自适应容积卡尔曼滤波器及其在雷达目标跟踪中的应用

发布时间：2018-04-26 17:33

本文选题：卡尔曼滤波 + 非线性滤波　；参考：《大连海事大学》2015年博士论文

【摘要】：容积卡尔曼滤波算法是近年提出的一种获得广泛关注的非线性滤波算法。为满足近似线性最小方差条件,CKF仍须假设过程与量测噪声均为已知独立的零均值高斯白噪声。然而实际噪声环境往往不满足上述严格假设,这可能导致标准CKF的性能退化甚至发散。为克服这一局限性,进而增强标准CKF算法的鲁棒性。本研究基于标准CKF算法,针对不同噪声条件,提出几种改进自适应CKF算法并应用于雷达目标跟踪仿真实验。主要研究工作如下：1.为克服统计特征未知的过程噪声在滤波过程中对SCKF算法性能的不利影响,提出基于Sage-Husa噪声估计器的自适应SCKF算法。机动目标跟踪的仿真实验结果表明该算法在未知恒定和时变过程噪声两种背景中都比同等条件下的SCKF算法具有更好的滤波精度和稳定性。2.有色量测噪声不符合CKF算法的零均值高斯白噪声假设,因此基于一阶马尔科夫有色噪声模型、高斯滤波器和三度容积规则提出同时适用于有色和白色量测噪声条件的自适应CKF算法及其平方根版本。随后的目标跟踪仿真实验结果表明两种改进算法的有效性。3.针对高度容积卡尔曼滤波算法运行过程中存在的不定噪声对滤波性能造成的影响,基于非线性H无穷滤波框架和五度容积规则提出高度容积H无穷滤波算法,仿真实验结果证明了该改进算法的有效性。4.基于两种解相关原则和五度容积规则,提出两种互相关噪声背景下的改进高度容积滤波器,用于解决同时互相关噪声背景下高度容积卡尔曼滤波算法的性能退化的问题。目标跟踪仿真实验结果表明两种改进滤波算法能有效克服高度容积卡尔曼滤波算法的上述局限性。5.通过稳定性分析和仿真实验比较分析基于极大后验噪声估计的自适应CKF、强跟踪CKF、以及容积H无穷滤波算法这三种自适应CKF算法的性能。稳定分析结果表明三者之中,容积H无穷滤波算法的稳定性主要取决于标量β；而另两个滤波器的稳定性都受到过程噪声初值的影响。仿真实验结果表明三种滤波器在过程噪声未知恒定和时变的条件下均能有效改进CKF算法,但性能各有优劣。
[Abstract]:Volumetric Kalman filter (VKF) is a nonlinear filtering algorithm which has received wide attention in recent years. In order to satisfy the approximate linear minimum variance condition (CKF), it is necessary to assume that both the process and the measurement noise are known to be independent of Gao Si white noise with zero mean. However, the actual noise environment often does not satisfy these strict assumptions, which may lead to the degradation or even divergence of the performance of the standard CKF. In order to overcome this limitation, the robustness of the standard CKF algorithm is enhanced. Based on the standard CKF algorithm, this paper proposes several improved adaptive CKF algorithms for different noise conditions and applies them to radar target tracking simulation experiments. The main research work is as follows: 1. In order to overcome the adverse effect of process noise with unknown statistical characteristics on the performance of SCKF algorithm in the filtering process, an adaptive SCKF algorithm based on Sage-Husa noise estimator is proposed. The simulation results of maneuvering target tracking show that the proposed algorithm has better filtering accuracy and stability than the SCKF algorithm under the same conditions in both unknown constant and time-varying process noise background. The colored measurement noise does not accord with the null mean Gao Si white noise assumption of the CKF algorithm, so it is based on the first-order Markov colored noise model. An adaptive CKF algorithm and its square root version for both colored and white measurement noise conditions are proposed by Gao Si filter and cubic volume rule. The simulation results of target tracking show the effectiveness of the two improved algorithms. Aiming at the influence of the uncertain noise in the operation of the high volume Kalman filter algorithm, a high volume H infinite filter algorithm is proposed based on the nonlinear H-infinity filter framework and the five-degree volumetric rule. Simulation results show the effectiveness of the improved algorithm. 4. 4. Based on two kinds of decorrelation principles and five degree volume rule, an improved height volume filter with cross-correlation noise is proposed to solve the degradation of the performance of the high volume Kalman filter algorithm in the background of simultaneous cross-correlation noise. The simulation results of target tracking show that the two improved filtering algorithms can effectively overcome the limitations of the high volume Kalman filtering algorithm. The performance of three adaptive CKF algorithms based on maximum posteriori noise estimation, strongly tracked CKF, and volume H-infinity filtering algorithm are compared and analyzed by stability analysis and simulation experiments. The results of stability analysis show that the stability of the volumetric H-infinity filter is mainly dependent on the scalar 尾, while the stability of the other two filters is affected by the initial value of the process noise. The simulation results show that the three filters can effectively improve the CKF algorithm under the condition that the process noise is unknown constant and time-varying, but the performance has its own advantages and disadvantages.
【学位授予单位】：大连海事大学
【学位级别】：博士
【学位授予年份】：2015
【分类号】：TN713;TN953

【参考文献】