http://dspace.cityu.edu.hk:80/handle/2031/746 2013-06-15T14:59:52Z http://dspace.cityu.edu.hk:80/handle/2031/7013 Title: Impact of antenna correlation on multi-user MIMO systems Authors: Wang, Hao (汪浩) Abstract: Multiple-input multiple-output (MIMO) techniques provide a promising solution to enhance the performance of wireless communication systems. Antenna correlation exists in practical MIMO systems for two reasons. First, there may be no sufficient space to separate antenna elements due to the limited physical sizes of the transmitters and/or the receivers. Second, the practical propagation environments may not provide sufficient scatters. In a conventional single-user MIMO system, since antenna correlation results in reduced degrees of freedom (DOF) and may severely degrade the system capacity, it has been generally regarded as a negative factor. However, the impact of antenna correlation in multi-user environments is still limited explored. In this thesis, we make a comprehensive study on theoretical and practical aspects of multi-user MIMO systems with correlated fading. In the first contribution, the capacity of correlated MIMO systems with full channel state information (CSI) at both the transmitters and the receiver over multiple access channels (MACs) (i.e. uplink multi-user MIMO systems) is analyzed. In contrast to the common views, we show that antenna correlation is potentially beneficial in a multi-user environment. The key is that the spatial diversity related to user locations (i.e., multi-user diversity) can compensate the loss of DOF due to antenna correlation. More specifically, it is shown numerically that there is a cross point between the capacity curves for systems with and without correlation. Below this point, correlation is advantageous and vice versa. Moreover, such a point occurs at a rate increasing with the number of mobile units (MUs) (denoted by K in this thesis), which implies that the range where antenna correlation is beneficial increases with K. We also quantify this advantage analytically in the limiting case of K → ∞. In the meanwhile, it is shown numerically and analytically that there is a similar advantage from antenna correlation for MIMO MACs with rate constraints. (We call this advantage correlation gain in this thesis.) In the second contribution, we study the impact of antenna correlation on the capacity of MIMO MACs with imperfect CSI at the transmitter (CSIT). We first consider the case of no CSIT.We prove that isotropic inputs (i.e., the covariance of transmitted signal for each MU is identity matrix) are the most robust and optimal ones, and therefore achieve the capacity of such systems. Both numerical results and theoretical analysis show that although antenna correlation is detrimental in all rate or power range, the capacity degradation decreases with K increasing and vanishes when K → ∞. These results imply that besides multi-user diversity, CSIT also plays an important role in exploiting correlation gain. We then study the systems with partial CSIT in form of channel covariance information (CCI). Our major finding is that, similar to the scenario of perfect CSIT, antenna correlation is potentially beneficial in a multi-user environment. We also prove that, when antennas at MUs are fully correlated, systems with CCI at transmitters can obtain exact the same correlation gain as that in systems with perfect CSIT. This indicates that CCI may be enough to exploit the potential benefit of antenna correlation. In the third contribution, we extend the results from the MAC scenario to the broadcast channel (BC) scenario (i.e., downlink multi-user MIMO systems). Numerical results show that, similar to MIMO MACs, antenna correlation can potentially improve the capacity of MIMO BCs. We point out that, besides multiuser diversity and power focusing effect, such a gain mainly comes from the fact that antenna correlation can increase the variance of the channel gain, which has already been regarded as a advantageous factor in systems with user scheduling. We also quantify the correlation gain in BCs for the limiting case of K → ∞ when antennas at the base station (BS) and/or MUs are fully correlated. Theoretical analysis shows that the asymptotic correlation gain for the case of full correlation at both the BS and MUs grows linearly and logarithmically with the antenna number at the BS, and logarithmically with the antenna number at each MU. In the final contribution, we consider the practical implementation aspects of correlation gain in coded MIMO multiple-access systems. To decrease the complexity of MIMO transmission under imperfect CSIT, we propose two types of low-cost but asymptotically optimal strategies, i.e. the instantaneous maximum eigenmode beamforming (MEB) and statistical MEB strategy. Interleave-division multiple-access (IDMA) technique, as a low-cost iterative multi-user detection (MUD) approach, is adopted to alleviate inter-user interference resulting from multi-user concurrent transmission. Simulation results demonstrate that our proposed transceiver (i.e., MEB-based IDMA system) is an effective platform in practice to obtain the aforementioned correlation gain. In summary, this thesis presents a comprehensive study on antenna correlation in multi-user MIMO systems. Both numerical and analytical results show that antenna correlation is potential advantageous in a multi-user environment. Such an advantage mainly comes from three aspects. First, the loss of the spatial DOF due to antenna correlation is compensated by multi-user diversity. Second, antenna correlation enables focusing power. Third, the variance of the channel gain is enlarged by antenna correlation, which is beneficial in systems with user scheduling. The finding in this thesis is useful in practice as minimizing the physical size of MUs and/or the BS is highly desirable, but it may result in antenna correlation. Notes: CityU Call Number: TK5103.4836 .W36 2012; xi, 91, iii leaves : ill. 30 cm.; Thesis (Ph.D.)--City University of Hong Kong, 2012.; Includes bibliographical references (leaves [83]-91) 2012-01-01T00:00:00Z http://dspace.cityu.edu.hk:80/handle/2031/6960 Title: An investigation into the use of coil-resonators in domino forms for wireless power transfer Authors: Zhong, Wenxing (鍾文興) Abstract: This thesis presents a study on the wireless power transfer systems using coil-resonators in domino forms (called domino-resonator systems or domino-systems in this thesis). The study is based on the lumped circuit model of domino-resonator systems and aims to analyze the operating principles and find out the optimum operations of the systems under different operating conditions. Straight and circular domino-resonator systems with one source resonator (transmitter) and one loaded resonator (receiver) are firstly studied. Then domino-resonator systems consisting of one transmitter and two receivers are investigated. Analysis methods based on superposition theorem, vector diagrams and power path are proposed. From theoretical analysis, simulation results and practical measurements, a number of findings are reported on the operation and optimization of the domino-resonator systems. Straight domino-resonator systems are studied firstly assuming that the resonators in each system are identical and the operating frequency of the system equals to the resonant frequency of the resonators. Analytical expressions for the optimum load resistance and maximum efficiency are derived for a system with more than 2 resonators when the cross-couplings between non-adjacent resonators are negligible. The optimum arrangement of the resonators in a straight domino-resonator system is then studied. It is shown from theoretical analysis and numerical simulations that the optimum arrangement of the resonators is dependent on the load resistance and that if the load resistance is also optimized then equal-spacing is the best arrangement for a system with 3 resonators, but for a system with more than 3 resonators, the optimum arrangement is not equal-spacing which is different from the equal-spacing arrangement used in domino-resonator systems for waveguide applications. Next the operating frequency is taken into consideration in the study on the straight domino-resonator system. Power flow analysis shows that there exist more than one power paths in a domino-system with more than 2 resonators due to the cross-couplings of non-adjacent resonators. As verified by analysis and practical measurements, the cross-coupling effects of the non-adjacent resonators cause the optimum operating frequencies for both the equally-spaced and the optimally-spaced resonator systems with identical resonators to shift away from the resonant frequency of the resonators so that all the power paths in the system can be utilized optimally to achieve the maximum efficiency. The cross-coupling effects of the non-adjacent resonators increase with increasing number of resonators and decreasing distances of the resonators. Apart from the straight domino-resonator system which normally contains only one main power flow path, the domino-resonator system with two main power flow paths is also studied. One typical example of such systems is the circular domino-resonator system. Based on superposition theorem, individual power flow path of a circular domino-system is analyzed and then their interactions are merged and explained with vector diagrams. It is demonstrated that the optimum operating frequency could be far away from the resonant frequency of the resonators. However, if the operating frequency is fixed, the individual resonant frequency of each resonator in the system should be adjusted, for example by varying the compensating capacitances of the resonators, in order to regulate the power flows of the circular domino-resonator system to an optimum distribution. As the simplest and typical multi-receiver domino-system, a system consisting of one transmitter and two receivers is studied. Theoretical analysis and numerical simulations show that the coupling between two receivers could be either helpful or harmful or even irrelevant to the power transfer efficiency of a system. All possibilities have been tabulated and explained. When the effect of the output power ratio is considered, three methods to obtain a desired output power ratio are introduced. It is proven from the simulation and experimental results that the efficiency can be improved significantly by optimizing the compensating capacitances in some cases, while in other cases operating at the resonant frequency of the resonators is already good enough. This phenomenon is favorably explained with the help of the power path concept. In addition, it has been found that a strong coupling between the receivers can help achieve a higher efficiency within a large variation of the output power ratio. Notes: CityU Call Number: TK7895.P68 Z45 2012; vii, 159 leaves : ill. 30 cm.; Thesis (Ph.D.)--City University of Hong Kong, 2012.; Includes bibliographical references (leaves 153-159) 2012-01-01T00:00:00Z http://dspace.cityu.edu.hk:80/handle/2031/6959 Title: System identification algorithms and techniques for systems biology Authors: Zhan, Choujun (詹儔軍) Abstract: Mathematical models for revealing the dynamics and interaction properties of biological systems play an important role in computational systems biology. This PhD work is motivated by the current difficulty in system identification of dynamic biochemical pathways, given limited highly noisy and spare time-course experimental data. In this thesis, the inverse problem of identifying unknown parameters of dynamical biological systems, which are modelled by ordinary differential equations (ODEs) or delay-differential equations (DDEs), is treated using experimental data. In some cases, even the model can sufficiently describe the measured data, it is still important to infer how well the model parameters are determined by the amount and quality of the available experimental data, which is essential for investigation of model prediction. For this reason, another key topic in this thesis is identifiability analysis. The main contributions of this PhD work are summarized as follows: 1. In many cases, bio-system models are autonomous systems, which are linear in parameters. For this type of models, an optimization-based parameter estimation approach is proposed. Spline and numerical differentiation methods are used to smooth noisy observations and to estimate the time derivative of the underlying dynamical system, respectively. Subsequently, the parameter estimation problem can be reduced to a Least-Squares Parameter Estimation (LSPE) or a Linear Programming Parameter Estimation (LPPE) problem, which can then be efficiently solved by many global optimization algorithms. 2. For general bio-system models, a parameter estimation method combining spline theory with Nonlinear Programming (NLP) is developed. This method removes the need for ODE solvers during the identification process. Our analysis shows that the augmented cost function surface used in the proposed method is smoother; which can ease the optimal searching process and hence enhance the robustness and speed of the search algorithm. Moreover, the core of our algorithms is NLP based, which is flexible and where consequently additional constraints can be embedded/removed easily. 3. In practice, time-delay feedback pathways exist in many biological systems, which can be modelled by continuous delay-differential equations (DDEs). In this work, a two-stage approach is adopted for parameter estimation: first, by combining spline theory and NLP, the parameter estimation problem is formulated as an optimization problem with only algebraic constraints; then, a new differential evolution (DE) algorithm is proposed to find a feasible solution. The approach is designed to handle problems of realistic sizes with noisy observation data. 4. Identifiability analysis of the so-called S-system is given. The basic theory is developed and the structural identifiability of the S-system is proved. This work also analyzes the limitation of existing structural identification approaches, revealing that these approaches face the risk of the overfitting/underfitting problem. Notes: CityU Call Number: QH324.2 .Z45 2012; xiv, 4, 160 leaves : ill. 30 cm.; Thesis (Ph.D.)--City University of Hong Kong, 2012.; Includes bibliographical references (leaves 149-160) 2012-01-01T00:00:00Z http://dspace.cityu.edu.hk:80/handle/2031/6958 Title: Parallelization of competitive networks Authors: Xiao, Yi (肖懿) Abstract: In this thesis, two kinds of parallel implementations of competitive networks are involved. Firstly, a GPU-based software parallel implementation is presented for the Linde-Buzo-Gray (LBG) and self-organizing map (SOM) methods. Based on the topological preserving property of a SOM network, a high dynamic range texture compression algorithm is also presented. Secondly, an analysis on the convergence time is done for the dual neural network based k-winners-take-all (kWTA) network, which is a hardware parallel implementation of the kWTA networks. A GPU implementation for LBG and SOM training The LBG and SOM methods are two popular learning algorithms in competitive networks. They are applicable in a wide range of areas, such as vector quantization, image compression and pattern recognition. However, the training process of them is time-consuming. Nowadays, desktop computers are usually equipped with programmable graphics processing units (GPUs), whose parallel data processing ability is ideal for training acceleration. Although there are some GPU algorithms for LBG and SOM training, their implementations suffer from a large amount of data transfer between CPU and GPU, and a large number of rendering passes within a training iteration. This thesis presents a novel GPU-based algorithm for LBG and SOM training. In each training iteration, we use a vertex shader to calculate the winner for each training vector. Afterwards, we utilize a fragment shader to update the network. Therefore, the overheads mentioned above are reduced. In the experiments, we apply the LBG and SOM in vector quantization to evaluate the training performance. Our experimental results show that our approach can run much faster than previous approaches. SOM-based color palette for high-dynamic range texture compression High dynamic range (HDR) images are commonly used in computer graphics for accurate rendering. However, it is inefficient to store these images because of their large data size. Although vector quantization approach can be used to compress them, a large number of representative colors are still needed to preserve acceptable image quality. To tackle the problem, an efficient color quantization approach is proposed to compress HDR images. In the proposed approach, a 1D/2D neighborhood structure is defined for the SOM approach. The SOM approach is then used to train a color palette. Afterwards, a virtual color palette which has more codevectors is simulated by interpolating the trained color palette. The interpolation process is hardwaresupported in the current graphics hardware. Hence there is no need to store the virtual color palette as the representative colors are constructed on the fly. Experimental results show that our approach can obtain good image quality with a moderate color palette. Analysis on the convergence time of dual neural network-based kWTA A k-winners-take-all network (kWTA) is able to select the k-largest numbers from n inputs. Recently, a dual neural network (DNN) approach was proposed to implement the kWTA process. Compared to conventional approaches, the DNN approach is with much less interconnections. Moreover, a rough upper bound on the convergence time of the DNN-kWTA model was given in terms of input variables. Instead of an upper bound, this thesis derives the exact convergence time of the DNN-kWTA model based on the theory of ordering statistics. With our result, we do not need to spend extensive time in simulating the network dynamics to estimate the convergence time. Once the sampling inputs are given, we can easily estimate the statistical behavior of the convergence time. Finally, we theoretically study the statistical properties of the convergence time when the inputs are uniformly distributed. From the basic probability theory, we can convert a non-uniform distribution into a uniform distribution. Also the conversion preserves the ordering of the inputs. Hence, our theoretical result on the convergence time is also valid for non-uniformly distributed inputs, if we convert the inputs to the uniformly distributed manner. Notes: CityU Call Number: QA76.87 .X53 2012; xvii, 105 leaves : ill. 30 cm.; Thesis (Ph.D.)--City University of Hong Kong, 2012.; Includes bibliographical references (leaves 95-105) 2012-01-01T00:00:00Z