CN102252677A - Time series analysis-based variable proportion self-adaptive federal filtering method - Google Patents

Abstract

The invention discloses a time series analysis-based variable proportion self-adaptive federal filtering method, and the method is utilized for underwater multi-sensor integrated navigation systems. The time series analysis-based variable proportion self-adaptive federal filtering method is characterized by establishing system state equations and measurement equations according to error equations belonging to all navigation sensor systems, carrying out discretization processing of the established equations, creating discrete state-space models respectively corresponding to the all navigation sensor systems, acquiring information weight of the navigation sensor systems through an autoregressive model according to historical data of the navigation sensor systems, acquiring an information distribution ratio according to the obtained information weight and a law of information conservation, realizing global optimal estimates, and resetting a filtering value and an evaluated error covariance matrix by the global optimal estimates. The time series analysis-based variable proportion self-adaptive federal filtering method improves system navigation precision, system stability and fault tolerance, and can satisfy requirements which belong to underwater navigation devices and comprise high precision and high reliability requirements.

Description

The federal filtering method of a kind of control with changed scale self-adaptation based on time series analysis

Technical field

The invention belongs to the multi-sensor combined navigation system technical field, relate to a kind of multi-sensor combined navigation system filtering method, specifically is the federal filtering method of a kind of control with changed scale self-adaptation based on time series analysis.

Background technology

Progress along with airmanship, the performance of any single navigator and range of application all show tangible limitation, can't satisfy the growing accuracy requirement of vehicle, also can't adapt to complicated applied environment, can't satisfy the requirement of system reliability fully.Horst Ahlers says: " belong to multisensor future." multi-sensor combined navigation system becomes a kind of inevitable trend.Along with the improvement of hardware condition, fusion method becomes the key factor of restriction combined system performance.The federal kalman filter method that is grown up by the N.A.Carlson proposition is the fusion method that success applies to engineering practice, and its adopts noise variance upper bound technology to eliminate sub-filtering correlativity, reaches global optimum.

But in existing federal filter structure, the information distribution ratio is fixing along with determining of structure, it can not reflect the variation of navigation sensor duty and data mode, can not accurately reflect navigation sensor information weight allocation, the situation that can not accurately reflect system really is unfavorable for improving the research of system accuracy, reliability and fault-tolerance.

Summary of the invention

The present invention is directed to information distribution ratio determining in the existing federal filtering method along with structure, the problem that can not accurately reflect navigation sensor information weight allocation, provide a kind of control with changed scale self-adaptation federal filtering method based on time series analysis, information distribution ratio and navigation sensor working condition and carrier running environment situation are adapted, reach the purpose that improves system accuracy, reliability and fault-tolerance.

The federal filtering method of a kind of control with changed scale self-adaptation based on time series analysis specifically may further comprise the steps:

Step 1, set up system state equation and measurement equation according to the error equation of each navigation sensor system of integrated navigation system;

Step 2, with the system state equation and the measurement equation discretize that obtain, set up the discrete state spatial model of each navigation sensor system correspondence;

Step 3, according to the historical data of navigation sensor system, obtain the information weights of this navigation sensor system by autoregressive model;

Step 4, obtain the information distribution ratio according to information weights and information conservation law;

Step 5, carry out federal filtering, obtain global optimum and estimate, estimate replacement filter value and estimation error variance battle array with global optimum at last according to the information distribution ratio.

Integrated navigation system in the described step 1, according to the navigation sensor configuring condition of certain submarine navigation device, this integrated navigation system is by inertial navigation system INS, electrostatically supported gyro monitor ESGM, NAVSTAR GPS and Doppler range rate measuring system DVL form.

Set up INS state equation, INS/ESGM measurement equation, INS/GPS measurement equation and INS/DVL system state equation and measurement equation respectively according to each systematic error equation:

Described INS state equation is:

Wherein, quantity of state

δ L, δ λ, δ h represent the variation of latitude L, longitude λ, height h, δ V respectively _E, δ V _N, δ V _URepresent east orientation speed V respectively _E, north orientation speed V _N, day to speed V _UVariation, φ _E, φ _N, φ _URepresent pitch angle, roll angle, crab angle respectively, ε _Bx, ε _By, ε _BzBe the component of gyroscopic drift arbitrary constant error on three, ε _Rx, ε _Ry, ε _RzBe the component of gyroscopic drift Markovian process error on three,

Be the component of accelerometer error on three; Process noise battle array W=[w _Bxw _Byw _Bzw _Gxw _Gyw _Gzw _Axw _Ayw _Az] ^T, w _Bx, w _By, w _BzThe component of white noise on three of expression gyroscopic drift arbitrary constant, w _Gx, w _Gy, w _GzThe component of white noise on three of expression gyroscopic drift Markovian process, w _Ax, w _Ay, w _AzThe component of white noise on three of expression accelerometer error; The state transitions battle array Wherein

f _E, f _N, f _UBe respectively east orientation acceleration, north orientation acceleration, sky to acceleration, ω _IeBe rotational-angular velocity of the earth; Noise drives battle array

Wherein

It is direction cosine matrix.

Described INS/ESGM measurement equation is: Z ₁(t)=H ₁(t) X ₁(t)+V ₁(t), wherein, Z ₁For the difference of ESGM and INS position, course information is observed quantity, V ₁(t)=[m ₁, m ₂, m ₃, m ₄] ^TIt is the white Gaussian noise of the zero-mean that produces by the site error of ESGM and azimuthal error; H ₁(t) be to measure battle array, INS/ESGM state equation X ₁(t)=X (t).

Described INS/GPS measurement equation is:

Wherein, position quantity is measured

Speed measurement amount v ₂(t) expression white noise, INS/GPS state equation X ₂(t)=X (t);

Described INS/DVL system state equation is:

Described INS/DVL measurement equation is:

Wherein, the noise of Doppler range rate measuring system drives battle array G _d(t)=and diag[1 1 0], F _d(t) be the augmented state transfer matrix of Doppler range rate measuring system, W _d(t) be the augmentation process noise battle array of Doppler range rate measuring system, X _d(t) be Doppler range rate measuring system augmented state amount, X _d=[δ V _dδ Δ δ C] ^T, δ V _dBe the velocity shifts error in the Doppler range rate measuring system, the δ Δ is the drift angle error in the Doppler range rate measuring system, and δ C is the calibration factor error in the Doppler range rate measuring system;

V ₃(t)=[m _VEm _VN] ^TBe observation white noise, V _DE, V _DNRepresent east orientation speed, north orientation speed in the Doppler range rate measuring system respectively, K ' is the angle, ship's head.

In the described step 2, the discrete state spatial model of each navigation sensor system correspondence is:

$\left\{\begin{array}{c}{X}_k={\mathrm{\Φ}}_{k/k-1}{X}_{k-1}+{\mathrm{\Γ}}_{k/k-1}{W}_{k-1}\\ Z_k=H_k{X}_k+V_k\end{array}\right.$

Wherein, k represents the index value of discrete time state, X _kExpression k quantity of state constantly, Z _kExpression k observed quantity constantly, H _kExpression k measurement battle array constantly, Γ _K/k-1Be that noise drives battle array, W _K-1Be k-1 system state noise constantly, and process noise covariance matrix Q _k=E[W _kW _k ^T], V _kBe k system measurements noise constantly, and measurement noise covariance matrix R _k=E[V _kV _k ^T], state transitions battle array Φ _K/k-1Adopt substep build up discrete method to obtain:

${\mathrm{\Φ}}_{k/k-1}=\underset{i-1}{\overset{J}{\mathrm{\Π}}}{\mathrm{\Φ}}_{(t_{k-1}+\mathrm{is})/(t_{k-1}+(i-1)s)}\≈\underset{i-1}{\overset{J}{\mathrm{\Π}}}(I+s{F}_{i-1})\≈I+s\underset{i=1}{\overset{J}{\mathrm{\Σ}}}{F}_{i-1}$

J=T/s wherein, T is the sampling time, and s is the step-length of dividing in the sampling time, and I is a unit matrix, F _I-1It is i-1 state transitions battle array constantly.

In the described step 3, with the predicated error of each navigation sensor system as information weights Θ _i:

${\mathrm{\Θ}}_i=y_i\left(\mathrm{\τ}\right)-{\hat{y}}_i\left(\mathrm{\τ}\right)(i=\mathrm{0,1,2,3})$

Wherein, y _i(τ) be the output sequence of i navigation sensor, Be the predicted value of the output sequence of i navigation sensor, i=0, the corresponding inertial navigation system of 1,2,3 difference, electrostatically supported gyro monitor, NAVSTAR and Doppler range rate measuring system.

Described information distribution ratio of step 4 and information weights are directly proportional, and according to the information conservation law, the weight of each navigation sensor system and be

β wherein ₀Be main system weight, β _iBe each subsystem weight, n is the number of all subsystems, and described main system is an inertial navigation system, and subsystem is electrostatically supported gyro monitor, NAVSTAR and Doppler range rate measuring system, and the weight that obtains each navigation sensor system at last is:

The described global optimum of step 5 is estimated as:

Wherein,

Be overall estimated value, P _gFor estimating square error, P _{I, k}Expression is with the estimation error variance battle array P of i navigation sensor system _kPassing ratio β _iDistribute the estimation error variance battle array after resetting,

The quantity of state of i navigation sensor system of expression prediction.

With globally optimal solution replacement filter value and estimation error variance battle array.

Because use variance upper bound technology when the distribution system noise, each subsystem is uncorrelated mutually, augmented system is not coupled yet.Though the filtering result of subsystem is a suboptimum, synthesized again when merging, so the overall situation is estimated optimum.

The advantage and the good effect of the inventive method are:

(1) selection of informative weight has been avoided estimating the error that each navigation sensor system is brought with the Filtering Estimation value based on the source information of each navigation sensor system, makes integrated navigation system structure adaptive system truth more; Obtaining of weights also used the measurement forward data simultaneously, makes the historical data of navigation sensor system obtain sufficient utilization, improved the utilization factor of information.

(2) the inventive method makes the navigation accuracy of integrated navigation system remain on suitable level, and error is little and do not disperse with respect to the INS error, has overcome the INS accumulation of error, especially the weakness of site error.

(3) use the inventive method can make under different navigation sensor configuration situation, the combined navigation system performance held stationary, make integrated navigation system have good filtering accuracy, have good station-keeping ability and error correction ability, have stronger antijamming capability and noise inhibiting ability, improved the reliabilty and availability of integrated navigation system.

Description of drawings

Fig. 1 is that the federal filtering method of the present invention is applied in the synoptic diagram in the multi-sensor combined navigation system;

Fig. 2 is the concrete steps process flow diagram of the federal filtering method of the present invention;

Fig. 3 is a federal filter structure synoptic diagram of using the federal filtering method of the present invention;

Fig. 4 is an information distribution ratio change curve synoptic diagram;

Fig. 5 is emerging system and INS alliance error correlation curve synoptic diagram;

Fig. 6 is emerging system and INS system speed error correlation curve synoptic diagram;

Fig. 7 is the site error correlation curve synoptic diagram of using the emerging system of the inventive method and using the emerging system of classical federal filtering method.

Embodiment

The present invention is described in further detail below in conjunction with drawings and Examples.

Applied multi-sensor combined navigation system is multi-sensor combined navigation system under water in the embodiment of the invention, this integrated navigation system is by inertial navigation system (Inertial Navigation System, be called for short INS), electrostatically supported gyro monitor (Electrostatic Supported Gyroscope Monitor, be called for short ESGM), NAVSTAR (Global Positioning System, be called for short GPS) and Doppler range rate measuring system (Doppler Velocity Log is called for short DVL) composition.

INS is made up of gyroscope and accelerograph, be a kind of autonomous navigation system that does not rely on external information, it can provide multiple navigational parameters such as comprising speed, position and attitude, has anti-interference, therefore round-the-clock advantage is suitable as the benchmark navigational system of underwater hiding-machine.The precision of INS depends primarily on inertia device, and the navigation accuracy error can constantly increase along with the accumulation of time, needs to consider the error stable problem of its long-time navigation.ESGM constitutes " dumbbell type " assembly by two double freedom electrostatic gyroes and an indirect stable platform, and it is as information source monitoring, the compensation inertial navigation system of high precision and long-time stability.ESGM utilizes the monitoring of celestial body navigation principle realization to the INS parameter, and INS then provides indirect stable platform and resolves parameter for ESGM.This array mode can suppress dispersing of navigation error greatly, prolongs the resetting cycle of system, and the long-time high precision of assurance submarine navigation device is cruised.The remarkable advantage of GPS is a real-time navigation, and positioning error and time are irrelevant, and higher location and rate accuracy are arranged.When antenna emerges or sounding buoy when launching, obtain the accurate position of current point, with it calibration navigation error, and the update the system time.DVL adopts four wave beam Doppler sonars of fixed transmit directions based on Doppler effect more, and it not only can compensate the attitude error of carrier in the drift angle measurement, and can measure the ground speed of carrier, thereby provides the space absolute velocity vector of carrier.

As shown in Figure 1, the navigation information process pre-service that integrated navigation system is measured, mainly be data to be picked wild value processing etc. by the distance function method, afterwards the navigation data that collects is carried out time, spatial calibration, through the federal filtering method of control with changed scale self-adaptation the valid data after calibrating are merged then, obtain optimum solution, export navigation data at last.

The federal filtering method of control with changed scale self-adaptation based on time series analysis of the present invention as shown in Figure 2, may further comprise the steps.

Step 1, set up system state equation and measurement equation according to each navigation sensor error equation of integrated navigation system.Need to set up INS state equation, INS/ESGM measurement equation, INS/GPS measurement equation and INS/DVL system state equation and measurement equation in the present embodiment.

Steps A, set up the INS state equation.

The error of inertial navigation system INS comprises triple channel position, speed, attitude error and inertia device error.The triple channel site error is characterized by latitude L, longitude λ and height h, and the error of triple channel speed is by east orientation speed V _E, north orientation speed V _NWith the sky to speed V _UCharacterize, three-channel attitude error is characterized by pitch angle, roll angle and crab angle, and each error equation is respectively:

$\mathrm{\δ}\stackrel{\·}{L}=\frac{\mathrm{\δ}{V}_N}{R_M+h}$

$\mathrm{\δ}\stackrel{\·}{\mathrm{\λ}}=\frac{\mathrm{\δ}{V}_E}{R_N+h}\sec L+\frac{V_E}{R_N+h}\sec \mathrm{LtgL\δL}$

$\mathrm{\δ}\stackrel{\·}{h}=\mathrm{\δ}{V}_U$

$\mathrm{\δ}{\stackrel{\·}{V}}_E=f_N{\mathrm{\φ}}_U-f_U{\mathrm{\φ}}_N+(\frac{V_E}{R_M+h}\mathrm{tgL}-\frac{V_U}{R_M+h}){\mathrm{\δV}}_E+({2\mathrm{\ω}}_{\mathrm{ie}}\sin L+\frac{V_E}{R_N+h}\mathrm{tgL}){\mathrm{\δV}}_N$

$+(2{\mathrm{\ω}}_{\mathrm{ie}}\cos L\·V_N+\frac{V_N{V}_E}{R_N+h}{\sec }^{2}L+2{\mathrm{\ω}}_{\mathrm{ie}}\sin L\·V_U)\mathrm{\δL}-(2{\mathrm{\ω}}_{\mathrm{ie}}\cos L+\frac{V_E}{R_N+h})\mathrm{\δ}{V}_U+{\▿}_E$

$\mathrm{\δ}{\stackrel{\·}{V}}_N=f_U{\mathrm{\φ}}_E-f_E{\mathrm{\φ}}_U-2({\mathrm{\ω}}_{\mathrm{ie}}\sin L+\frac{V_E}{R_N+h}\mathrm{tgL}){\mathrm{\δV}}_E-\frac{V_U}{R_M+h}{\mathrm{\δV}}_N-\frac{V_N}{R_M+h}\mathrm{\δ}{V}_U$

$-(2{\mathrm{\ω}}_{\mathrm{ie}}\cos L+\frac{V_E}{R_N+h}{\sec }^{2}L)V_E\mathrm{\δL}+{\▿}_N$

$\mathrm{\δ}{\stackrel{\·}{V}}_U=f_E{\mathrm{\φ}}_N-f_N{\mathrm{\φ}}_E+2({\mathrm{\ω}}_{\mathrm{ie}}\cos L+\frac{V_E}{R_N+h}\mathrm{tgL})\mathrm{\δ}{V}_E+2\frac{V_N}{R_M+h}\mathrm{\δ}{V}_N-2{\mathrm{\ω}}_{\mathrm{ie}}\cos L\·V_E\mathrm{\δL}+{\▿}_U$

${\stackrel{\·}{\mathrm{\φ}}}_E=-\frac{\mathrm{\δ}{V}_N}{R_M+h}+({\mathrm{\ω}}_{\mathrm{ie}}\sin L+\frac{V_E}{R_N+h}\mathrm{tgL}){\mathrm{\φ}}_N-({\mathrm{\ω}}_{\mathrm{ie}}\cos L+\frac{V_E}{R_N+h}\mathrm{tgL}){\mathrm{\φ}}_U+{\mathrm{\ϵ}}_E$

${\stackrel{\·}{\mathrm{\φ}}}_N=\frac{\mathrm{\δ}{V}_E}{R_N+h}-{\mathrm{\ω}}_{\mathrm{ie}}\sin \mathrm{L\δL}-({\mathrm{\ω}}_{\mathrm{ie}}\sin L+\frac{V_E}{R_N+h}\mathrm{tgL}){\mathrm{\φ}}_E-\frac{V_E}{R_N+h}{\mathrm{\φ}}_U+{\mathrm{\ϵ}}_N$

${\stackrel{\·}{\mathrm{\φ}}}_U=\frac{\mathrm{\δ}{V}_E}{R_N+h}\mathrm{tgL}+({\mathrm{\ω}}_{\mathrm{ie}}\cos L+\frac{V_E}{R_N+h}{\sec }^{2}L)\mathrm{\δL}+({\mathrm{\ω}}_{\mathrm{ie}}\cos L+\frac{V_E}{R_N+h}){\mathrm{\φ}}_E+\frac{V_N}{R_N+h}{\mathrm{\φ}}_N+{\mathrm{\ϵ}}_U$

Wherein,

Be respectively latitude L, longitude λ, height h, east orientation speed V _E, north orientation speed V _N, day to speed V _U, pitch angle, roll angle, crab angle rate of change; δ L, δ λ, δ h represent the variation of latitude L, longitude λ, height h, δ V respectively _E, δ V _N, δ V _URepresent east orientation speed V respectively _E, north orientation speed V _N, day to speed V _UVariation; f _E, f _N, f _UBe respectively east orientation acceleration, north orientation acceleration, sky to acceleration, ω _IeBe rotational-angular velocity of the earth, R _MBe radius of curvature of meridian, R _NBe radius of curvature in prime vertical.

The inertia device error comprises alignment error, calibration factor sum of errors stochastic error etc., and alignment error and calibration factor error have nothing in common with each other because of conditions such as part category, model, environment, technologies, do not possess unitarity, so only consider stochastic error among the present invention.Stochastic error need be considered gyroscopic drift and accelerometer error.

Gyroscopic drift is: ε=ε _b+ ε _r+ ε _gWherein, ε _bThe expression arbitrary constant, ε _rThe expression first-order Markov process, ε _gWhite noise for gyroscopic drift.Three gyroscopic drift equations are:

${\stackrel{\·}{\mathrm{\ϵ}}}_b=0$

${\stackrel{\·}{\mathrm{\ϵ}}}_r=-\frac{1}{T_g}{\mathrm{\ϵ}}_r+w_g$

T wherein _gThe correlation time of three gyroscopic drifts of expression, w _gThe white noise of expression gyroscopic drift Markovian process.

Accelerometer error can be thought of as first-order Markov process, and the error equation of accelerometer is:

${\stackrel{\·}{\▿}}_a=-\frac{1}{T_a}{\▿}_a+w_a$

T wherein _aThe correlation time of expression accelerometer error, w _aThe white noise of expression accelerometer error.

By the error equation of inertial navigation system INS, can determine that the INS state equation is:

$\stackrel{\·}{X}\left(t\right)=F\left(t\right)X\left(t\right)+G\left(t\right)W\left(t\right)$

Wherein, quantity of state

Triple channel position, speed and attitude and gyro error and accelerometer error have been comprised, ε _Bx, ε _By, ε _BzBe the component of gyroscopic drift arbitrary constant error on three, ε _Rx, ε _Ry, ε _RzBe the component of gyroscopic drift Markovian process error on three,

Be the component of accelerometer error on three, described three refer to x axle, y axle and z axle.

Process noise battle array W=[w _Bxw _Byw _Bzw _Gxw _Gyw _Gzw _Axw _Ayw _Az] ^T, w _Bx, w _By, w _BzThe component of white noise on three of representing the gyroscopic drift arbitrary constant respectively, w _Gx, w _Gy, w _GzThe component of white noise on three of representing the gyroscopic drift Markovian process respectively, w _Ax, w _Ay, w _AzThe component of white noise on three of representing accelerometer error respectively.

Noise drives battle array

Wherein

Be direction cosine matrix, b refers to the hull coordinate system, is connected on the hull, and n refers to navigation coordinate system, chooses according to the navigational system duty.

The state transitions battle array

Wherein,

T wherein _Gx, T _Gy, T _GzThe component of correlation time on three of representing gyroscopic drift respectively, T _Ax, T _Ay, T _AzThe component of correlation time on three of representing accelerometer error respectively;

Step B, set up the INS/ESGM measurement equation.

Electrostatically supported gyro monitor ESGM monitoring, compensation inertial navigation system INS, but it can only provide position and course information.The INS/ESGM system state equation uses the INS state equation, makes X ₁=X.The difference of ESGM and INS position, course information is observed quantity Z ₁:

${\mathrm{<figure-callout\; id="1"\; label="Z"\; filenames="HDA0000055797430000022.png"\; state="\{\{state\}\}">Z</figure-callout>}}_1=\left[\begin{array}{c}{L}_E-L\\ {\mathrm{\&lambda;}}_E-\mathrm{\&lambda;}\\ h_E-h\\ K_E-K\end{array}\right]+\left[\begin{array}{c}{m}_1\\ m_2\\ m_3\\ m_4\end{array}\right]$

Wherein, L _E, L represents the latitude of ESGM and INS, λ respectively _E, λ represents the longitude of ESGM and INS, h respectively _E, h represents the height of ESGM and INS respectively; K _E, K represents the course angle of ESGM and INS respectively.[m ₁, m ₂, m ₃, m ₄] ^TIt is the white Gaussian noise of the zero-mean that produces by the site error of ESGM and azimuthal error.

Corresponding INS/ESGM measurement equation is:

Z ₁(t)＝H ₁(t)X ₁(t)+V ₁(t)

Wherein, V ₁(t)=[m ₁, m ₂, m ₃, m ₄] ^TIt is the white Gaussian noise of the zero-mean that produces by the site error of ESGM and azimuthal error.H ₁(t) be to measure battle array, INS/ESGM state equation X ₁(t)=X (t).

Step C, determine the INS/GPS measurement equation.

The GPS error comprises: the SA error, and the ionosphere transmission error, the troposphere transmission error, ephemeris error, and the clock synchronization error etc.Under position, velocity composition pattern, the above-mentioned error of GPS is considered as measurement noise, do not carry out state and expand, get X ₂(t)=X (t).

Observed quantity Z ₂Be GPS and INS position difference and speed difference.The INS/GPS system state equation uses INS state equation: X ₂(t)=and X (t), the INS/GPS measurement equation is:

$Z_2=\left[\begin{array}{c}{Z}_p\left(t\right)\\ Z_V\left(t\right)\end{array}\right]=H_2\left(t\right)X_2\left(t\right)+v_2\left(t\right)$

Wherein, position quantity is measured

Speed measurement amount

Measure battle array

v ₂(t) the white Gaussian noise battle array of the zero-mean that produces by the site error of GPS and azimuthal error of expression.L _GThe latitude of expression GPS, λ _GThe longitude of expression GPS, h _GThe height of expression GPS, V _GEEast orientation speed among the expression GPS, V _GNNorth orientation speed among the expression GPS, V _GUSky among the expression GPS is to speed.

Step D, determine the INS/DVL system equation.

Doppler log can provide ground speed and drift angle, and its measuring error mainly contains velocity shifts error delta V _d, calibration factor error delta C and drift angle error delta Δ, range rate error and drift angle measuring error can be described with first-order Markov process, and the calibration factor error is an arbitrary constant, and error equation is as follows:

$\left\{\begin{array}{c}\mathrm{\&delta;}{\stackrel{\&CenterDot;}{V}}_d=-{\mathrm{\&beta;}}_d\mathrm{\&delta;}{V}_d+w_d\\ \mathrm{\&delta;}\stackrel{\&CenterDot;}{\mathrm{\&Delta;}}=-{\mathrm{\&beta;}}_{\mathrm{\&Delta;}}\mathrm{\&delta;\&Delta;}+w_{\mathrm{\&Delta;}}\\ \mathrm{\&delta;}\stackrel{\&CenterDot;}{C}=0\end{array}\right.$

$\left[\begin{array}{c}\mathrm{\&delta;}{\stackrel{\&CenterDot;}{V}}_d\\ \mathrm{\&delta;}\stackrel{\&CenterDot;}{\mathrm{\&Delta;}}\\ \mathrm{\&delta;}\stackrel{\&CenterDot;}{C}\end{array}\right]=\left[\begin{array}{ccc}-{\mathrm{\&beta;}}_d& & \\ & -{\mathrm{\&beta;}}_{\mathrm{\&Delta;}}& \\ & & 0\end{array}\right]\left[\begin{array}{c}{\mathrm{\&delta;V}}_d\\ \mathrm{\&delta;\&Delta;}\\ \mathrm{\&delta;C}\end{array}\right]+\left[\begin{array}{c}{w}_d\\ w_{\mathrm{\&Delta;}}\\ 0\end{array}\right]$

Wherein, δ V _d, δ Δ, δ C represent velocity error, drift angle error, calibration factor error change rate, β respectively _d, β _ΔBe respectively velocity error Markovian process, drift angle error Markovian process correlation time constant inverse, w _d, w _ΔThe white noise of expression velocity error, drift angle error.So with X _d=[δ V _dδ Δ δ C] ^TAs the expanding system state array, form the system state equation of INS/GPS with the INS error equation:

${\stackrel{\&CenterDot;}{X}}_3=\left[\begin{array}{c}\stackrel{\&CenterDot;}{X}\\ {\stackrel{\&CenterDot;}{X}}_d\end{array}\right]=\left[\begin{array}{cc}F& \\ & F_d\end{array}\right]\left[\begin{array}{c}X\\ X_d\end{array}\right]+\left[\begin{array}{cc}G& \\ & G_d\end{array}\right]\left[\begin{array}{c}W\\ W_d\end{array}\right]$

Matrix G wherein _d=diag[1 1 0].F _dBe the augmented state transfer matrix of Doppler range rate measuring system, W _dIt is the augmentation process noise battle array of Doppler range rate measurement subsystem.

Observed quantity Z ₃(t) be the speed difference of DVL and INS, the observation equation of DVL/INS is:

$Z_3\left(t\right)=\left[\begin{array}{c}{V}_E-V_{\mathrm{dE}}\\ V_N-V_{\mathrm{dN}}\end{array}\right]=H_3\left(t\right)X_3\left(t\right)+V_3\left(t\right)$

Wherein, measure battle array

V ₃(t)=[m _VEm _VN] ^TBe the observation white noise, K ' is the angle, ship's head, V _DE, V _DNEast orientation speed, the north orientation speed of expression Doppler measurement.

Step 2, with state equation and measurement equation discretize, set up the discrete state spatial model.

Federal filtering method of the present invention adapts the effect weight of senior filter and subfilter and the working condition of navigator and the effective condition of navigation data by adjusting the control with changed scale weight, and the master is filtered into sub-filtering provides feedback information, has good performance.The federal filter structure of the control with changed scale self-adaptation of application the inventive method as shown in Figure 3.Inertial navigation system INS offers senior filter with quantity of state, and electrostatically supported gyro monitor ESGM and INS produce observed quantity Z ₁Send signal to first subfilter, NAVSTAR GPS and INS produce observed quantity Z ₂Send signal to second subfilter, Doppler range rate measuring system DVL and INS produce observed quantity Z3 and send signal to the 3rd subfilter, and the signal behind three sub-filter filterings sends to senior filter.Historical data according to inertial navigation system INS, electrostatically supported gyro monitor ESGM, NAVSTAR GPS and Doppler range rate measuring system DVL is carried out the ratio weights and is calculated information weights Θ by autoregression (AutoRegressive is called for short AR) model _i, offer the foundation that senior filter calculates as the information distribution ratio then, finish all after date senior filters of a filtering and provide overall estimated value to INS and three subfilters

Estimate square error P _gWith the information distribution ratio beta _iTo reset.Through not break time renewal and measurement renewal, draw optimal estimation.

With senior filter and each subsystem state equation and measurement equation discretize all, in the embodiment of the invention INS state equation, INS/ESGM measurement equation, INS/GPS measurement equation and INS/DVL state equation and measurement equation are all carried out discretize.The discrete state spatial model of setting up after corresponding states equation and the measurement equation discretize is:

$\left\{\begin{array}{c}{X}_k={\mathrm{\&Phi;}}_{k/k-1}{X}_{k-1}+{\mathrm{\&Gamma;}}_{k/k-1}{W}_{k-1}\\ Z_k=H_k{X}_k+V_k\end{array}\right.$

Wherein, k represents the index value of discrete time state, X _kExpression k quantity of state constantly, Z _kExpression k observed quantity constantly, H _kExpression k measurement battle array constantly, Γ _K/k-1Be that noise drives battle array, W _K-1Be the system state noise, and process noise covariance matrix Q _k=E[W _kW _k ^T], V _kBe the system measurements noise, and measurement noise covariance matrix R _k=E[V _kV _k ^T].State transitions battle array Φ _K/k-1Adopt substep build up discrete method to obtain:

${\mathrm{\&Phi;}}_{k/k-1}=\underset{i-1}{\overset{J}{\mathrm{\&Pi;}}}{\mathrm{\&Phi;}}_{(t_{k-1}+\mathrm{is})/(t_{k-1}+(i-1)s)}\&ap;\underset{i-1}{\overset{J}{\mathrm{\&Pi;}}}(I+s{F}_{i-1})\&ap;I+s\underset{i=1}{\overset{J}{\mathrm{\&Sigma;}}}{F}_{i-1}$

J=T/s wherein, T is the sampling time, and s is the step-length of dividing in the sampling time, and I is a unit matrix, F _I-1It is i-1 state transitions battle array constantly.

Step 3, according to the historical data that records before each navigation sensor system, draw the information weights by autoregressive model.

The calculating of navigation sensor system information weights draws by autoregressive model (AR model) analysis based on the some historical datas that record before the navigation sensor system.Be that example illustrates with some navigation sensors system below.

If y (τ) is the output sequence of this navigation sensor system, the AR model representation of y (τ) is:

Wherein, τ represents τ point of sequences y (τ), and N represents the exponent number of AR model, a _kFor average is zero, variance is σ ²White Gaussian noise, a _kThe representation model error,

Represent N rank AR model coefficient, can obtain the power spectrum density S of y (τ) _y(ω) be:

Expression τ rank AR model coefficient, e ^{-j ω τ}The frequency response of expression system, ω represents angular frequency, j represents imaginary part unit.The autocorrelation function battle array R of y (τ) _N+1For:

$R_{N+1}=\left[\begin{array}{ccccc}{\hat{r}}_y\left(0\right)& {\hat{r}}_y\left(1\right)& {\hat{r}}_y\left(2\right)& ...& {\hat{r}}_y\left(N\right)\\ {\hat{r}}_y\left(1\right)& {\hat{r}}_y\left(0\right)& {\hat{r}}_y\left(2\right)& ...& {\hat{r}}_y(N-1)\\ {\hat{r}}_y\left(2\right)& {\hat{r}}_y\left(1\right)& {\hat{r}}_y\left(0\right)& ...& {\hat{r}}_y(N-2)\\ .& .& .& .& .\\ .& .& .& .& .\\ .& .& .& .& .\\ {\hat{r}}_y\left(N\right)& {\hat{r}}_y(N-1)& {\hat{r}}_y(N-2)& ...& {\hat{r}}_y\left(0\right)\end{array}\right]$

According to the Yule-Waker equation, the autocorrelation function R of known y (τ) _N+1Can obtain the model coefficient of AR model

Suppose

Be the AR model coefficient

Estimated value, i=1,2 ... N,

Be the error power of N rank AR model,

Be the least error power of prediction,

According to the character of Toeplita matrix, determine model coefficient by the Levinson-Durbin recursive algorithm:

Obtain exponent number by final predicated error (FPE) rule:

$\mathrm{FPE}\left(N\right)={\widehat{\mathrm{\&rho;}}}_N\frac{\mathrm{\&Xi;}+(N+1)}{\mathrm{\&Xi;}-(N+1)}$

Wherein, Ξ is the length of measurement data.

Thereby obtain the predicted value of sensor output:

Because the AR model is the all-pole modeling that is based upon on the stationary solution space, so predicated error can reflect the flatness of sensor output.With the predicated error of each sensor as information weights Θ _i, informative weight is more little, and predicated error is more little, and navigation sensor output is level and smooth more.Described information weights are:

${\mathrm{\&Theta;}}_i=y_i\left(\mathrm{\&tau;}\right)-{\hat{y}}_i\left(\mathrm{\&tau;}\right)(i=\mathrm{0,1,2,3})$

y _i(τ) be the output sequence of i navigation sensor,

Be the predicted value of the output sequence of i navigation sensor, i=0, the corresponding inertial navigation system of 1,2,3 difference, electrostatically supported gyro monitor, NAVSTAR and Doppler range rate measuring system.

Step 4, according to information weights and information conservation law computing information allocation proportion.

Information distribution ratio selection principle is: the subsystem precision is high more, exports reliably more, will reduce the effect of main system estimated value more, and information distribution is than more little.Described main system is meant baseline system, is inertial navigation system in the embodiment of the invention, and subsystem is that electrostatically supported gyro monitor, NAVSTAR and Doppler range rate measuring system are from system.Therefore the breath allocation proportion of winning the confidence and information weights are directly proportional.Simultaneously, according to information conservation: β wherein ₀Be main system weight, β _iBe each subsystem weight, n is the number of all subsystems.Therefore the weight that obtains each navigation sensor system at last is:

${\mathrm{\&beta;}}_i=\frac{{\mathrm{\&Theta;}}_i}{\underset{i=0}{\overset{n}{\mathrm{\&Sigma;}}}{\mathrm{\&Theta;}}_i}$

Step 5, carry out federal filtering, obtain global optimum and estimate according to the information distribution ratio.

At first obtain information distribution:

$P_{i,k-1}={\mathrm{\&beta;}}_i^{-1}{P}_{k-1}$

$Q_{i,k-1}={\mathrm{\&beta;}}_i^{-1}{Q}_{k-1}$

X _i，k-1＝X _k-1

Estimation error variance battle array P with i navigation sensor system _K-1Passing ratio β _iBe assigned as the estimation error variance battle array P after the replacement _{I, k-1}, again with the process noise covariance matrix Q of i navigation sensor system _K-1Passing ratio β is assigned as the process noise covariance matrix Q after the replacement _{I, k-1}X _{I, k-1}Expression is with the quantity of state of i navigation sensor system, and k represents the index value of discrete time state.

Next time of carrying out upgrades:

${\hat{X}}_{i,k/k-1}={\mathrm{\&Phi;}}_{i,k/k-1}{\hat{X}}_{i,k-1}$

$P_{i,k/k-1}={\mathrm{\&Phi;}}_{i,k/k-1}{P}_{i,k-1}{\mathrm{\&Phi;}}_{i,k/k-1}^T+{\mathrm{\&Gamma;}}_{i,k/k-1}{Q}_{i,k-1}{\mathrm{\&Gamma;}}_{i,k/k-1}^T$

Be the state one-step prediction battle array of i navigation sensor system, P _{I, k/k-1}Be the one-step prediction square error battle array of i navigation sensor system, Φ _{I, k/k-1}Be the state transitions battle array of i navigation sensor system, Γ _{I, k/k-1}The noise that is i navigation sensor system drives battle array,

Be quantity of state X _{I, k-1}Predicted value.

Observe renewal then:

$K_{i,k}=P_{i,k/k-1}{H}_{i,k}^T{(H_{i,k}{P}_{i,k/k-1}{H}_{i,k}^T+R_{i,k})}^{-1}$

${\hat{X}}_{i,k}={\hat{X}}_{i,k/k-1}+K_{i,k}(Z_{i,k}-H_{i,k}{\hat{X}}_{i,k/k-1})$

$P_{i,k}=(I-K_{i,k}{H}_{i,k})P_{i,k/k-1}{(I-K_{i,k}{H}_{i,k})}^T+K_{i,k}{R}_{i,k}{K}_{i,k}^T$

K _{I, k}Be the filter gain of i navigation sensor system, H _{I, k}Be the measurement battle array of i navigation sensor system, R _{I, k}Be system measurements noise covariance battle array by i navigation sensor system, Z _{I, k}It is the observed quantity of i navigation sensor system.K represents the index value of discrete time state.

Obtain the global optimization value at last:

${\hat{X}}_g=P_g\underset{i=1}{\overset{m}{\mathrm{\&Sigma;}}}{P}_{i,k}^{-1}{\hat{X}}_{i,k}$

$P_g={\left(\underset{i=1}{\overset{m}{\mathrm{\&Sigma;}}}{P}_{i,k}^{-1}\right)}^{-1}$

Be overall estimated value, P _gFor estimating square error, m is the subfilter number, and embodiment of the invention m is 3.

With globally optimal solution replacement filter value and estimation error variance battle array.Because use variance upper bound technology when the distribution system noise, each subsystem is uncorrelated mutually, augmented system is not coupled yet.Though the filtering result of subsystem is a suboptimum, synthesized again when merging, so the overall situation is estimated optimum.

Carry out emulation experiment below by the integrated navigation system under the mathematical analysis software matlab environment.Emulation experiment is 8 hours.Following parameter is set, and the error result of federal filtering method of control with changed scale provided by the invention and INS compares.0.1 °/h of INS constant value drift, drift Markovian process time constant T _gBe 300s, add table zero and be 0.1mg partially, add table Markovian process time constant T _aBe 1000s.0.04 °/h of ESGM white noise.GPS output adds position, the velocity error of stochastic distribution.DVL error 0.4m/s.

Ratio weight change curve such as Fig. 4.From Fig. 4, can reflect the ratio weights situation of change under the sensor different operating condition, when time spent, the weight beta of INS are mainly done in the INS performance ₀Be lower than other ratio values, when GPS or DVL carry out timing, its ratio weights will be higher than other system, have given prominence to the effect of corrective system.Obtain systematic error curve such as Fig. 5 and Fig. 6.Wherein red solid line is represented the error of the multi-sensor combined navigation system under water of the embodiment of the invention, and black dotted lines is represented the INS systematic error.As shown in Figure 5 and Figure 6, the absolute value of combined system site error peak value is 20m, and the absolute value of velocity error peak value is 0.8m/s, and does not disperse.And the INS site error is 500m, velocity error 5m/s, and constantly increase.As can be seen, the navigation information precision that federal filtering method provided by the invention obtains is better than the single-sensor system, and system output held stationary and not dispersing in certain amplitude, overcome the weakness of the inertial navigation system accumulation of error (especially site error), improved the reliability and stability of system.Constantly change in the navigation sensor informative weight, especially the site error variance that calculates under the situation than big saltus step that GPS and DVL correction is switched is 8.92, as seen velocity error variance 0.065 uses the combined navigation system performance of the inventive method more steady.As shown in Figure 7, use the federal filtering method of control with changed scale self-adaptation of the present invention the dynamic property and the site error amplitude of integrated navigation system, compare with the integrated navigation system that uses conventional federal filtering method, the site error changes in amplitude of the integrated navigation system of use the inventive method is little, have clear superiority, can reflect the situation of integrated navigation system more truly.

The present invention adopts time series analysis to obtain the information weights by analyzing the plurality of data of navigation sensor forward direction, and the calculating filter weights with this ratio that changes information distribution, reach the purpose that improves system accuracy, reliability and fault-tolerance again.Than the federal filtering of other self-adaptation, determining of native system informative weight is source information according to navigation sensor, but not the estimated value after Kalman filtering is handled makes the truth near navigation sensor determined more of weights like this, thereby reflects the situation of system more truly.

Claims (7)

1. one kind based on the federal filtering method of the control with changed scale self-adaptation of time series analysis, it is characterized in that this method specifically may further comprise the steps:

Step 1, set up system state equation and measurement equation according to the error equation of each navigation sensor system of integrated navigation system;

Step 2, with the system state equation and the measurement equation discretize that obtain, set up the discrete state spatial model of each navigation sensor system correspondence;

Step 3, according to the historical data of navigation sensor system, obtain the information weights of this navigation sensor system by autoregressive model;

Step 4, obtain the information distribution ratio according to information weights and information conservation law;

Step 5, carry out federal filtering, obtain global optimum and estimate, estimate replacement filter value and estimation error variance battle array with global optimum at last according to the information distribution ratio.

2. the federal filtering method of a kind of control with changed scale self-adaptation according to claim 1 based on time series analysis, it is characterized in that, described integrated navigation system is multi-sensor combined navigation system under water, is made up of four navigation sensor systems: inertial navigation system INS, electrostatically supported gyro monitor ESGM, NAVSTAR GPS and Doppler range rate measuring system DVL.

3. the federal filtering method of a kind of control with changed scale self-adaptation according to claim 1 and 2 based on time series analysis, it is characterized in that described system state equation of step 1 and measurement equation comprise: INS state equation, INS/ESGM measurement equation, INS/GPS measurement equation and INS/DVL system state equation and measurement equation;

Described INS state equation is:

Wherein, quantity of state

δ L, δ λ, δ h represent the variation of latitude L, longitude λ, height h, δ V respectively _E, δ V _N, δ V _URepresent east orientation speed V respectively _E, north orientation speed V _N, day to speed V _UVariation, φ _E, φ _N, φ _URepresent pitch angle, roll angle, crab angle respectively, ε _Bx, ε _By, ε _BzBe the component of gyroscopic drift arbitrary constant error on three, ε _Rx, ε _Ry, ε _RzBe the component of gyroscopic drift Markovian process error on three,

Be the component of accelerometer error on three; Process noise battle array W=[w _Bxw _Byw _Bzw _Gxw _Gyw _Gzw _Axw _Ayw _Az] ^T, w _Bx, w _By, w _BzThe component of white noise on three of expression gyroscopic drift arbitrary constant, w _Gx, w _Gy, w _GzThe component of white noise on three of expression gyroscopic drift Markovian process, w _Ax, w _Ay, w _AzThe component of white noise on three of expression accelerometer error; Described three is x axle, y axle and z axle; Noise drives battle array

Wherein

It is direction cosine matrix; The state transitions battle array

Wherein,

T _Gx, T _Gy, T _GzThe component of correlation time on three of representing gyroscopic drift respectively, T _Ax, T _Ay, T _AzThe component of correlation time on three of representing accelerometer error respectively,

Wherein, f _EBe east orientation acceleration, f _NBe north orientation acceleration, f _UFor the sky to acceleration, ω _IeBe rotational-angular velocity of the earth, R _MBe radius of curvature of meridian, R _NBe radius of curvature in prime vertical;

Described INS/ESGM measurement equation is: Z ₁(t)=H ₁(t) X ₁(t)+V ₁(t);

Wherein, V ₁(t)=[m ₁, m ₂, m ₃, m ₄] ^TBe the white Gaussian noise of the zero-mean that produces by the site error of ESGM and azimuthal error, H ₁(t) be to measure battle array, observed quantity Z ₁(t) be the difference of ESGM and INS position, course information; INS/ESGM state equation X ₁(t)=X (t);

Described INS/GPS measurement equation is:

Wherein, position quantity is measured

Speed measurement amount Measure battle array

v ₂(t) the white Gaussian noise battle array of the zero-mean that produces by the site error of GPS and azimuthal error of expression, observed quantity Z ₂(t) be the difference of GPS and INS position and speed; L _GThe latitude of expression GPS, λ _GThe longitude of expression GPS, h _GThe height of expression GPS, V _GEEast orientation speed among the expression GPS, V _GNNorth orientation speed among the expression GPS, V _GUSky among the expression GPS is to speed; INS/GPS state equation X ₂(t)=X (t);

Described INS/DVL system state equation is:

${\stackrel{\&CenterDot;}{X}}_3\left(t\right)=\left[\begin{array}{c}\stackrel{\&CenterDot;}{X\left(t\right)}\\ \stackrel{\&CenterDot;}{X_d\left(t\right)}\end{array}\right]=\left[\begin{array}{cc}F\left(t\right)& \\ & F_d\left(t\right)\end{array}\right]\left[\begin{array}{c}X\left(t\right)\\ X_d\left(t\right)\end{array}\right]+\left[\begin{array}{cc}G\left(t\right)& \\ & G_d\left(t\right)\end{array}\right]\left[\begin{array}{c}W\left(t\right)\\ W_d\left(t\right)\end{array}\right];$

Wherein, the noise of Doppler range rate measuring system drives battle array G _d(t)=and diag[1 1 0], F _d(t) be the augmented state transfer matrix of Doppler range rate measuring system, W _d(t) be the augmentation process noise battle array of Doppler range rate measuring system, X _d(t) be Doppler range rate measuring system augmented state amount, X _d=[δ V _dδ Δ δ C] ^T, δ V _dBe the velocity shifts error in the Doppler range rate measuring system, the δ Δ is the drift angle error in the Doppler range rate measuring system, and δ C is the calibration factor error in the Doppler range rate measuring system;

Described INS/DVL measurement equation is:

Wherein, measure battle array

V ₃(t)=[m _VEm _VN] ^TBe observation white noise, V _DE, V _DNRepresent east orientation speed, north orientation speed in the Doppler range rate measuring system respectively, K ' is the angle, ship's head.

4. the federal filtering method of a kind of control with changed scale self-adaptation based on time series analysis according to claim 1 and 2 is characterized in that the discrete state spatial model of each the navigation sensor system correspondence described in the step 2 is:

$\left\{\begin{array}{c}{X}_k={\mathrm{\&Phi;}}_{k/k-1}{X}_{k-1}+{\mathrm{\&Gamma;}}_{k/k-1}{W}_{k-1}\\ Z_k=H_k{X}_k+V_k\end{array}\right.$

Wherein, k represents the index value of discrete time state, X _kExpression k quantity of state constantly, Z _kExpression k observed quantity constantly, H _kExpression k measurement battle array constantly, Γ _K/k-1Be that noise drives battle array, W _K-1Be k-1 system state noise constantly, and process noise covariance matrix Q _k=E[W _kW _k ^T], V _kBe k system measurements noise constantly, and measurement noise covariance matrix R _k=E[V _kV _k ^T], state transitions battle array Φ _K/k-1Adopt substep build up discrete method to obtain:

${\mathrm{\&Phi;}}_{k/k-1}=\underset{i-1}{\overset{J}{\mathrm{\&Pi;}}}{\mathrm{\&Phi;}}_{(t_{k-1}+\mathrm{is})/(t_{k-1}+(i-1)s)}\&ap;\underset{i-1}{\overset{J}{\mathrm{\&Pi;}}}(I+s{F}_{i-1})\&ap;I+s\underset{i=1}{\overset{J}{\mathrm{\&Sigma;}}}{F}_{i-1}$

J=T/s wherein, T is the sampling time, and s is the step-length of dividing in the sampling time, and I is a unit matrix, F _I-1It is i-1 state transitions battle array constantly.

5. according to claim 1 or 2 or the federal filtering method of described a kind of control with changed scale self-adaptation based on time series analysis, it is characterized in that step 3 is described obtains the information weights by autoregressive model, is specially:

At first, determine the autoregressive model of navigation sensor system:

Wherein, y (τ) is the output sequence of navigation sensor system, and τ represents τ point of sequences y (τ), and N represents the exponent number of autoregressive model, a _kThe representation model error is for average is zero, variance is σ ²White Gaussian noise,

Expression N rank autoregressive model coefficient;

Then, determine the autoregressive model coefficient by the Levinson-Durbin recursive algorithm:

Wherein,

Be N rank autoregressive model coefficient Estimated value, i=1,2 ... N,

Be N-1 rank autoregressive model coefficient Estimated value,

Be N-1 rank autoregressive model coefficient Estimated value,

Be the least error power of N rank autoregressive model of prediction, Be the least error power of N-1 rank autoregressive model of prediction,

Be intermediate parameters,

Be respectively capable the 1st element and i+1 element of N in the autocorrelation function battle array of output sequence y (τ) of navigation sensor;

Obtain exponent number by final predicated error rule:

$\mathrm{FPE}\left(N\right)={\widehat{\mathrm{\&rho;}}}_N\frac{\mathrm{\&Xi;}+(N+1)}{\mathrm{\&Xi;}-(N+1)}$

Wherein, Ξ is the length of measurement data,

Error power for N rank AR model;

Further obtain the predicted value of navigation sensor system output sequence:

With the predicated error of each navigation sensor system as information weights Θ _i:

${\mathrm{\&Theta;}}_i=y_i\left(\mathrm{\&tau;}\right)-{\hat{y}}_i\left(\mathrm{\&tau;}\right)(i=\mathrm{0,1,2,3})$

y _i(τ) be the output sequence of i navigation sensor,

Be the predicted value of the output sequence of i navigation sensor, i=0, the corresponding inertial navigation system of 1,2,3 difference, electrostatically supported gyro monitor, NAVSTAR and Doppler range rate measuring system.

6. the federal filtering method of a kind of control with changed scale self-adaptation according to claim 1 and 2 based on time series analysis, it is characterized in that described information distribution ratio of step 4 and information weights are directly proportional, and according to the information conservation law, the weight of each navigation sensor system and be

β wherein ₀Be main system weight, β _iBe each subsystem weight, n is the number of all subsystems, and described main system is an inertial navigation system, and subsystem is electrostatically supported gyro monitor, NAVSTAR and Doppler range rate measuring system, and the weight that obtains each navigation sensor system at last is:

${\mathrm{\&beta;}}_i=\frac{{\mathrm{\&Theta;}}_i}{\underset{i=0}{\overset{n}{\mathrm{\&Sigma;}}}{\mathrm{\&Theta;}}_i}$

Wherein, i=0, the corresponding inertial navigation system of 1,2,3 difference, electrostatically supported gyro monitor, NAVSTAR and Doppler range rate measuring system.

7. the federal filtering method of a kind of control with changed scale self-adaptation based on time series analysis according to claim 1 and 2 is characterized in that the described global optimum of step 5 estimates, specifically obtains by following process:

At first, obtain information distribution:

$P_{i,k-1}={\mathrm{\&beta;}}_i^{-1}{P}_{k-1}$

$Q_{i,k-1}={\mathrm{\&beta;}}_i^{-1}{Q}_{k-1}$

X _i，k-1＝X _k-1

Wherein, P _{I, k-1}Expression is with the estimation error variance battle array P of i navigation sensor system _K-1Passing ratio β _iDistribute the estimation error variance battle array after resetting, Q _{I, k-1}Expression is with the process noise covariance matrix Q of i navigation sensor system _K-1Passing ratio β _iDistribute the process noise covariance matrix after resetting, X _{I, k-1}Expression is with the quantity of state of i navigation sensor system, and k represents the index value of discrete time state; I=0, the corresponding inertial navigation system of 1,2,3 difference, electrostatically supported gyro monitor, NAVSTAR and Doppler range rate measuring system;

The time of carrying out again upgrades:

${\hat{X}}_{i,k/k-1}={\mathrm{\&Phi;}}_{i,k/k-1}{\hat{X}}_{i,k-1}$

$P_{i,k/k-1}={\mathrm{\&Phi;}}_{i,k/k-1}{P}_{i,k-1}{\mathrm{\&Phi;}}_{i,k/k-1}^T+{\mathrm{\&Gamma;}}_{i,k/k-1}{Q}_{i,k-1}{\mathrm{\&Gamma;}}_{i,k/k-1}^T$

Wherein,

Be the state one-step prediction battle array of i navigation sensor system, P _{I, k/k-1}Be the one-step prediction square error battle array of i navigation sensor system, Φ _{I, k/k-1}Be the state transitions battle array of i navigation sensor system, Γ _{I, k/k-1}The noise that is i navigation sensor system drives battle array,

Be quantity of state X _{I, k-1}Predicted value;

Observe renewal then:

$K_{i,k}=P_{i,k/k-1}{H}_{i,k}^T{(H_{i,k}{P}_{i,k/k-1}{H}_{i,k}^T+R_{i,k})}^{-1}$

${\hat{X}}_{i,k}={\hat{X}}_{i,k/k-1}+K_{i,k}(Z_{i,k}-H_{i,k}{\hat{X}}_{i,k/k-1})$

$P_{i,k}=(I-K_{i,k}{H}_{i,k})P_{i,k/k-1}{(I-K_{i,k}{H}_{i,k})}^T+K_{i,k}{R}_{i,k}{K}_{i,k}^T$

Wherein, K _{I, k}Be the filter gain of i navigation sensor system, H _{I, k}Be the measurement battle array of i navigation sensor system, R _{I, k}Be the system measurements noise covariance battle array of i navigation sensor system, Z _{I, k}It is the observed quantity of i navigation sensor system;

Obtaining global optimization at last estimates:

${\hat{X}}_g=P_g\underset{i=1}{\overset{m}{\mathrm{\&Sigma;}}}{P}_{i,k}^{-1}{\hat{X}}_{i,k}$

$P_g={\left(\underset{i=1}{\overset{m}{\mathrm{\&Sigma;}}}{P}_{i,k}^{-1}\right)}^{-1}$

Wherein,

Be overall estimated value, P _gFor estimating square error, m is a number of filter.

Images