Research On MIMO Detection Technology In Wireless Communications

Multiple-input multiple-output (MIMO) which is based on the system deploy with multi-antennas in transmit and receive side, is regarded to be one of the key technologies for next generation wireless communication systems. It exploits the spatial potential of wireless channels so as to make use of random fading and spatial multi-paths to achieve the speedup of data rate and the improvement of reliability. This dissertation pays attention to the MIMO receiver under uncoded spatial multiplexing environment, i.e., transmit sends separately the uncoded data streams to each antenna, and the interesting area is detection algorithm.Due to the specialty of channel in wireless communication systems as well as restrict of system cost, the design and realization of algorithms in base-band digital processing always meets with the conflict between performance and complexity. While in the case of spatial multiplexing, such conflict becomes even worse. Because generally the MIMO technology accompanies with quite high data throughput. Meanwhile, spatial multiplexing implies the increase of the number of data streams, in other words, several times of throughput increase. At the same time, data services require extremely low link error probability, while unfortunately the spatial multiplexing will introduce the impact of the multi-stream interference (MSI). In summary, consider the MIMO detection, the requirements from system level upon performance and complexity both grow strict; hence make the aforementioned conflict worse. Therefore strictly speaking, this dissertation addresses the problem of the conflict between performance and complexity within the MIMO detection. Keep above research idea and objective in mind, facing with several MIMO detection techniques, we aim to provide effective approach of dealing with above conflict. In this dissertation, the fundamental model of MIMO system is firstly introduced, after that the discussions on sphere decoding (SD) and its surrounding techniques, search space predetermined detector (SSPD) algorithm, the accelerating effect of parallel computing on algorithm implementation as well as the alleviation effect of automatic retransmission request (ARQ) on complexity pressure are performed. The main contributions and innovative research findings are listed as follows:1 As far as is known, few literature deals with the quantified relationship between complexity and detection performance of SD. Sphere radius is proposed to be a transition parameter, thus the quantified relationship is analyzed theoretically. This part of work can be divided into three parts:1) In order to analyze the performance of SD, the asymptotic bound of optimal performance is utilized, and latter the performance loss caused by search space reduction is quantified with the help of above bound. Since the performance loss is affected by the sphere radius directly, sphere radius can be used to represent the performance feature of SD.2) Based on the characteristic of complexity of SD, derive an upper-bound of average complexity; therefore the sphere radius becomes the metric of average complexity.3) By making use of sphere radius, the bridge between performance and complexity can be built. The compute simulations verify above analysis for different MIMO use cases.2 It is difficult for existing radius determine methods to maintain their detection performance while keeping the complexity of SD under control. Due to the tight connections of sphere radius with performance and complexity, the study of its setting problem is interesting for us. Two methods of radius determine are proposed which called controllable performance loss based radius determine method and radius determine method without diversity loss respectively. Meanwhile, an extended application of the latter is shown, i.e. set radius to remove the redundancy of breadth-first search style detectors so as to reduce the power consuming. The simulations validate these two methods.3 Since the heuristic search method leads to uncontrollable complexity and existing detection algorithm cannot give consideration to both performance and complexity, the SSPD algorithm is proposed, which determines the search space even before the processing of search and this behavior is evidently different from that of SD. The search space determine strategy behind SSPD, firstly takes the channel condition into consideration then can adaptively change the size of search space, secondly, use the hardware capability to bound the search space. The impacts caused by imperfect channel and noise estimates are investigated as well, in order to evaluate the algorithm in real environments. Furthermore, extend the SSPD to generate the soft output, which incur tiny extra complexity. Finally, in the block process model, introduce the idea of dynamic computing power allocation to SSPD, and then make it work better.4 The ideas of accelerating the implementing of algorithm and alleviating the complexity pressure through the approaches outside the algorithm itself are provided. The capability of parallel computing to speedup the implementing is exploited. After providing the general methodology of parallel design for complicated algorithms, apply it to KSP based MIMO detector and obtain a memory cost efficient parallel algorithm. Show that the ARQ can alleviate the complexity pressure of MIMO detection, because ARQ is able to bring in the time diversity gain. A practical example is provided as well...