Optimal data fusion method suitable for ballistic missile INS/CNS/GNSS combined navigation system
阅读说明:本技术 一种适用于弹道导弹ins/cns/gnss组合导航系统的最优数据融合方法 (Optimal data fusion method suitable for ballistic missile INS/CNS/GNSS combined navigation system ) 是由 陈熙源 柳笛 刘晓 石春凤 马振 于 2019-09-24 设计创作,主要内容包括:本发明公开了一种适用于弹道导弹INS/CNS/GNSS组合导航系统的最优数据融合方法,该方法包括以下步骤:构造INS/CNS/GNSS组合导航系统模型;在广义高阶CKF的时间更新阶段和量测更新阶段分别引入自适应渐消因子和最大相关熵准则进行INS/CNS子系统和INS/GNSS子系统的局部状态估计;根据最小方差原理和容积准则融合INS/CNS子系统和INS/GNSS子系统的局部估计得到全局最优状态估计。本发明可以同时抑制过程建模误差和非高斯量测噪声对状态估计的影响,提高弹道导弹INS/CNS/GNSS组合导航的自适应性和鲁棒性,获得全局最优的状态估计值。(The invention discloses an optimal data fusion method suitable for a ballistic missile INS/CNS/GNSS combined navigation system, which comprises the following steps: constructing an INS/CNS/GNSS integrated navigation system model; respectively introducing a self-adaptive fading factor and a maximum correlation entropy criterion to carry out local state estimation of the INS/CNS subsystem and the INS/GNSS subsystem in a time updating stage and a measurement updating stage of the generalized high-order CKF; and according to the minimum variance principle and the volume criterion, local estimation of the INS/CNS subsystem and the INS/GNSS subsystem is fused to obtain global optimal state estimation. The method can simultaneously inhibit the influence of process modeling errors and non-Gaussian measurement noise on state estimation, improve the adaptivity and robustness of ballistic missile INS/CNS/GNSS combined navigation, and obtain the globally optimal state estimation value.)
1. An optimal data fusion method suitable for a ballistic missile INS/CNS/GNSS combined navigation system is characterized by comprising the following steps: the method comprises the following steps:
s1: constructing an INS/CNS/GNSS combined navigation system filtering model;
s2: respectively introducing a self-adaptive fading factor and a maximum correlation entropy criterion to carry out local state estimation of the INS/CNS subsystem and the INS/GNSS subsystem in a time updating stage and a measurement updating stage of the generalized high-order CKF;
s3: and according to the minimum variance principle and the volume criterion, local estimation of the INS/CNS subsystem and the INS/GNSS subsystem is fused to obtain global optimal state estimation.
2. The optimal data fusion method applicable to ballistic missile INS/CNS/GNSS combined navigation system according to claim 1, wherein: the constructing of the INS/CNS/GNSS integrated navigation system filtering model in step S1 includes the following steps:
s1-1: setting the state vector of the INS/CNS/GNSS integrated navigation system as x ═ phi [ [ phi ] ]x,φy,φz,δvx,δvy,δvz,x,y,z,εx,εy,εz,Δx,Δy,Δz]TWherein phix,φy,φzRepresenting the angular misalignment of the attitude, δ v, under the inertial system of the emission pointx,δvy,δvzRepresenting the velocity error in the inertial system of the emission point, x, y, z representing the position error in the inertial system of the emission point, epsilonx,εy,εzRepresenting the gyro constant drift, Delta, in a projectile coordinate systemx,Δy,ΔzRepresenting the constant bias of the accelerometer in a missile body coordinate system, wherein T is a transposition symbol;
s1-2: establishing a state equation of the system according to a 15-dimensional state vector x of the INS/CNS/GNSS combined navigation system:
xk=f(xk-1)+vk-1
wherein f (-) is a nonlinear system function, xk-1And xkState vectors, v, representing time k-1 and k, respectivelyk-1Representing process noise, vk-1Has a covariance of
S1-3: respectively establishing measurement equations of an INS/GNSS subsystem and an INS/CNS subsystem:
in an INS/GNSS subsystem, taking the difference between the positions and the speed output by the INS and the GNSS as measurement information, and establishing a measurement equation of the subsystem:
z1,k=H1,kxk+ω1,k
wherein z is1,kRepresentation of INS/GNSSMeasurement vector of system k time, H1,kMeasurement matrix, ω, representing INS/GNSS subsystem k time1,kRepresents the measurement noise at the moment k of the INS/GNSS subsystem, the variance of which is
In an INS/CNS subsystem, the difference value of the attitude angles output by the INS and the CNS is used as measurement information, and a measurement equation of the subsystem is established:
z2,k=H2,kxk+ω2,k
wherein z is2,kThe measurement vector, H, representing the INS/CNS subsystem at time k2,kMeasurement matrix, ω, representing INS/CNS subsystem k time2,kRepresents the measured noise at the moment k of the INS/CNS subsystem with a variance of
3. The optimal data fusion method applicable to the ballistic missile INS/CNS/GNSS combined navigation system according to claim 2, wherein the optimal data fusion method comprises the following steps: the step S2 specifically includes the following steps:
s2-1: since the INS/CNS subsystem and the INS/GNSS subsystem use the same filtering process for local state estimation, only the filtering process of the INS/GNSS subsystem will be specifically described herein to avoid repetition, and therefore, z in step S1-3 will be described1,k、z2,kIs uniformly written as zk,H1,kAnd H2,kWrite uniformly as Hk,ω1,kAnd ω2,kWrite uniformly to omegak,R1,kAnd R2,kIs written uniformly as RkSetting the value of the kernel width gamma according to the local state estimation requirement of the INS/GNSS subsystem, and setting the initial state vector, the state error covariance and the fading factor as
s2-2: according to the formula
s2-3: predicting state vectors at time k
S2-4: according toAnd Sk|k-1Generating new volume points
S2-5: measurement equations and from INS/GNSS subsystemsIn step S2-3
wherein, I represents an identity matrix,
in the equation
S2-6: updating the measurement noise variance:
s2-7: computing a cross-covariance matrix P between state information and metrology informationxz,k|k-1,
s2-8: order to
S2-9:
if it is notComputing
s2-10: using an adaptive fading factor taukUpdate predicted state prediction covariance:
4. The optimal data fusion method applicable to the ballistic missile INS/CNS/GNSS combined navigation system according to claim 3, wherein; the step S3 includes the steps of:
s3-1: after the INS/GNSS subsystem and the INS/CNS subsystem are respectively executed with filtering processes, corresponding local posterior state estimation can be obtained
S3-2: obtaining global optimal state estimation according to the minimum variance principle and the volume criterion:
5. The optimal data fusion method applicable to ballistic missile INS/CNS/GNSS combined navigation system according to claim 4, wherein: p in the step S3-212And P21The volume is obtained by approximation through a volume criterion, and specifically comprises the following steps:
wherein the content of the first and second substances,respectively representing the propagation volume points obtained by the INS/GNSS subsystem and the INS/CNS subsystem executing the step S2-2
Technical Field
The invention relates to a data fusion method, in particular to an optimal data fusion method suitable for a ballistic missile INS/CNS/GNSS combined navigation system.
Background
Traditional EKF, UKF based federated multi-sensor data fusion methods use the upper bound of process noise covariance to eliminate the correlation between local state estimates rather than the process noise covariance itself, so the resulting global state estimate is suboptimal. Furthermore, this type of method requires that the nonlinear system model must achieve sufficient accuracy. However, in the process of ballistic missile flight, on one hand, the missile-borne INS/CNS/GNSS integrated navigation system is easily affected by a complex environment, which causes measurement information of the INS/CNS/GNSS integrated navigation system to be affected by non-gaussian noise, and in this case, if the conventional non-linear filtering method based on gaussian assumption is continuously used, the navigation accuracy of the integrated navigation system will be seriously reduced; on the other hand, because the real dynamic system model is very complex, the established system process model can only be a theoretical approximation of the real model, so that a modeling error exists, and the modeling error can also cause the reduction of the navigation precision of the integrated navigation system.
Therefore, a new technical solution is needed to solve the above problems.
Disclosure of Invention
The purpose of the invention is as follows: in order to overcome the defect of poor navigation accuracy of an integrated navigation system in the prior art, an optimal data fusion method suitable for a ballistic missile INS/CNS/GNSS integrated navigation system is provided, and global optimal state estimation can still be obtained under the condition that a process model contains uncertainty and measurement noise is non-Gaussian noise.
The technical scheme is as follows: the invention provides an optimal data fusion method suitable for a ballistic missile INS/CNS/GNSS combined navigation system, which comprises the following steps:
s1: constructing an INS/CNS/GNSS combined navigation system filtering model;
s2: respectively introducing a self-adaptive fading factor and a maximum correlation entropy criterion to carry out local state estimation of the INS/CNS subsystem and the INS/GNSS subsystem in a time updating stage and a measurement updating stage of the generalized high-order CKF;
s3: and according to the minimum variance principle and the volume criterion, local estimation of the INS/CNS subsystem and the INS/GNSS subsystem is fused to obtain global optimal state estimation.
Further, the constructing an INS/CNS/GNSS integrated navigation system filter model in step S1 includes the following steps:
s1-1: setting the state vector of the INS/CNS/GNSS integrated navigation system as x ═ phi [ [ phi ] ]x,φy,φz,δvx,δvy,δvz,x,y,z,εx,εy,εz,Δx,Δy,Δz]TWherein phix,φy,φzRepresenting the angular misalignment of the attitude, δ v, under the inertial system of the emission pointx,δvy,δvzRepresenting the velocity error in the inertial system of the emission point, x, y, z representing the position error in the inertial system of the emission point, epsilonx,εy,εzRepresenting the gyro constant drift, Delta, in a projectile coordinate systemx,Δy,ΔzRepresenting the constant bias of the accelerometer in a missile body coordinate system, wherein T is a transposition symbol;
s1-2: establishing a state equation of the system according to a 15-dimensional state vector x of the INS/CNS/GNSS combined navigation system:
xk=f(xk-1)+vk-1
wherein f (-) is a nonlinear system function, xk-1And xkState vectors, v, representing time k-1 and k, respectivelyk-1Representing process noise, vk-1Has a covariance of
S1-3: respectively establishing measurement equations of an INS/GNSS subsystem and an INS/CNS subsystem:
in an INS/GNSS subsystem, taking the difference between the positions and the speed output by the INS and the GNSS as measurement information, and establishing a measurement equation of the subsystem:
z1,k=H1,kxk+ω1,k
wherein z is1,kMeasurement vector, H, representing INS/GNSS subsystem time k1,kMeasurement matrix, ω, representing INS/GNSS subsystem k time1,kRepresents the measurement noise at the moment k of the INS/GNSS subsystem, the variance of which is
In an INS/CNS subsystem, the difference value of the attitude angles output by the INS and the CNS is used as measurement information, and a measurement equation of the subsystem is established:
z2,k=H2,kxk+ω2,k
wherein z is2,kThe measurement vector, H, representing the INS/CNS subsystem at time k2,kMeasurement matrix, ω, representing INS/CNS subsystem k time2,kRepresents the measured noise at the moment k of the INS/CNS subsystem with a variance of
Further, the step S2 specifically includes the following steps:
s2-1: since the INS/CNS subsystem and the INS/GNSS subsystem use the same filtering process for local state estimation, only the filtering process of the INS/GNSS subsystem will be specifically described herein to avoid repetition. Therefore, z in step S1-31,k、z2,kIs uniformly written as zk,H1,kAnd H2,kWrite uniformly as Hk,ω1,kAnd ω2,kWrite uniformly to omegak,R1,kAnd R2,kIs written uniformly as RkSetting the value of the kernel width gamma according to the local state estimation requirement of the INS/GNSS subsystem, and setting the initial state vector, the state error covariance and the fading factor as
τk(0)=1,S0|0=chol(P0|0) Chol (·) denotes the cholesky decomposition operation;s2-2: according to the formula
Calculating propagation volume pointsWhere i 1., 2n2+1, n represents the dimension of the state vector,[·]ia set of representations [ ·]The ith column;s2-3: predicting state vectors at time k
Error covariance matrixAnd calculating Sk|k-1=chol(Pk|k-1) WhereinS2-4: according to
And Sk|k-1Generating new volume pointsAnd predict the measurement information at time kS2-5: according to the INS/GNSS subsystem' S measurement equation and in step S2-3
And Pk|k-1The following equation was constructed:
wherein, I represents an identity matrix,
Mp,k|k-1=chol(Pk|k-1)、Mr,k=chol(Rk) Andin the equation
Is multiplied byObtaining:order toThe above equation can be rewritten as: dk=BkXk+ek;S2-6: updating the measurement noise variance:
wherein diag (-) denotes the diagonalization of the matrix, m is the dimension of the measurement information, di,krepresents DkThe ith element of (b)i,kIs represented by BkRow i element of (1);s2-7: computing a cross-covariance matrix P between state information and metrology informationxz,k|k-1,And let i equal to 0;
s2-8: order to
And setting chi-square distributionWhere θ represents the degree of freedom of the chi-squared distribution and α represents the quantile of the chi-squared distribution ifThen calculate Returning to the step S2-2 to continue executing the next filtering cycle;S2-9:
if it is not
ComputingWherein, i ═ i +1, i ═ i · is obtained according to the adaptive fading factor τkUpdate ak(i) A value of (a) is judgedk(i) Whether or not to satisfyIf not, returning to the step S2-9 to continue execution; if so, obtaining the optimal adaptive fading factor taukThen continuing to perform subsequent steps;s2-10: using an adaptive fading factor taukUpdating predicted state prediction covariance:
Calculating a Kalman gain:estimating a posterior state and a posterior covariance matrix:returning to step S2-2 continues to execute the next filtering cycle.Further, the step S3 includes the following steps:
s3-1: after the INS/GNSS subsystem and the INS/CNS subsystem are respectively executed with filtering processes, corresponding local posterior state estimation can be obtained
For the convenience of differentiation, the INS/GNSS subsystem is made to obtain a local posterior state estimate asThe INS/CNS subsystem obtains a local posterior state estimate of53-2: obtaining global optimal state estimation according to the minimum variance principle and the volume criterion:
wherein let beta be [ beta ]1,β2]T,P11P derived from performing local state estimation for INS/GNSS subsystemsk|k,P22P derived from performing local state estimation for INS/CNS subsystemsk|k,P12To representAndcross covariance matrix of estimation errors, P21To representAnda cross-covariance matrix of the errors is estimated.Further, P12And P21Approximated by the volume criterion:
wherein the content of the first and second substances,
respectively representing the propagation volume points obtained by the INS/GNSS subsystem and the INS/CNS subsystem executing the step S2-2Andrespectively representing the predicted values of the state vectors obtained by the INS/GNSS subsystem and the INS/CNS subsystem executing the step S2-3,andrepresents the new volume point X obtained by the INS/GNSS subsystem and the INS/CNS subsystem executing the step S2-4 respectivelyi,k|k-1,Andrespectively indicating the predicted values of the measurement information obtained by the INS/GNSS subsystem and the INS/CNS subsystem executing the step S2-4 Andrespectively representing the filter gains obtained by the INS/GNSS subsystem and the INS/CNS subsystem executing the local state estimation process I denotes an n-dimensional identity matrix.Has the advantages that: compared with the prior art, the invention respectively introduces the self-adaptive fading factor and the maximum correlation entropy criterion to carry out the local state estimation of the INS/GNSS subsystem and the INS/CNS subsystem in the time updating stage and the measurement updating stage of the generalized high-order CKF, and finally fuses the local estimation of the INS/CNS subsystem and the INS/GNSS subsystem according to the minimum variance principle and the volume criterion, under the condition that the ballistic missile INS/CNS/GNSS combined navigation system has process modeling error and measurement noise is in non-Gaussian distribution in a system model, the integrated navigation system can still obtain the global optimal state estimation, the invention can simultaneously inhibit the influence of process modeling errors and non-Gaussian measurement noise on the state estimation, improve the adaptivity and robustness of ballistic missile INS/CNS/GNSS integrated navigation, therefore, the navigation precision of the INS/CNS/GNSS combined navigation system is ensured.
Drawings
FIG. 1 is a flow chart of the method of the present invention.
Detailed Description
The invention is further elucidated with reference to the drawings and the embodiments.
As shown in fig. 1, the present invention provides an optimal data fusion method suitable for a ballistic missile INS/CNS/GNSS integrated navigation system, comprising the following steps:
s1: constructing an INS/CNS/GNSS combined navigation system filtering model:
s1-1: setting the state vector of the INS/CNS/GNSS integrated navigation system as follows:
x=[φx,φy,φz,δvx,δvy,δvz,x,y,z,εx,εy,εz,Δx,Δy,Δz]T (1)
wherein phix,φy,φzRepresenting the angular misalignment of the attitude, δ v, under the inertial system of the emission pointx,δvy,δvzRepresenting the velocity error in the inertial system of the emission point, x, y, z representing the position error in the inertial system of the emission point, epsilonx,εy,εzRepresenting the gyro constant drift, Delta, in a projectile coordinate systemx,Δy,ΔzAnd T is a transposed symbol and represents the constant bias of the accelerometer in the missile coordinate system.
S1-2: establishing a state equation of the system according to a 15-dimensional state vector x of the INS/CNS/GNSS combined navigation system:
xk=f(xk-1)+vk-1 (2)
wherein f (-) is a nonlinear system function, xk-1And xkState vectors, v, representing time k-1 and k, respectivelyk-1Representing process noise, vk-1Has a covariance of
S1-3: respectively establishing measurement equations of an INS/GNSS subsystem and an INS/CNS subsystem:
in an INS/GNSS subsystem, the INS and the GNSS output position and speed are respectively differenced and used as measurement information to establish a measurement equation of the subsystem:
z1,k=H1,kxk+ω1,k (3)
wherein z is1,kMeasurement vector, H, representing INS/GNSS subsystem time k1,kMeasurement matrix, ω, representing INS/GNSS subsystem k time1,kRepresents the measurement noise at the moment k of the INS/GNSS subsystem, the variance of which is
In an INS/CNS subsystem, a measurement equation of the subsystem takes the difference value of the attitude angles output by the INS and the CNS as measurement information, and the measurement equation of the subsystem is established as follows:
z2,k=H2,kxk+ω2,k (4)
wherein z is2,kThe measurement vector, H, representing the INS/CNS subsystem at time k2,kMeasurement matrix, ω, representing INS/CNS subsystem k time2,kRepresents the measured noise at the moment k of the INS/CNS subsystem with a variance of
S2: respectively introducing a self-adaptive fading factor and a maximum correlation entropy criterion to carry out local state estimation of an INS/CNS subsystem and an INS/GNSS subsystem in a time updating stage and a measurement updating stage of the generalized high-order CKF, wherein the specific process is as follows:
s2-1: because the INS/GNSS subsystem and the INS/CNS subsystem perform local state estimation through the local filter 1 and the local filter 2 respectively by using the same filtering process, in order to avoid repetition, only the filtering process of the INS/GNSS subsystem is specifically described here: therefore, the present embodiment converts z in step S11,k、z2,kIs uniformly written as zk,H1,kAnd H2,kWrite uniformly as Hk,ω1,kAnd ω2,kWrite uniformly to omegak,R1,kAnd R2,kIs written uniformly as Rk. According to INS/GNSSThe local state estimation of the system requires setting the value of kernel width gamma, and setting the initial state vector, state error covariance and fading factor as
τk(0)=1,S0|0=chol(P0|0) Chol (·) denotes the choles decomposition operation.S2-2: calculating a propagation volume point according to the formula (5)
Where i 1., 2n2+1, n represents the dimension of the state vector,[·]ia set of representations [ ·]Column i.
S2-3: predicting a state vector and an error covariance matrix at the k moment:
wherein
S2-4: according to
Yang Sk|k-1Generate new volume points:
wherein S isk|k-1=chol(Pk|k-1)。
S2-5: prediction of measurement information at time k:
s2-6: measurement equations and from INS/GNSS subsystemsAnd Pk|k-1The following equation was constructed:
wherein, I represents an identity matrix,Mp,k|k-1=chol(Pk|k-1)、Mr,k=chol(Rk) And
in the equation
Is multiplied byObtaining:
order to
Equation (11) can be rewritten as:Dk=BkXk+ek (12)
s2-7: updating the measurement noise variance:
wherein diag (-) denotes the diagonalization of the matrix, m is the dimension of the measurement information,
di,krepresents DkThe ith element of (b)i,kIs represented by BkRow i element of (1).S2-8: computing a cross-covariance matrix P between state information and metrology informationxz,k|k-1:
Let i be 0 and/or n be 0,
and setting chi-square distributionWhere θ represents the degree of freedom of the chi-squared distribution and α represents the quantile of the chi-squared distribution;if it is not
Then calculate:
Sk|k=chol(Pk|k) (19)
continuing to execute k-k +1, and returning to the position of the formula (5) in the step S2-2 to start executing the next filtering cycle;
if it is not
And (3) calculating:
according to the obtained adaptive fading factor taukUpdate ak(i) A value of (a) is judgedk(i) Whether or not to satisfy
If the value is not satisfied, i is equal to i +1, the formula (20) is returned to update the extinction factor τkA value of (d); if so, obtaining the optimal adaptive fading factor taukThen the following steps are continued.S2-9: using the obtained adaptive fading factor taukUpdate predicted state prediction covariance:
calculating a Kalman gain:
s2-10: estimating a posterior state and a posterior covariance matrix:
Sk|k=chol(Pk|k) (25)
k is executed as k +1, and the execution of the next filtering cycle is started returning to the position of the formula (5) in step S2-2.
S3: and (3) fusing local estimation of the INS/GNSS subsystem and the INS/CNS subsystem according to a minimum variance principle and a volume criterion to obtain a global optimal state estimation:
s3-1: after the INS/GNSS subsystem and the INS/CNS subsystem are respectively executed with the filtering process, corresponding local posterior state estimation can be obtained
For the convenience of differentiation, the INS/GNSS subsystem is made to obtain a local posterior state estimate asThe INS/CNS subsystem obtains a local posterior state estimate ofS3-2: obtaining global optimal state estimation according to the minimum variance principle and the volume criterion:
wherein let beta be [ beta ]1,β2]T,
P11P derived from performing local state estimation for INS/GNSS subsystemsk|k,P22P derived from performing local state estimation for INS/CNS subsystemsk|k,P12To representAndcross covariance matrix of estimation errors, P21To representAndcross covariance matrix of estimation errors, P12And P21Approximated by the volume criterion:
wherein the content of the first and second substances,respectively representing the propagation volume points obtained during the INS/GNSS subsystem and the INS/CNS subsystem executing the step S2
Andrespectively representing the predicted values of the state vectors obtained by the INS/GNSS subsystem and the INS/CNS subsystem executing the step S2,andrespectively representing INS/GNSS sub-elementsThe system and the INS/CNS subsystem perform the new volume point X obtained during step S2i,k|k-1,Andrespectively representing the predicted values of the measurement information obtained by the INS/GNSS subsystem and the INS/CNS subsystem executing the step S2 Andrespectively representing the filter gains obtained during the INS/GNSS subsystem and the INS/CNS subsystem executing the step S2 I denotes an n-dimensional identity matrix.