| Smart antennas can improve the quality of communication links, greatly increase the capacity of communication systems and enhance the efficiency of frequency, which has become a "hotspot" of research in mobile communication. This dissertation dissertated the basic theory and relevant techniques of smart antennas in mobile communication systems, and detailedly studied on some key problems, which including the spatial-temporal channel property and modeling theory, adaptive digital beamforming algorithms, super-resolution algorithms for DOA estimation (especially, effective algorithms for coherent multipaths environment), interference suppression techniques (especially suppression of coherent interference), joint multi-dimension parameter estimation in spatial-temporal channel, and the 2D spatial-temporal processing.As for the spatial-temporal channel property and modeling theory, firstly, the thesis introduced the property of wireless spatial-temporal channel, and theoretically analyzed the effects on the frequency correlativity, temporal correlativity and spatial correlativity caused by delay spreading, Doppler frequency spreading and angle spreading, respectively. (1) For different scattering environments in mobile communications, we theoretically deduced and numerically simulated the correlativity of smart antennas, and obtained the relationship between antennas correlativity and the distributing property of scattering, which can be provided as theoretical reference for smart antennas designs. (2) As for channel modeling theory, the thesis simply introduced the conventional models including the Lee's model and its improved model, the Geometrically Based Single Bounce (GBSB) model, the Gaussian Wide Sense Stationary Uncorrelated Scattering (GWSUS) model and the Expanded Tapped-Delay-Line model. Then, base on the power-delay spectrum and power-angle spectrum suggested by ITU for different environments, we proposed a method based on statistical property to simulate the spatial-temporal channel. The method is simple and effective, can easily construct different channels. Simulation results confirmed the model can reflect the channel property in spatial domain, temporal domain and frequency domain simultaneously.As to the beamforming techniques, the thesis simply introduced two kind of non-blind algorithms—LMS and RLS, detailedly studied on the two kind of blind algorithms—Constant Modules Algorithm (CMA) and Least Square De-spread Re-spread Multi Target Constant Modules Algorithm (LS-DRMTCMA), and analyzed their performance via simulation experiments. (1) Aimed at the contradictory between convergence speed and the steady-state errs in the conventional CMA, based on a new cost function, the thesis proposed a novel method to control the step-size adaptively. The algorithm uses Gauss function to realize the mapping from constant modulus err to step-size, its parameters are simple to be control, and it has super-linear accelerating effect, can obtain excellent convergence speed, small steady-state err and excellent noise resisting ability simultaneously, which can improve the performance of constant modulus array. (2) In non-stationary channel, the relative power of the signals input in the same stage will change quickly owing to the random fading, however, the CMA simply tends to capture the strongest signal, which might cause the signals captured by each stages to fluctuate, and each stage can't maintain a steadily capture or tracing to the target user. So, based on the improved algorithm proposed before, we further introduced a dissymmetrical mechanism to control the step-size, which can not only ensure each stage to capture the target user quickly, but also ensure each stage to trace the target user steadily, and enhance the anti-fading ability of the array.As to DOA estimation, the thesis simply introduced the commonly used DOA estimation algorithms. The thesis noticed the problems that, MUSIC and ESPRIT can't deal with coherent sources, the number of DOAs that can be estimated can't exceed the element number of array, the algorithms are not fitted for colored noise, and it's difficult to pair the DOAs with the users. Aiming at the cellular system, according to uniform linear array and uniform circular array respectively, the thesis studied on algorithms that suitable for coherent mutipath environment in cellular systems. (1) For uniform linear array, using the fourth order cumulant of array outputs, we realized the blind estimation of spatial signature in uplink, and use it to estimate the multi-user's DOAs in coherent mutipath environment. The algorithm is not depended upon the specific property of signals, can be employed to additive arbitrary Gaussian noise, can estimate 2M2/3 DOAs by M elements, can breakthrough the limit of conventional MUSIC and ESPRIT, and can pair the multipath with users automatically. Through simulation experiments, the relations between estimation variance, SNR and snapshot number were studied, which confirmed the availability and robustness of the algorithm. Based on spatial signature estimation, the thesis constructed a spatial filter bank, and realized the separation of multi-user and the recovery of sources. (2) For uniform circular array, the thesis firstly introduced the element space MUSIC, real beam space UCA-RB-MUSIC and UCA-ESPRIT, then studied and evaluated their performance via simulation experiment. Because all algorithms mentioned above are invalid in coherent environment, and uniform circular arrays are not fit for directly spatial smoothing, the thesis used a pre-processing technique to transform the UCA of element space into a virtual uniform linear array of mode space, and then used spatial smoothing to realize the DOA estimation for coherent sources in UCA.In coherent multipath environment, the performance of smart antenna will decline dramatically. The thesis proposed an improved method to overcome the problem. (1) To overcome the shortcoming of the existing spatial smoothing algorithm, and base on the idea that fully utilize each auto-correlation matrix and cross-correlation matrix of sub-arrays, the thesis proposed an algorithm calledadaptive-globally-weighted-spatial-smoothing (AGWSS), and greatly improved the de-correlation ability of small array. The algorithm is completely adaptive and does not need the prior knowledge of the source angles or pre-estimation of source angles. (2) In AGWSS, for the first time, the thesis proposed an idea of symmetrical processing (in page 121) to ensure the Hermitian property of the smoothed correlation matrix, and, base on this idea, the thesis deduced the closed-form solution of the optimal smoothing weight matrix for the AGWSS (i.e., explicit formulation, not implicit formulation as some other algorithms). (3) Furthermore, considering the requirement of semi-definite to the smoothed correlation matrix in some applications, using symmetrical processing techniques, under the constraints that the smoothed correlation matrix should be Toeplitzian, Hermitian and semi-definite, and based on the Primal-Dual-Interior-Point-Barrier-Function method, the thesis proposed an algorithm called semi-definite adaptive globally weighted spatial smoothing (SD-AGWSS). The algorithm was used to estimate DOAs in coherent environment, and greatly improved the estimating performance, especially, when the coherent sources are close or the SNR is very low, the resolution can be enhanced greatly.In coherent multipath environment, the commonly used beamforming algorithms will become ineffective, and the desired signal in array might be cancelled. The conventional spatial smoothing technique has poor de-correlating ability, and it will reduce the "effective aperture" of array. (1) The thesis proposed an improved beamforming scheme: Firstly, adopting the AGWSS algorithm proposed in this thesis, which, through globally weighted averaging of the auto-correlation matrices and cross-correlation matrix of each subarray, can de-correlate the coherency to the utmost extent; and then, using the linear-constrained-minimum-variance (LCMV) criterion for directly matrix inverse, the optimum weight vector of subarray beamformer can be found; at last, in order to avoid the aperture loss caused by smoothing and fully utilize the residual freedom degrees, the thesis proposed a technique of second-weighting under the criterion of null deepening or noise gain minimization respectively, which greatly improved the ability of array to suppress coherent (or correlative) interference and noise. (2) By theoretical analysis and simulations, the thesis compared the de-correlating ability of different smoothing algorithms for the first time. (3) Using freedom degree theory, for the first time, the thesis analyzed and compared the conditions for subarray division in different smoothing algorithms: for an array with M elements, WSS can suppress ((M-1)1/2-1) coherent interferences, but AGWSS can suppress (M/2-1) coherent interferences. Theoretical analysis and simulations showed that the AGWSS outperforms the existing spatial smoothing algorithms obviously, and especially it's suitable for small smart antennas in mobile communications.The thesis further studied the problem of joint multi-dimension parameters estimation of spatial-temporal channel in smart antennas and its techniques of interference suppression. (1) For circular array in TD-SCDMA system, it provided an architecture of expanded spatial-temporal array, and realized the joint 4D parameters estimation of elevation, azimuth, delay and doppler frequency-offset for multipaths in spatial-temporal channel. The method does not need training sequences or pre-estimation of channel impulse response, can work in multi-user environments. The multipaths number of each user can exceed the element number of array, and all parameters are paired automatically. During the procedure, from PN code domain, spatial domain, temporal domain and frequency domain, the method can simultaneously suppressed the MAI, ISI and Doppler frequency interference in the channel, which greatly improved the interference suppression ability of smart antennas. (2) In order to increase the searching speed in multi-dimension parameters estimation, Particle Swarm Optimization (PSO) algorithm was employed to realize multi-dimension searching. Considering the limit that the basic PSO is only fit to search the global optimum, the thesis transformed the original spectrum function into an equal-peaks function, and, through some measures, such as modification to the particle speed updating formula and "congestion handling", it realized the application of PSO in the peaks searching for spatial-temporal spectrum.At last, the thesis analyzed the SNR, system capacity of mobile communication systems employing smart antennas, and discussed the possible effects on original system caused by smart antennas.
|
PDF Full Text Request