返回
类型 应用研究 预答辩日期 2017-11-15
开始(开题)日期 2013-05-04 论文结束日期 2017-09-25
地点 中心楼3楼会议室315 论文选题来源 国家自然科学基金项目     论文字数 7.5 (万字)
题目 非线性滤波方法在AUV导航系统中的应用研究
主题词 Gauss-Hermite求积分卡尔曼滤波,稀疏网格高斯滤波器,初始对准,组合导航,粗对准
摘要 本文以非线性滤波方法在AUV导航系统中的应用研究为主题,主要围绕SINS的惯性系粗对准方法、AUV导航系统的非线性间接法初始对准和非线性直接法信息滤波方法开展相关研究工作。主要研究工作和成果如下: (1) 研究了不考虑经度以及考虑经度的两种惯性系粗对准方法。由于地球自转角速率很小,在AUV晃动于扰或者运动环境下,传统粗对准方法的误差很大,甚至不可用。对线加速度干扰对惯性系粗对准精度的影响进行了理论分析,并通过多种运动状态下的数学仿真验证了理论分析的正确性,其中考虑经度的惯性系粗对准表现较优,不仅能在AUV存在大幅度的角运动时实现对准,而且也能在AUV匀速运动时完成粗对准,适应性强。 (2) 提出了基于Gauss-Hermite求积分卡尔曼滤波(GHQKF)的SINS非线性间接法初始对准方法。当AUV受海浪影响时,特别是海况不佳时,粗对准精度一般达不到线性滤波器适用的条件,考虑采用非线性滤波方法来实现SINS动基座初始对准。详细阐述了Gauss点和Gauss系数的单变量配置及多变量配置方法,给出了GHQKF滤波算法,针对SINS大方位失准角非线性误差模型,数学仿真验证了该算法的有效性。基于GHQKF滤波算法的对准性能优于基于UKF的非线性初始对准方法,而且该算法的对准精度随着单变量Gauss点数的增加可以进一步提高,但运算量也随着增加,需综合考虑对准性能和计算量。 (3) 针对基于GHQKF的初始对准方法的运算量较大的问题,提出了基于稀疏网格高斯滤波器(SGQKF)的SINS非线性间接法初始对准方法。该算法在不降低对准性能的基础上,可以减少运算量,提高实用性。详细阐述了采用稀疏网格和Gauss-Hermite数值积分理论配置多维积分点的方法。半物理仿真实验验证了该算法的优越性。该算法在失准角较小时,对准性能优于Kalman;在失准角较大时,该算法和基于GHQKF的对准方法的性能相当,但运算量小很多。综合考虑对准精度、收敛速度、运算量和适用条件,可采用3阶积分精度水平的SGQKF进行SINS初始对准。 (4) 提出了基于载体系速度匹配和欧拉角微分方程的AUV导航系统的非线性直接法信息滤波方法。和基于误差模型的间接法不同,直接法能将SINS导航解算和滤波结合在一起,更好的跟踪非线性。AUV导航系统中一般都会装备DVL,提供载体系速度,来辅助SINS导航。量测量直接采用载体系速度而不是传统上采用的导航系速度误差,解决了间接法中存在的噪声交叉问题。系统方程是基于欧拉角微分方程,而不是四元数微分方程,避免了四元数归一化的非线性约束问题。本文详细介绍了基于载体系速度匹配和欧拉角微分方程的直接法的建模过程,针对状态方程和量测方程均为非线性的模型特点,采用UKF滤波算法实现状态估计,最后通过装备有里程计与GPS的跑车实验数据验证了该算法的有效性,直接法比间接法的定位精度高,为以SINS为主的AUV导航系统信息滤波方法提供了一种新的解决方案。 (5) 研究了直接法中欧拉角的奇异问题,提出了基于双欧拉角的AUV无奇异的非线性直接法初始对准方法。该算法用于解决基于单欧拉角的直接法信息滤波方法存在的奇异问题。分析了正反欧拉角微分方程的奇异性正好互补的特性,研究了基于双欧拉角的SINS非线性直接法初始对准方法,给出了双欧拉角直接法的对准流程,并进行了半物理仿真验证。双欧拉角SINS非线性直接初始对准方法中,采用正、反欧拉角分区接替运算进行滤波,可以实现以欧拉角为状态量的直接法无奇异初始对准,双欧拉角直接法成功解决了单欧拉角直接法的奇异性问题,可以适用于AUV大机动的情况。 (6) 为了进一步的将惯性系粗对准的思想用于SINS的精对准阶段,分别研究了基于惯性系的间接法和直接法精对准方法,受地理坐标系中直接法的启发,对模型进行改造,建立了惯性系的间接法滤波模型,提出了基于惯性系的直接法精对准方法。首先基于不同的量测量建立了两种间接法惯性系误差模型,给出了它们的精对准流程图,半物理仿真对比验证了惯性系间接法精对准方法。然后建立了惯性系直接法初始对准方法的非线性滤波模型,并采用UKF实现状态估计,数学仿真验证了该算法的有效性,提高了航向收敛速度,缩短了收敛时间。
英文题目 NONLINEAR FILTERING APPLICATION FOR AUV NAVIGATION SYSTEM
英文主题词 Gauss-Hermite quadrature Kalman filter(GHQKF),sparse grid quadrature Kalman filter(SGQKF),initial alignment, integrated navigation,coarse alignment
英文摘要 The dissertation focuses on the application of nonlinear filtering method in AUV navigation systems. The coarse alignment methods in the inertial frame of SINS, the indirect nonlinear initial alignment approach and the direct nonlinear filtering approach for AUV navigation systems are studied. The main research work and results are as follows: (1) With or without the consideration of longitude, two coarse alignment methods in the inertial frame of SINS are investigated. Since the earth’s angular velocity is very small, the error of traditional coarse alignment method is very large and not even applicable when the AUV is swinging in disturbing or moving environment. We theoretically analyze the influence of the linear acceleration on the alignment accuracy of the coarse alignment methods in the inertial frame. The coarse alignment methods in the inertial frame are compared with the traditional analytical coarse alignment by means of the mathematical simulation in the condition of various motions. The correctness of the theoretical analysis is verified through the mathematical simulation of various motions, and the coarse alignment method in which the longitude is considered can achieve better performance, and it can adapt to a variety of environments. The coarse alignment performs well not only when the AUV has a large angular movement, but also when the AUV has the constant linear velocity. (2) The SINS indirect nonlinear initial alignment method using GHQKF is presented. When AUV is influenced by the waves, especially in the poor sea conditions, the coarse alignment accuracy is too low to use the linear filters, and the nonlinear filters is considered to achieve the SINS initial alignment on the moving base. The configuration method for one dimensional Gauss points and their corresponding coefficients in Gauss-Hermite quadrature is derived, which is then extended to form the multidimensional Gauss points and their coefficients using the direct tensor method. A GHQKF is presented, and the SINS nonlinear error model with large azimuth misalignment angle is used to verify the new algorithm. The new algorithm based on the GHQKF is better than the non-linear initial alignment based on the UKF. With the increase of the number of one dimensional Gauss-Hermite quadrature points, alignment accuracy can be further improved. However, the calculation load also increases a lot, so we need to take into account the performance and the calculation load. (3) In order to solve the problem that the initial alignment method based on the GHQKF has big calculation load, the SINS nonlinear initial alignment using the SGQKF is presented. The algorithm based on SGQKF can keep the alignment performance, and can also reduce the calculation load and improve the practicability. The theory of sparse-grid and Gauss-Hermite quadrature is adopted to generate multidimensional quadrature points and weights. The semi-physical simulation experiment verifies the superiority of the new algorithm. When the initial azimuth errors are small, the new algorithm can achieve better alignment performance than Kalman. If the initial azimuth errors become larger, the new algorithm shows similar alignment performance with GHQKF. The calculation of the new algorithm is much smaller. Considering the alignment accuracy, convergence speed, computation and applicable conditions, the SGQKF at the level-3 quadrature accuracy can be used in SINS initial alignment. (4) A direct nonlinear approach for the information filtering is presented, using the measured velocity in the body frame as the filter measurement and Euler angle kinematics as the system model for AUV navigation systems. Comparing with the indirect approach based on the error models, the direct approach can combine the SINS navigation calculation and the filtering, and it also can better track the system nonlinearity. The AUV navigation system is usually equipped with DVL, providing the velocity in the body frame to assist the SINS. The velocity in the body frame is used as the filter measurements without the velocity projection from the body frame to the local-level frame as in traditional approaches, so the cross-noise problem is avoided. The proposed approach is based on Euler angle kinematics equations instead of the quaternion kinematics equations, so the constraint in the quaternion no longer exists. The model, using the measured velocity in the body frame as the filter measurement and Euler angle kinematics as the system model, is established in detail. The direct approach uses an UKF because the system model and measurement model are nonlinear. The performance of the proposed approach is verified by road test data from a land vehicle equipped with an odometer and GPS. The proposed approach can achieve higher navigation accuracy than the commonly-used indirect approach. As an alternative, it can be considered as a supplement to the conventional integration navigation algorithm. (5) The problem how to avoid the singularities when using Euler angles in the direct approach is discussed, and a non-singular and nonlinear approach for the AUV initial alignment based on the dual-Euler method is presented. The new algorithm is established to solve the singularity problem which exists in the direct approach using the single Euler angle set. The characteristics that the singularity is complementary of the positive and passive Euler angle differential equations are analyzed. The direct approach using dual-Euler method is applied for the nonlinear SINS initial alignment, the alignment flow is presented, and the semi-physical simulation is carried out. The singularity problem, which exists in the Euler angles, can be efficiently solved by means of the dual-Euler method. The proposed approach is non-singular because the positive and passive Euler angle direct approaches operate alternately according to the switch flag to remove the singularities, and the non-singular direct approach can be applied to the case of large maneuver for the AUVs. (6) The indirect and direct approaches for the SINS initial alignment in the inertial frame are investigated, which further extends the idea of the coarse alignment in the inertial frame to the fine alignment. The indirect error model and the direct model in the geographic frame are modified, and inspired by the direct approach in the geographic frame the direct SINS initial alignment in the inertial frame is presented. The indirect error models are established by means of two different measurements, and their corresponding fine alignment flow charts are presented. The semi-physical simulation test is conducted to verify the proposed indirect approach. The steps to establish the nonlinear direct filter model in the inertial frame are presented in detail. The UKF is applied to estimate the states. The validity of the proposed direct approach is verified by mathematical simulation, the heading convergence speed is improved, and its convergence time is also shortened.
学术讨论
主办单位时间地点报告人报告主题
东南大学 2011.3.29 中心楼525会议室 冉昌艳 光纤气体传感器
东南大学 2012.3.14 中心楼525会议室 冉昌艳 基于模型辅助的AUV组合导航技术研究
东南大学 2014.9.10 中心楼525会议室 冉昌艳 四元数二次映射
东南大学 2014.5.15 中心楼525会议室 朱倚娴博士 基于高斯过程回归的联邦滤波故障检测方法
东南大学 2011.11.10 中心楼525会议室 冉昌艳 容积卡尔曼滤波
东南大学 2016.3.23 中心楼525会议室 周玲博士 一种基于SINS/TAN紧组合的水下潜器导航定位方法
东南大学 2014.3.26 中心楼525会议室 王磊博士 一种交互式多模型联邦滤波算法
东南大学 2017.3.13 中心楼525会议室 胡杰博士 单轴旋转捷联惯导系统精对准方法
     
学术会议
会议名称时间地点本人报告本人报告题目
2013 Instrumentation and Measurement Technology Conference. IEEE 2013.5.6-9 Minneapolis, MN, USA Sparse-grid Quadrature Kalman Filter based on the Kronrod-Patterson rule
2014年第三届自动控制和机电工程国际会议 2014.6.13-14 中国厦门 SINS Alignment Using Velocity Matching and Simplified Cubature Kalman Filter
     
代表作
论文名称
Sparse-grid Quadrature Kalman Filter based on the Kronrod-Patterson rule
稀疏网格高斯滤波器在SINS初始对准中的应用
基于Gauss-Hermite求积分卡尔曼滤波的SINS非线性初始对准方法
SINS Alignment Using Velocity Matching and Simplified Cubature Kalman Filter
A Direct and Non-Singular UKF Approach Using Euler Angle Kinematics for Integrated Navigation System
 
答辩委员会组成信息
姓名职称导师类别工作单位是否主席备注
何秀凤 正高 教授 博导 河海大学
朱欣华 正高 教授 硕导 南京理工大学
李宏生 正高 教授 博导 东南大学
陈熙源 正高 教授 博导 东南大学
刘锡祥 正高 教授 博导 东南大学
      
答辩秘书信息
姓名职称工作单位备注
汤新华 其他 讲师 东南大学