CN107421543B - Implicit function measurement model filtering method based on state dimension expansion - Google Patents
Implicit function measurement model filtering method based on state dimension expansion Download PDFInfo
- Publication number
- CN107421543B CN107421543B CN201710478805.0A CN201710478805A CN107421543B CN 107421543 B CN107421543 B CN 107421543B CN 201710478805 A CN201710478805 A CN 201710478805A CN 107421543 B CN107421543 B CN 107421543B
- Authority
- CN
- China
- Prior art keywords
- state
- measurement
- state quantity
- covariance matrix
- filtering
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Images
Classifications
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01C—MEASURING DISTANCES, LEVELS OR BEARINGS; SURVEYING; NAVIGATION; GYROSCOPIC INSTRUMENTS; PHOTOGRAMMETRY OR VIDEOGRAMMETRY
- G01C21/00—Navigation; Navigational instruments not provided for in groups G01C1/00 - G01C19/00
- G01C21/20—Instruments for performing navigational calculations
Abstract
The invention relates to a hidden function measurement model filtering method based on state dimension expansion, namely IAUKF. In this method, the quantity measurement is extended into the state quantity, while the zero vector is treated as an equivalent quantity measurement for filter updating. The IAUKF achieves better estimation performance than the IAEKF and IEKF. Especially when the measurement noise increases, the performance can be greatly improved compared with the implicit UKF.
Description
Technical Field
The invention belongs to the field of autonomous navigation of spacecrafts, and relates to a filtering method of an implicit function measurement model of state dimension expansion.
Background
Kalman in 1960 proposed a linear optimal recursive filtering method, Kalman Filter (KF). Initially, KF is only applicable to linear systems, and as the demand for filtering nonlinear systems expands, filtering methods such as Extended Kalman Filter (EKF), Unscented Kalman Filter (uvf), Particle Filter (PF), etc. are gradually proposed and continuously developed. The measurement models in the classical kalman filtering algorithm all have explicit expressions, however, in many practical problems, constraints of state quantity and measurement quantity are often implicit, and explicit measurement models are not easy or cannot be obtained, which is an implicit measurement model filtering problem.
There are two main types of methods for solving the state estimation problem including the implicit measurement model at home and abroad. The first type is the IEKF proposed by sotto et al, obtained by linearizing an Implicit measurement equation at a reference point and taking a second order form, in combination with the conventional EKF algorithm with an explicit measurement equation, through an Implicit Extended Kalman Filter (IEKF). The second type is a filtering method of an implicit measurement model containing Iterative measurement updating, namely Iterative IEKF, which is provided by Steffen on the basis of analyzing the IEKF method. Both the methods are established on the basis of EKF, a Jacobian matrix needs to be calculated during application, linearization errors can reduce the precision of a filtering algorithm, divergence of filtering results can be caused, and the calculation of the Jacobian matrix is usually complex.
Disclosure of Invention
The technical problem to be solved by the invention is as follows: the method overcomes the defects of the prior art, provides a hidden function measurement model filtering method based on state dimension expansion, and obtains better estimation performance compared with IAEKF and IEKF. Especially when the measurement noise increases, the estimation performance can be greatly improved compared with the implicit UKF.
The invention provides a hidden function measurement model filtering method based on state dimension expansion, namely IAUKF. In the method, the quantity measurement is extended into the state quantity, and meanwhile, the zero vector is regarded as the equivalent quantity measurement to perform filtering updating, and the method specifically comprises the following steps:
first, the true quantity of k time is measuredExtending to state quantity XkIn the structure of the expanded state quantityAndestablishing a system model meeting the dimension expansion state quantity;
and secondly, initializing, solving the state quantity of the initial moment according to the system model, and substituting the state quantity of the initial moment and the covariance matrix thereof into the expanded state quantity constructed in the first stepAndin each case are denoted byAnd
thirdly, time updating, namely calculating the predicted state according to the expanded state quantity of the initial time and the covariance matrix obtained in the second stepAnd error covariance matrix thereofAt the rest of the time, in the state of calculating the predictionAnd error covariance matrix thereofThen, the state quantity obtained in the fourth step is updated according to the measurementAnd corresponding error covariance matrixTo obtain;
fourthly, measuring and updating, and solving the updated state estimation according to the state prediction value obtained by time updating and solvingAnd corresponding error covariance matrixAnd then returning to the third step, and realizing loop solution until the filtering is finished.
The first step is to construct an expanded state quantity, and establish a system model meeting the dimension expansion state quantity as follows:
wherein, XkAnd ZkRespectively the state quantity and the quantity measurement at the time k,the state quantity after the expansion is represented,indicating measurement of Z from actual quantitykAnd the measurement noise vkMeasuring the true quantity obtained by mixing;
in the formula, Fa(. a) anda state model representing the extended state quantities and their state errors, respectively, h (-) is a non-linear explicit function.
In the second step, the state quantity after the dimension expansionAnd covariance matrix thereofThe initialization is as follows:
wherein the content of the first and second substances,E[x]denotes the expected value of x, Z1And N1And n and m respectively represent the dimensionality of the state vector and the measurement vector.
In the third step, the time is updated as follows:
in the formula (I), the compound is shown in the specification,is satisfied with a mean value ofCovariance ofPoint of (a), wiIs the weight of the ith Sigma point,is composed ofThe covariance matrix of (a), wherein,
in the formula, naA dimension representing the state quantity after expansion, tau is a scaling parameter,represents the square root of the matrixThe ith dimensional column vector of (1).
In the fourth step, the measurement is updated as follows:
wherein the content of the first and second substances,for filtering the gain matrix, Ykk-1In order to measure the predicted value of the measurement, respectively corresponding error covariance matrices.
Compared with the prior art, the invention has the advantages that:
(1) according to the invention, the state quantity and the real quantity measurement are expanded into a new state quantity, and meanwhile, the zero vector is regarded as an equivalent quantity measurement to carry out filtering updating, so that better estimation performance is obtained compared with IAEKF and IEKF, and the estimation precision is improved; especially when the measurement noise increases, the performance can be greatly improved compared with the implicit UKF.
(2) The method omits the process of solving the Jacobian matrix in the existing method, and reduces the calculated amount.
Drawings
FIG. 1 is a flow chart of a filtering method of a hidden function measurement model based on state dimension expansion according to the present invention;
Detailed Description
Fig. 1 shows a flow chart of a filtering method of an implicit function measurement model based on state dimension expansion. The following detailed description of the embodiments of the present invention:
common nonlinear systems all contain explicit metrology models, and such systems can be described as:
in the formula, the state equation f (-) and the measurement equation h (-) are both non-linear explicit functions. XkAnd wkRespectively representing the state vector at time k and its noise, Xk+1Representing the state vector at time k + 1. ZkAnd vkRespectively representing the measurement vector at time k and its noise. In practical application, it can be considered that both the state vector and the measurement vector are affected by zero-mean and uncorrelated white gaussian noise, that is, the state quantity noise and the measurement noise respectively obey:
wherein Q iskAnd NkRespectively representing the covariance matrix corresponding to the state noise and the measurement noise, and the specific value is determined by engineering experience or system parameters.
However, in many practical problems, the state quantity and the actual quantity measurement are constrained in a measurement model in an implicit function form, and an explicit measurement equation is not easy or available, and such problems are filtering problems of the implicit measurement model. Such implicit metrology models can be described as a system as follows:
wherein the state quantity XkSum quantity measurement ZkThe constituent implicit function h (·) ═ 0 constraint.
First, to deal with the nonlinear system problem with implicit metrology models, the state quantities and true quantity measurements can be expanded into a new state quantity:
in the formula, the superscript a marks the state extension,indicating measurement of Z from actual quantitykAnd the measurement noise vkThe actual amount of measurements made by the mixing is measured,indicating the expanded state quantities.
Second, note h (X) in System (3)k,Zk+vk) Zero vector equal to m dimensions, so the zero vector can be regarded as an equivalent quantity measurement YkNamely:
due to expanded state quantityIs derived from the previous state quantity XkAnd true quantity measurementThe structure is that the measurement model (5) can be rewritten as:
therefore, the system model for establishing the dimension expansion state quantity satisfaction is as follows:
in the formula, Fa(. a) andstate model representing the extended state quantities and its state errors, Fa(. a) andthe calculation formula of (a) is as follows:
The IAUKF method comprises the following concrete implementation steps:
1. initialization
Initial state estimation and corresponding error covariance matrixAnd P0Respectively setting as follows:
in the formula, E [ x ]]Expected value, state error covariance matrix Q representing xkCovariance matrix N with measurement errorskThe corresponding values at the initial time are selected respectively.
Initial extended state quantityShould be constructed according to equation (4), however, in practical applications, the true quantity is measuredNot available, the present invention measures Z with an actual quantity1Measurement in place of a true quantityAt the moment, a noise covariance matrix is introduced as N1Thereby, the state quantity after dimension expansionAnd covariance matrix thereofMay be respectively configured as:
in the formula, Z1And N1And n and m respectively represent the dimensionality of the state vector and the measurement vector.
2. Time updating
At time k, the estimation value of the expansion state quantity obtained at the last time is needed firstThe correction is performed according to equation (4). However, in practical applications, the true quantity is measuredNot available, we measured Z with the actual quantitykMeasurement in place of a true quantityAt this time, a measurement noise v is introducedkAnd its noise covariance matrix Nk. Therefore, the dimension expansion state quantityAnd covariance matrix thereofThe correction can be made as follows:
in the formula (I), the compound is shown in the specification,and Pk-1Are respectively shown to be included inAndthe unexpanded state quantity estimate in (1) and its error covariance matrix.
Likewise, the IAUKF method performs probability deduction based on UT transforms. Satisfy the mean value ofCovariance of2n ofa+1 Sigma points are equivalent toThe Sigma points are propagated through the system model (7) to obtain corresponding propagated Sigma points, and the propagated Sigma points can be used for calculating the predicted stateAnd error covariance matrix thereofThis particular set of Sigma is determined according to the following formula:
in the formula, naA dimension representing the state quantity after expansion, numerically equal to n + m, τ being a scaling parameter,represents the square root of the matrixOf the ith-dimensional column vector, wiIs the weight of the ith Sigma point.
The Sigma point-by-state model passes as:
the state prediction value and its error covariance matrix can be calculated as follows:
3. measurement update
According to equation (6), the Sigma point of the estimator measurement can be calculated according to the following equation:
due to the prediction of the Sigma points obtained according to equation (16)Value ofNot the true state at time k, contains errors. Therefore, equation (18) is also not equal to its true value of 0, which provides information about the predicted stateCan be used for state correction.
Thus, the predicted value of the measurement can be calculated as:
the corresponding error covariance matrix can be obtained by:
then, the gain matrix is filteredUpdated state estimationAnd corresponding error covariance matrixCan be obtained by calculation according to a UKF method respectively:
table 1 and Table 2 show the comparison of the navigation results of the IAUKF, the IEKF, the IAEKF, and the implicit UKF.
TABLE 1 navigation results of four filtering methods
TABLE 2 navigation results of four filtering methods under different measurement noises
Table 1 compares the filtering results of the four filtering methods when the equivalent measurement noise is relatively small and is 1'; table 2 compares the filtering results of the four filtering methods with different measured noise. It can be seen that: when the measurement noise is small, compared with the IAEKF and the IEKF, the estimation precision of the position of the IAUKF is greatly improved; when the measurement noise is gradually increased, the estimation performance of the IAUKF can be greatly improved compared with the implicit UKF.
Those skilled in the art will appreciate that the invention may be practiced without these specific details.
Claims (5)
1. A hidden function measurement model filtering method based on state dimension expansion is used for autonomous navigation of a spacecraft and is characterized in that a state model is constructed by taking the position and the speed of the spacecraft as state quantities, and the method comprises the following steps:
first, the true quantity of k time is measuredExtending to state quantity XkIn the structure of the expanded state quantityAndestablishing a system model meeting the dimension expansion state quantity;
and secondly, initializing, solving the state quantity of the initial moment according to the system model, and substituting the state quantity of the initial moment and the covariance matrix thereof into the expanded state quantity constructed in the first stepAndin each case are denoted byAnd
thirdly, time updating, namely calculating the predicted state according to the expanded state quantity of the initial time and the covariance matrix obtained in the second stepAnd error covariance matrix thereofAt the rest of the time, in the state of calculating the predictionAnd error covariance matrix thereofThen, the state quantity obtained in the fourth step is updated according to the measurementAnd corresponding error covariance matrixTo obtain;
2. The method of claim 1, wherein the filtering method comprises: the first step is to construct an expanded state quantity, and establish a system model meeting the dimension expansion state quantity as follows:
wherein, XkAnd ZkRespectively the state quantity and the quantity measurement at the time k,the state quantity after the expansion is represented,indicating measurement of Z from actual quantitykAnd the measurement noise vkMeasuring the true quantity obtained by mixing;
3. The method of claim 1, wherein the filtering method comprises: in the second step, the state quantity after the dimension expansionAnd covariance matrix thereofThe initialization is as follows:
4. The method of claim 1, wherein the filtering method comprises: in the third step, the time is updated as follows:
in the formula (I), the compound is shown in the specification,is satisfied with a mean value ofCovariance ofPoint of (a), wiIs the weight of the ith Sigma point,is composed ofThe covariance matrix of (a), wherein,
5. The method of claim 1, wherein the filtering method comprises: in the fourth step, the measurement is updated as follows:
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201710478805.0A CN107421543B (en) | 2017-06-22 | 2017-06-22 | Implicit function measurement model filtering method based on state dimension expansion |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201710478805.0A CN107421543B (en) | 2017-06-22 | 2017-06-22 | Implicit function measurement model filtering method based on state dimension expansion |
Publications (2)
Publication Number | Publication Date |
---|---|
CN107421543A CN107421543A (en) | 2017-12-01 |
CN107421543B true CN107421543B (en) | 2020-06-05 |
Family
ID=60426530
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201710478805.0A Active CN107421543B (en) | 2017-06-22 | 2017-06-22 | Implicit function measurement model filtering method based on state dimension expansion |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN107421543B (en) |
Families Citing this family (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109001699B (en) * | 2018-01-30 | 2021-07-16 | 哈尔滨工业大学 | Tracking method based on noisy destination information constraint |
CN111735458B (en) * | 2020-08-04 | 2020-11-24 | 西南石油大学 | Navigation and positioning method of petrochemical inspection robot based on GPS, 5G and vision |
Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102289557A (en) * | 2011-05-17 | 2011-12-21 | 杭州电子科技大学 | Battery model parameter and residual battery capacity joint asynchronous online estimation method |
CN103065037A (en) * | 2012-11-13 | 2013-04-24 | 杭州电子科技大学 | Nonlinear system target tracking method based on distributed volume information filtering |
CN103294931A (en) * | 2013-06-28 | 2013-09-11 | 上海交通大学 | System state estimation method based on improved nonlinear robust filtering algorithm |
CN104252178A (en) * | 2014-09-12 | 2014-12-31 | 西安电子科技大学 | Strong maneuver-based target tracking method |
CN106017509A (en) * | 2016-05-30 | 2016-10-12 | 北京航空航天大学 | Method for determining anti-interference attitude in multi-source interference environment, and test platform |
CN106352876A (en) * | 2016-07-25 | 2017-01-25 | 北京航空航天大学 | Airborne distributed POS transfer alignment method based on H infinity and CKF hybrid filtering |
-
2017
- 2017-06-22 CN CN201710478805.0A patent/CN107421543B/en active Active
Patent Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102289557A (en) * | 2011-05-17 | 2011-12-21 | 杭州电子科技大学 | Battery model parameter and residual battery capacity joint asynchronous online estimation method |
CN103065037A (en) * | 2012-11-13 | 2013-04-24 | 杭州电子科技大学 | Nonlinear system target tracking method based on distributed volume information filtering |
CN103294931A (en) * | 2013-06-28 | 2013-09-11 | 上海交通大学 | System state estimation method based on improved nonlinear robust filtering algorithm |
CN104252178A (en) * | 2014-09-12 | 2014-12-31 | 西安电子科技大学 | Strong maneuver-based target tracking method |
CN106017509A (en) * | 2016-05-30 | 2016-10-12 | 北京航空航天大学 | Method for determining anti-interference attitude in multi-source interference environment, and test platform |
CN106352876A (en) * | 2016-07-25 | 2017-01-25 | 北京航空航天大学 | Airborne distributed POS transfer alignment method based on H infinity and CKF hybrid filtering |
Non-Patent Citations (3)
Title |
---|
四元数扩维无迹卡尔曼滤波算法及其在大失准角快速传递对准中的应用;周卫东等;《控制理论与应用》;20111130;第28卷(第11期);1583-1588 * |
基于地磁定轨和扩维卡尔曼滤波的导航算法;赵敏华等;《西安交通大学学报》;20041231;第38卷(第12期);1315-1318 * |
基于滤波扩维融合的传感器定位测距方法改进;孙媛等;《计算机仿真》;20161231;第33卷(第12期);265-269 * |
Also Published As
Publication number | Publication date |
---|---|
CN107421543A (en) | 2017-12-01 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
WO2018014602A1 (en) | Volume kalman filtering method suitable for high-dimensional gnss/ins deep coupling | |
CN109782240B (en) | Multi-sensor system error registration method and system based on recursive correction | |
Cao et al. | Huber second-order variable structure predictive filter for satellites attitude estimation | |
CN108562290B (en) | Navigation data filtering method and device, computer equipment and storage medium | |
CN110275193B (en) | Cluster satellite collaborative navigation method based on factor graph | |
CN108413986B (en) | Gyroscope filtering method based on Sage-Husa Kalman filtering | |
CN111181529B (en) | Smooth constraint extended Kalman filtering method applied to nonlinear Gaussian model | |
CN107421543B (en) | Implicit function measurement model filtering method based on state dimension expansion | |
CN107565931A (en) | A kind of self calibration Unscented kalman filtering method | |
Narasimhappa et al. | An innovation based random weighting estimation mechanism for denoising fiber optic gyro drift signal | |
CN116595897B (en) | Nonlinear dynamic system state estimation method and device based on message passing | |
CN114139109A (en) | Target tracking method, system, equipment, medium and data processing terminal | |
CN106871905B (en) | Gaussian filtering substitution framework combined navigation method under non-ideal condition | |
CN116734864B (en) | Autonomous relative navigation method for spacecraft under constant observed deviation condition | |
CN110532517A (en) | Gas pipeline method for parameter estimation based on improved ARUKF | |
CN111310110B (en) | Hybrid state estimation method for high-dimensional coupling uncertain system | |
CN109655081B (en) | On-orbit adaptive correction method and system for star sensor optical system parameters | |
KR101502721B1 (en) | Method and apparatus for providing precise positioning information using adaptive interacting multiple model estimator | |
CN110912535A (en) | Novel pilot-free Kalman filtering method | |
CN109115228B (en) | Target positioning method based on weighted least square volume Kalman filtering | |
CN108050997B (en) | Fiber-optic gyroscope filtering method based on volume Kalman | |
Wang et al. | Iterated posterior linearization filters and smoothers with cross-correlated noises | |
CN108225373B (en) | Large misalignment angle alignment method based on improved 5-order cubature Kalman | |
US20100161284A1 (en) | Nonlinear Variable Lag Smoother | |
KR101860810B1 (en) | System and method for navigation with nonlinear measurement model |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant | ||
GR01 | Patent grant |