运载器组合导航高性能滤波算法研究

发布时间：2018-06-20 14:34

本文选题：运载器组合导航 + 卡尔曼滤波　；参考：《西北工业大学》2014年博士论文

【摘要】：随着现代科学技术的快速发展,人们对运载器导航解算的实时性和快速性的要求越来越高。常用的运载器导航解算主要采用卡尔曼滤波方法,应用卡尔曼滤波进行导航解算时,要求动力学系统的数学模型必须为线性。当系统模型为非线性时,若采用卡尔曼滤波算法进行导航解算,将会引起较大的误差,甚至导致滤波发散。为了提高运载器组合导航的解算精度,研究适用于运载器组合导航的高精度、非线性滤波算法,是交通信息工程与控制领域一项重要而又需要迫切研究的任务。本文在认真研究现有导航滤波算法的基础上,提出了一套适合运载器组合导航的高性能滤波算法,包括非线性模型预测Unscented粒子滤波算法、衰减记忆平方根Unscented粒子滤波算法、模糊抗差自适应Unscented粒子滤波算法、基于状态相关系数的抗差自适应滤波算法、非线性抗差自适应状态相关黎卡提方程滤波算法、以及动力学模型误差的Sage随机加权自适应滤波算法。将提出的算法应用到运载器组合导航系统中进行仿真验证,并与现有的滤波算法进行比较,结果表明,提出的算法不但计算量小,而且滤波精度高,滤波性能明显优于现有的滤波算法。论文主要研究内容和创新性贡献如下 (1)提出一种新的非线性模型预测Unscented粒子滤波算法。该算法在建立系统模型时顾及了模型误差,利用估计出的模型误差对含有误差的非线性、非高斯系统模型进行修正,再利用Unscented粒子滤波进行解算。仿真结果表明,提出算法的滤波性能明显优于模型预测滤波和Unscented粒子滤波,提高了导航解算精度。 (2)在研究Unscented粒子滤波的基础上,吸收了衰减记忆滤波和平方根滤波的优点,提出一种新的衰减记忆平方根Unscented粒子滤波算法。在该算法中,通过衰减因子调节当前量测信息对估计值的影响,减小历史信息对滤波的作用。然后用协方差矩阵的平方根阵代替协方差矩阵进行迭代计算,保证了协方差矩阵的对称性和正定性。研究结果表明,提出的算法能有效改善滤波性能,提高了导航系统的解算精度。 (3)在研究模糊控制理论的基础上,吸收了Unscented粒子滤波、自适应滤波和抗差估计的优点,提出一种新的模糊抗差自适应Unscented粒子滤波算法。该算法顾及了量测量中的粗差对滤波的影响,基于模糊理论构造等价权函数,利用等价权函数和模糊抗差自适应因子调节粗差对导航解的影响,有效地控制粗差对导航解的影响。将提出的算法应用到组合导航系统中进行仿真验证,结果表明,提出算法不但实时性好,而且滤波精度明显提高。 (4)提出一种新的基于状态相关系数的抗差自适应滤波算法。采用状态相关系数将非线性系统转换为状态相关系统,在处理非线性动力学模型与量测模型时不必进行线性化,从而减小了由线性化系统模型所带来的误差。建立抗差自适应滤波模型,利用等价权矩阵和自适应因子进行信息分配,从而控制动力学模型异常和观测异常对导航解的影响。仿真结果表明,提出的算法不仅能够有效地抑制动态系统模型状态噪声和观测噪声干扰,而且计算简单,滤波精度明显优于EKF和UKF算法。 (5)提出一种非线性抗差自适应状态相关黎卡提方程滤波算法。该方法采用状态相关系数法将非线性系统转换成类似线性系统结构,减小了由线性化系统模型所带来的误差。在一定的条件下证明了该算法的稳定性。仿真结果表明,提出的算法不仅能够有效地抑制非线性系统模型状态噪声和观测噪声的干扰,而且滤波精度明显优于UKF和SDRE滤波算法。 (6)现有文献研究中,对新息向量和观测残差向量的协方差阵采用算术平均值估计,其估计的观测噪声向量协方差阵中含有状态预测值的误差,若状态预测值的误差较大,预测残差必然大,从而由预测残差计算的新息向量和观测残差向量的协方差阵的估计精度就变差。为了克服这一缺陷,本文提出用一种新的随机加权估计算法,对观测噪声协方差阵和状态噪声协方差阵进行估计,以控制观测异常和动态模型噪声异常对状态参数估值的影响。仿真结果表明,提出的算法不仅计算简单,而且能提高动态导航解算得滤波精度。 (7)提出动力学模型误差的Sage随机加权自适应估计方法。该方法利用Sage滤波的开窗平滑方法,求取观测残差向量和预测残差向量的协方差阵,用随机加权因子对观测残差和预测残差进行调节,以控制观测残差和预测残差对导航解算精度的影响。仿真结果证明,提出的算法对状态扰动带来的误差具有较强的抑制能力。本文所取得的研究成果对运载器组合导航滤波解算、多源信息融合、误差估计和计算机仿真等领域的研究都有一定贡献。研究结果不但可以应用于军用和民用领域运载器导航定位的滤波解算,而且经过推广,还可以用于航空航天领域其它飞行器导航定位的滤波解算。
[Abstract]:With the rapid development of modern science and technology, the demand for the real-time and rapidity of the navigation solution of the carrier is getting higher and higher. The Calman filtering method is mainly used in the navigation solution of the carrier. When the navigation is solved by the Calman filter, the mathematical model of the dynamic system must be linear. When the system model is non line In sex, if the Calman filter algorithm is used to calculate navigation, it will cause large error and even lead to filtering divergence. In order to improve the accuracy of the integrated navigation of the carrier, it is an important and urgent research to study the high precision and nonlinear filtering algorithm which is suitable for the integrated navigation of the carrier. The task of studying.
On the basis of studying the existing navigation filtering algorithm, a set of high performance filtering algorithm suitable for carrier integrated navigation is proposed, including nonlinear model prediction Unscented particle filter algorithm, attenuated memory square root Unscented particle filter algorithm, fuzzy adaptive adaptive Unscented particle filter algorithm and state correlation system. The robust adaptive filtering algorithm, the nonlinear adaptive state dependent Riccati equation filtering algorithm and the Sage random weighted adaptive filtering algorithm for the dynamic model error are applied to the carrier integrated navigation system to be simulated and verified, and compared with the existing filtering algorithms. The results show that the algorithm is proposed. The algorithm not only has a small amount of computation, but also has high filtering accuracy, and the filtering performance is much better than the existing filtering algorithm.
The main research content and innovative contribution of this paper are as follows
(1) a new nonlinear model prediction Unscented particle filter algorithm is proposed. The algorithm takes into account the model error when establishing the system model, uses the estimated model error to modify the nonlinear model containing errors, and then uses the Unscented particle filter to solve the model. The simulation results show that the algorithm is filtered. Wave performance is better than model predictive filtering and Unscented particle filtering, which improves navigation accuracy.
(2) on the basis of studying Unscented particle filter and absorbing the advantages of attenuated memory filter and square root filter, a new attenuation memory square root Unscented particle filter algorithm is proposed. In this algorithm, the effect of the historical information on the filter is reduced by the attenuation factor and the effect of historical information on the filtering is reduced. The square root matrix of variance matrix is used for iterative calculation instead of covariance matrix, which ensures the symmetry and positive stability of the covariance matrix. The results show that the proposed algorithm can effectively improve the filtering performance and improve the accuracy of the navigation system.
(3) on the basis of the study of fuzzy control theory and absorbing the advantages of Unscented particle filter, adaptive filtering and robust estimation, a new fuzzy adaptive Unscented particle filtering algorithm is proposed. The algorithm takes into account the influence of the gross error in the measurement on the filtering, constructs the equivalent weight function based on the fuzzy theory, and uses the equivalent weight function. The effect of gross error on navigation solution is regulated by number and fuzzy tolerance adaptive factor, and the effect of gross error on navigation solution is effectively controlled. The proposed algorithm is applied to the integrated navigation system for simulation verification. The results show that the proposed algorithm not only has good real-time performance, but also improves the filtering accuracy.
(4) a new adaptive filtering algorithm based on state correlation coefficient is proposed. Using the state correlation coefficient, the nonlinear system is converted into a state related system. The linearization is not necessary when dealing with the nonlinear dynamic model and the measurement model, thus reducing the error caused by the linearized system model. The simulation results show that the proposed algorithm can not only effectively suppress the state noise and the observation noise interference of the dynamic system model, but also the calculation is simple, and the filtering precision is obviously superior to the EK. F and UKF algorithms.
(5) a nonlinear differential adaptive state correlation Riccati equation filtering algorithm is proposed. The method uses the state correlation coefficient method to convert the nonlinear system into a similar linear system structure, and reduces the error caused by the linearized system model. The stability of the algorithm is proved under certain conditions. The simulation results show that the proposed method is proposed. The algorithm not only effectively suppresses the interference of the state noise and the observation noise of the nonlinear system model, but also has better filtering accuracy than the UKF and SDRE filtering algorithms.
(6) in the existing literature study, the covariance matrix of the new interest vector and the observed residual vector is estimated by arithmetic mean, and the estimated error of the estimated value of the state of the observed noise vector covariance is that if the error of the state prediction is larger and the predicted residual is inevitable, the new interest vector and the observation residual vector are calculated from the predicted residual. In order to overcome this defect, a new random weighting estimation algorithm is proposed to estimate the observation noise covariance matrix and the state noise covariance matrix to control the effect of abnormal observation and dynamic model noise on the state parameter estimation. The simulation results show that the proposed algorithm is proposed. It is not only simple in calculation but also improves the accuracy of filtering in dynamic navigation.
(7) the Sage random weighting adaptive estimation method for dynamic model error is proposed. This method uses the window smoothing method of Sage filtering to obtain the covariance matrix of the observation residual vector and the predicted residual vector, and adjusts the observation residual and the prediction residual by the random weighting factor, so as to control the observation residual and the prediction residual to the navigation solution precision. Simulation results show that the proposed algorithm has a strong ability to suppress the errors caused by state disturbance.
The research results obtained in this paper have some contribution to the research of carrier integrated navigation filtering, multi source information fusion, error estimation and computer simulation. The research results can be applied not only to the filtering and calculation of navigation and positioning of military and civil carrier, but also in the field of aviation and aerospace. The filtering of the navigation and positioning of the aircraft is calculated.
【学位授予单位】：西北工业大学
【学位级别】：博士
【学位授予年份】：2014
【分类号】：TN96.2

【参考文献】