to be indispensable.
To achieve this goal, error rate models for various MIMO detection algorithms are derived. As these analyses are performed on-the-fly, we apply the proposed delta lattice that simplifies their computations. From these models, a channel-adaptive-MIMO (CA-MIMO) receiver with detector-switching strategies for non-iterative receiver is roposed. Compared with the sophisticated sphere decoder (SD) MIMO detector, the average saving of the CA-MIMO at moderate
signal-to-noise ratio (SNR) is around 65% to 76%, depending on different modulation alphabets. At high SNR, the CA-MIMO detector almost switches to simple detection methods, where its complexity is several orders lower than the SD.
Furthermore, we reduce the complexity of iterative receiver
inevitable for low SNR regime. We follow the Divide-and-Conquer principle. An efficient algorithm of soft-input SD is first proposed that achieves optimal error rate at every iteration. Then, the delta lattice mentioned perviously is adopted to realize a iterative MIMO detector, namely the delta lattice detector(DLD). This DLD is regarded as an exemplar to show that in the iterative scenario, the optimal error rate performance is achieved without performing optimal algorithms at every iterations. Last, with the theoretical error rate analysis for iterative MIMO systems, the JAER concept translates the unnecessary error rate performance to more than 90% saving by imposing the LLR clipping trick.
|
|---|
