A generalized minimum error entropy Kalman method for underwater unmanned vehicle positioning

Through the generalized minimum error entropy Kalman filtering method, the non-Gaussian noise is processed by using the deformable shape kernel function, which solves the problem of low accuracy of the water acoustic positioning system in the marine environment, and improves the positioning accuracy and inertial navigation system of the unmanned aerial vehicle. performance.

CN115711622BInactive Publication Date: 2024-11-15UNIV OF ELECTRONICS SCI & TECH OF CHINA
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202211455385.1
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2022-11-21
Publication Date
2024-11-15
Estimated Expiration
Not applicable · inactive patent

AI Technical Summary

Technical Problem

The existing hydroacoustic positioning system has reduced positioning accuracy due to non-Gaussian noise interference in complex marine environments. The existing Kalman filtering algorithm cannot effectively process non-Gaussian noise, resulting in low estimation accuracy.

Method used

The generalized minimum error entropy Kalman filtering method is used to construct a generalized minimum error entropy Kalman filtering model, and use a deformable shape kernel function to process non-Gaussian noise of different distribution types to improve the positioning accuracy of underwater unmanned aerial vehicles.

Benefits of technology

It improves the accuracy of underwater navigation positioning, enhances the accuracy of the inertial navigation system, and can flexibly deal with different types of non-Gaussian noise, significantly improving the estimation accuracy of water acoustic positioning.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN115711622B_ABST
    Figure CN115711622B_ABST
Patent Text Reader

Abstract

The present invention provides a generalized minimum error entropy Kalman method for positioning an underwater unmanned vehicle, comprising the following steps: S1, collecting position and speed data of the underwater unmanned vehicle to construct a positioning data sample set; S2, using a generalized minimum error entropy Kalman filter to filter the positioning data samples in the positioning data sample set to obtain corrected positioning data; the present invention solves the problem that the existing filtering method for non-Gaussian noise can only process certain specific types of noise, and the non-Gaussian noise cannot be completely filtered out, resulting in low estimation accuracy of underwater acoustic positioning.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the technical field of unmanned vehicle positioning, and in particular to a generalized minimum error entropy Kalman positioning method for underwater unmanned vehicles. Background Art

[0002] With the development of ocean exploration, the application of underwater unmanned vehicles has become more and more extensive, and the high-precision hydroacoustic positioning technology of underwater unmanned vehicles is attracting more attention. In an actual marine environment, the hydroacoustic positioning system of underwater unmanned vehicles often uses sound wave signals for measurement and positioning. However, due to the complex underwater environment and the noise interference generated by passing ships, the hydroacoustic positioning system of underwater unmanned vehicles usually produces abnormal measurements. The noise generated by the marine environment and passing ships is generally non-Gaussian noise. These unknown distributions of non-Gaussian noise will seriously affect the accuracy of the hydroacoustic positioning system of the unmanned vehicle and have a great negative impact on the positioning of the underwater unmanned vehicle.

[0003] The original Kalman filter (KF) algorithm commonly used for underwater acoustic positioning is only applicable to Gaussian noise conditions. In order to solve the impact of non-Gaussian noise on state estimation (underwater acoustic positioning): (1) Recently, the maximum correlation entropy criterion (MCC) in information theoretic learning (ITL) takes into account high-order statistics and is a good non-Gaussian noise state estimation (underwater acoustic positioning) method. A new KF algorithm based on MCC, called maximum correlation entropy KF (MCKF), is proposed. This method is also extended to the state estimation (underwater acoustic positioning) of nonlinear systems. In addition, some KFs based on the modified correlation entropy criterion have been developed. (2) The minimum error entropy (MEE) criterion in ITL is superior to MCC in dealing with complex non-Gaussian noise with multi-peak distribution. In order to further improve the ability of the KF algorithm to deal with non-Gaussian noise, some new Kalman filter algorithms based on the MEE criterion are proposed. However, whether it is MCC or MEE, the shape of its error entropy cannot be changed freely because its kernel function is a Gaussian function, which makes the algorithms based on maximum correlation entropy and error entropy only able to deal with certain specific types of noise. These non-Gaussian noises with unknown distribution will inevitably reduce the estimation accuracy of the system's hydroacoustic positioning. Summary of the invention

[0004] In view of the above-mentioned deficiencies in the prior art, a generalized minimum error entropy Kalman positioning method for underwater unmanned vehicles provided by the present invention solves the problem that the existing filtering method for non-Gaussian noise can only process certain specific types of noise, and the non-Gaussian noise cannot be completely filtered out, resulting in low estimation accuracy of underwater acoustic positioning.

[0005] In order to achieve the above-mentioned invention object, the technical solution adopted by the present invention is: a generalized minimum error entropy Kalman underwater unmanned vehicle positioning method, comprising the following steps:

[0006] S1, collecting the position and speed data of the underwater unmanned vehicle and constructing a positioning data sample set;

[0007] S2. Use generalized minimum error entropy Kalman filtering to filter the positioning data samples in the positioning data sample set to obtain corrected positioning data.

[0008] Furthermore, the step S2 includes the following sub-steps:

[0009] S21, bringing the initial value of the positioning data sample into the prediction equation to obtain the prediction value of the positioning data sample;

[0010] S22, calculating a positioning data sample error value based on the positioning data sample prediction value and the positioning data sample concentrated positioning data sample;

[0011] S23, calculating the components of the covariance matrix of the augmented noise matrix according to the positioning data sample error value and the augmented noise matrix;

[0012] S24, constructing a generalized minimum error entropy Kalman filter model according to the components of the covariance matrix of the augmented noise matrix and the predicted values ​​of the positioning data samples;

[0013] S25, updating the estimated positioning data sample according to the generalized minimum error entropy Kalman filter model to obtain an updated estimated positioning data sample;

[0014] S26, judging whether the updated estimated positioning data sample meets the error condition, if so, updating the estimated positioning data sample to the corrected positioning data, if not, directly jumping to step S25.

[0015] Furthermore, the prediction equation in step S21 is:

[0016]

[0017] in, is the predicted value of the kth positioning data sample, A k-1 is the k-1th state transfer matrix, is the predicted value of the k-1th positioning data sample. When k=0, is the initial value of the positioning data sample.

[0018] Furthermore, the formula for calculating the positioning data sample error value in step S22 is:

[0019]

[0020] Among them, ε k|k-1 is the kth positioning data sample error value, x k is the kth positioning data sample in the positioning data sample set, is the predicted value of the kth positioning data sample.

[0021] Furthermore, the step S23 includes the following sub-steps:

[0022] S231, calculating an augmented noise matrix of the positioning data sample according to the positioning data sample error value;

[0023] S232. Calculate components of a covariance matrix of the augmented noise matrix according to the augmented noise matrix of the positioning data samples.

[0024] Furthermore, the formula for calculating the augmented noise matrix of the positioning data sample in step S231 is:

[0025]

[0026] Among them, μ k is the augmented noise matrix of the kth positioning data sample, ε k|k-1 is the kth positioning data sample error value, v k is the observation noise of the kth positioning data sample.

[0027] Furthermore, the formula for calculating the components of the covariance matrix of the augmented noise matrix in step S232 is:

[0028]

[0029] Among them, Θ k is the component of the covariance matrix of the augmented noise matrix of the kth positioning data sample, is the covariance matrix of the augmented noise matrix of the kth positioning data sample, T is the transpose operation, μ k is the augmented noise matrix of the kth positioning data sample.

[0030] Furthermore, the generalized minimum error entropy Kalman filter model in step S24 is:

[0031]

[0032] in,

[0033]

[0034]

[0035]

[0036]

[0037] Right D k , W k and e k Solving it, we get:

[0038]

[0039] L=m+n

[0040] Will Rewrite it as:

[0041]

[0042] Θ k Breaks down into:

[0043]

[0044] but:

[0045]

[0046]

[0047] in, is the estimated positioning data sample updated at the tth iteration, is the predicted value of the kth positioning data sample, is the Kalman gain of the predicted value of the kth positioning data sample, y k is the true measurement vector of the kth positioning data sample, C k is the observation transfer matrix of the kth positioning data sample, W k is the transformation matrix of the covariance matrix of the predicted value of the kth positioning data sample, is the sum of squares of the generalized Gaussian probability density function values ​​of the predicted values ​​of the k-th positioning data sample, is the weighted sum of the generalized Gaussian probability density function values ​​of the predicted values ​​of the k-th positioning data sample, is the generalized Gaussian probability density function value of the predicted value of the kth positioning data sample, for The element in the i-th row and j-th column of G α,β () is the Parzen window function, e j;k is the error matrix e of the predicted value of the kth positioning data sample k The jth element of i;k is the error matrix e of the predicted value of the kth positioning data sample kThe i-th element of , || is the absolute value, α is the shape parameter, sign() is the sign function, for The element in row i and column j in is the real number field, m is the column number of the matrix, n is the row number of the matrix, L is the number of the elements in the statistical matrix, is the error matrix e k The i-th element of is the error matrix e k The gth element of i;k is the transformation matrix D of the predicted value of the kth positioning data sample k The i-th element in w i;k The transformation matrix W is the covariance matrix of the predicted value of the kth positioning data sample k The i-th row matrix of is the estimated positioning data sample updated at the t-1th iteration, D k is the transformation matrix of the predicted value of the kth positioning data sample, x k is the kth positioning data sample in the positioning data sample set, e k is the error matrix of the predicted value of the kth positioning data sample, Θ k is the component of the covariance matrix of the augmented noise matrix of the kth positioning data sample, I m is a unit vector, is the field of real numbers, v k is the observation noise of the kth positioning data sample, T is the transposition operation, d 1;k is the transformation matrix D of the predicted value of the kth positioning data sample k The first element in d 2;k is the transformation matrix D of the predicted value of the kth positioning data sample k The second element in d L;k is the transformation matrix D of the predicted value of the kth positioning data sample k The Lth element in w 1;k The transformation matrix W is the covariance matrix of the predicted value of the kth positioning data sample k The first row of the matrix, w 2;k The transformation matrix W is the covariance matrix of the predicted value of the kth positioning data sample k The second row matrix, w i;k The transformation matrix W is the covariance matrix of the predicted value of the kth positioning data sample k The i-th row matrix, w L;k The transformation matrix W is the covariance matrix of the predicted value of the kth positioning data sample k The L-th row matrix of is the error matrix e of the predicted value of the kth positioning data samplek The first element of is the error matrix e of the predicted value of the kth positioning data sample k The second element of is the error matrix e of the predicted value of the kth positioning data sample k The Lth element of is the square sum matrix of the m-order Gaussian probability density function values ​​of the k-th positioning data sample prediction value, is the square sum matrix of the n-order Gaussian probability density function values ​​of the k-th positioning data sample prediction value, is the square sum matrix of the n×m dimensional generalized Gaussian probability density function values ​​of the k-th positioning data sample prediction value, is the square sum matrix of the m×n dimensional generalized Gaussian probability density function values ​​of the k-th positioning data sample prediction value, For the matrix The positive definite matrix of For the matrix The positive definite matrix, Θ q;k is θ k The first matrix block, Θ v;k is θ k The second matrix is ​​divided into blocks.

[0048] Furthermore, the error condition in step S26 is:

[0049]

[0050] in, is the estimated positioning data sample updated at the tth iteration, is the estimated positioning data sample updated at the t-1th iteration, |||| is the two-norm operation, and τ is the positive threshold.

[0051] The beneficial effects of the present invention are as follows: the present invention constructs a generalized minimum error entropy Kalman filter model, the kernel function shape of the model is variable, so that the generalized minimum error entropy Kalman filter model can flexibly handle non-Gaussian noise of different distribution types, thereby improving the accuracy of underwater navigation positioning, thereby improving the accuracy of the inertial navigation system. BRIEF DESCRIPTION OF THE DRAWINGS

[0052] Figure 1 It is a flowchart of a generalized minimum error entropy Kalman positioning method for underwater unmanned vehicles;

[0053] Figure 2 This is an experimental comparison chart. DETAILED DESCRIPTION

[0054] In order to make the purpose, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Generally, the components of the embodiments of the present invention described and shown in the drawings here can be arranged and designed in various different configurations.

[0055] like Figure 1 As shown, a generalized minimum error entropy Kalman positioning method for underwater unmanned vehicles includes the following steps:

[0056] S1, collecting the position and speed data of the underwater unmanned vehicle and constructing a positioning data sample set;

[0057] S2. Use generalized minimum error entropy Kalman filtering to filter the positioning data samples in the positioning data sample set to obtain corrected positioning data.

[0058] The step S2 comprises the following sub-steps:

[0059] S21, bringing the initial value of the positioning data sample into the prediction equation to obtain the prediction value of the positioning data sample;

[0060] The prediction equation in step S21 is:

[0061]

[0062] in, is the predicted value of the kth positioning data sample, A k-1 is the k-1th state transfer matrix, is the predicted value of the k-1th positioning data sample. When k=0, is the initial value of the positioning data sample.

[0063] S22, calculating a positioning data sample error value based on the positioning data sample prediction value and the positioning data sample concentrated positioning data sample;

[0064] The formula for calculating the positioning data sample error value in step S22 is:

[0065]

[0066] Among them, ε k|k-1 is the kth positioning data sample error value, x k is the kth positioning data sample in the positioning data sample set, is the predicted value of the kth positioning data sample.

[0067] S23, calculating the components of the covariance matrix of the augmented noise matrix according to the positioning data sample error value and the augmented noise matrix;

[0068] The step S23 comprises the following sub-steps:

[0069] S231, calculating an augmented noise matrix of the positioning data sample according to the positioning data sample error value;

[0070] The formula for calculating the augmented noise matrix of the positioning data sample in step S231 is:

[0071]

[0072] Among them, μ k is the augmented noise matrix of the kth positioning data sample, ε k|k-1 is the kth positioning data sample error value, v k is the observation noise of the kth positioning data sample.

[0073] S232. Calculate components of a covariance matrix of the augmented noise matrix according to the augmented noise matrix of the positioning data samples.

[0074] The formula for calculating the components of the covariance matrix of the augmented noise matrix in step S232 is:

[0075]

[0076] Among them, Θ k is the component of the covariance matrix of the augmented noise matrix of the kth positioning data sample, is the covariance matrix of the augmented noise matrix of the kth positioning data sample, T is the transpose operation, μ k is the augmented noise matrix of the kth positioning data sample.

[0077] S24, constructing a generalized minimum error entropy Kalman filter model according to the components of the covariance matrix of the augmented noise matrix and the predicted values ​​of the positioning data samples;

[0078] The generalized minimum error entropy Kalman filter model in step S24 is:

[0079]

[0080] in,

[0081]

[0082]

[0083]

[0084]

[0085] Right D k , W k and e k Solving it, we get:

[0086]

[0087] L=m+n (11)

[0088] Will Rewrite it as:

[0089]

[0090] Θ k Breaks down into:

[0091]

[0092] but:

[0093]

[0094]

[0095] in, is the estimated positioning data sample updated at the tth iteration, is the predicted value of the kth positioning data sample, is the Kalman gain of the predicted value of the kth positioning data sample, y k is the true measurement vector of the kth positioning data sample, C k is the observation transfer matrix of the kth positioning data sample, W k is the transformation matrix of the covariance matrix of the predicted value of the kth positioning data sample, is the sum of squares of the generalized Gaussian probability density function values ​​of the predicted values ​​of the k-th positioning data sample, is the weighted sum of the generalized Gaussian probability density function values ​​of the predicted values ​​of the k-th positioning data sample, is the generalized Gaussian probability density function value of the predicted value of the kth positioning data sample, for The element in the i-th row and j-th column of G α,β () is the Parzen window function, e j;k is the error matrix e of the predicted value of the kth positioning data sample k The jth element of i;k is the error matrix e of the predicted value of the kth positioning data sample kThe i-th element of , || is the absolute value, α is the shape parameter, sign() is the sign function, for The element in row i and column j in is the real number field, m is the column number of the matrix, n is the row number of the matrix, L is the number of the elements in the statistical matrix, is the error matrix e k The i-th element of is the error matrix e k The gth element of i;k is the transformation matrix D of the predicted value of the kth positioning data sample k The i-th element in w i;k The transformation matrix W is the covariance matrix of the predicted value of the kth positioning data sample k The i-th row matrix of is the estimated positioning data sample updated at the t-1th iteration, D k is the transformation matrix of the predicted value of the kth positioning data sample, x k is the kth positioning data sample in the positioning data sample set, e k is the error matrix of the predicted value of the kth positioning data sample, Θ k is the component of the covariance matrix of the augmented noise matrix of the kth positioning data sample, I m is a unit vector, is the field of real numbers, v k is the observation noise of the kth positioning data sample, T is the transposition operation, d 1;k is the transformation matrix D of the predicted value of the kth positioning data sample k The first element in d 2;k is the transformation matrix D of the predicted value of the kth positioning data sample k The second element in d L;k is the transformation matrix D of the predicted value of the kth positioning data sample k The Lth element in w 1;k The transformation matrix W is the covariance matrix of the predicted value of the kth positioning data sample k The first row of the matrix, w 2;k The transformation matrix W is the covariance matrix of the predicted value of the kth positioning data sample k The second row matrix, w i;k The transformation matrix W is the covariance matrix of the predicted value of the kth positioning data sample k The i-th row matrix, w L;k The transformation matrix W is the covariance matrix of the predicted value of the kth positioning data sample k The L-th row matrix of is the error matrix e of the predicted value of the kth positioning data samplek The first element of is the error matrix e of the predicted value of the kth positioning data sample k The second element of is the error matrix e of the predicted value of the kth positioning data sample k The Lth element of is the square sum matrix of the m-order Gaussian probability density function values ​​of the k-th positioning data sample prediction value, is the square sum matrix of the n-order Gaussian probability density function values ​​of the k-th positioning data sample prediction value, is the square sum matrix of the n×m dimensional generalized Gaussian probability density function values ​​of the k-th positioning data sample prediction value, is the square sum matrix of the m×n dimensional generalized Gaussian probability density function values ​​of the k-th positioning data sample prediction value, For the matrix The positive definite matrix of For the matrix The positive definite matrix, Θ q;k is θ k The first matrix block, Θ v;k is θ k The second matrix is ​​divided into blocks.

[0096] S25, updating the estimated positioning data sample according to the generalized minimum error entropy Kalman filter model to obtain an updated estimated positioning data sample;

[0097] S26, judging whether the updated estimated positioning data sample meets the error condition, if so, updating the estimated positioning data sample to the corrected positioning data, if not, directly jumping to step S25.

[0098] The error condition in step S26 is:

[0099]

[0100] in, is the estimated positioning data sample updated at the tth iteration, is the estimated positioning data sample updated at the t-1th iteration, |||| is the two-norm operation, and τ is the positive threshold.

[0101] The generalized Gaussian kernel function of the generalized minimum error entropy Kalman filter model proposed in the present invention is formula (18):

[0102]

[0103]

[0104] Among them, G σ (e) is the ordinary Gaussian kernel function, σ is the kernel bandwidth, exp() is the exponential function, e is the natural constant, Γ() is the gamma function, α is the shape parameter, β is the bandwidth range parameter, G α,β (e) is the generalized Gaussian kernel function, α>0, β>0.

[0105] According to formula (18), when the value of the shape parameter α changes, the shape of the generalized Gaussian kernel function will change significantly, so that the generalized minimum error entropy Kalman filter model can flexibly handle noises of different non-Gaussian distributions, thereby improving the accuracy of underwater navigation positioning estimation. When the shape parameter is set to 1 or 2, the distribution of the generalized Gaussian kernel function becomes a Laplace distribution or a Gaussian distribution.

[0106] In the generalized minimum error entropy criterion, the error e is measured by Renyi's entropy:

[0107]

[0108] Where, μ is the order of Renyi entropy; V μ (e) is information potential energy, which is in the form of:

[0109] V μ (e) = ∫p μ (x)dx=E[p μ-1 (e)], (20)

[0110] Among them, p μ () is the probability density function, p μ-1 (e) is the probability density of the function, x is the variable, e is the error, and E[] represents the mathematical expectation. In practical applications, the Parzen window method can be used to roughly estimate the probability density function:

[0111]

[0112] in, is a rough estimate of the probability density function, L is the number of error data, G α,β () is the Parzen window function, e i is the ith error data.

[0113] When the order of information potential is 2, we can get:

[0114]

[0115] in, is the information potential energy value of the second order, L is the number of error data, for e iA rough estimate of the probability density function, e i is the ith error data, e j is the ith error data.

[0116] experiment:

[0117] Under the condition of mixed Gaussian noise, the generalized minimum error entropy Kalman filter model (GMEEKF) constructed in the present invention is compared with the Kalman filter algorithm (KF), the maximum correlation entropy Kalman filter algorithm (MCKF), the MEEKF algorithm, and the robust student's t-based Kalman filter algorithm (RSTKF):

[0118] The mixed Gaussian noise model is as follows:

[0119]

[0120] Among them, r is the mixed Gaussian noise, λ is the weighting coefficient, is a Gaussian distribution with mean a and variance μ1, is a Gaussian distribution with mean a and variance μ2.

[0121] If its distribution obeys equation (23), then it is said to obey a mixed Gaussian noise distribution, that is, r~M(λ,a,μ1,μ2).

[0122] In the following experiment, a uniform velocity trajectory tracking model is considered, and the positioning data sample equation and trajectory equation are as follows:

[0123]

[0124]

[0125] Among them, x k =[x 1;k x 2;k x 3;k x 4;k ] T , x 1;k , x 2;k Respectively represent the displacement information in the x-axis and y-axis directions, x 3;k , x 4;k Respectively represents the speed information in the x-axis and y-axis directions; time interval ΔT = 0.1 second; q 1;k-1 is the first row element of the state noise matrix, q 2;k-1 is the second row element of the state noise matrix, q 3;k-1 is the third row element of the state noise matrix, q 4;k-1 is the fourth row element of the state noise matrix; the covariance matrix of the state noise Initial value x0~N(0,I m ), P 0|0 ~N(x0,I m ), positive threshold τ=10 -6 , (I m is an m-order identity matrix), x0 is x k The initial value of x k is the kth positioning data sample in the positioning data sample set, x0~N(0,I m ) means that x0 has a mean of 0 and a covariance matrix of I m Gaussian process (normal process), for The initial value of P 0|0 P k|k The initial value of express The mean is 0 and the covariance matrix is ​​I m Gaussian process (normal process), P 0|0 ~N(x0,I m ) indicates P 0|0 From the mean of 0, the covariance matrix is ​​I m Gaussian process (normal process).

[0126] Consider the process noise to be Gaussian distribution q k ~N(0,0.01), the observation noise is: v k ~M(0.9, 0, 0.01, 100), where σ, χ, η, and N are the state parameters of the RSTKF algorithm; the true value x and estimated value of the state variable are obtained. The simulation results of the mean square error are as follows Figure 2 shown.

[0127] Depend on Figure 2 It can be seen that the generalized minimum error entropy Kalman filter model GMEEKF performs best in mixed Gaussian noise and has the highest accuracy in positioning underwater unmanned vehicles.

[0128] In summary, the effect of the embodiment of the present invention is: according to formula (18), it can be seen that the values ​​of α and β in the present invention can be changed, so the shape of the generalized Gaussian kernel function can be flexibly changed, so that the generalized minimum error entropy Kalman filter model can flexibly handle non-Gaussian noise of different distribution types, thereby improving the accuracy of underwater navigation positioning.

[0129] The above are only preferred embodiments of the present invention and are not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A generalized minimum error entropy Kalman positioning method for underwater unmanned vehicles, characterized in that: The following steps are involved: S1, collecting the position and speed data of the underwater unmanned vehicle and constructing a positioning data sample set; S2, use the generalized minimum error entropy Kalman filter to filter the positioning data samples in the positioning data sample set to obtain the corrected positioning data The step S2 comprises the following sub-steps: S21, bringing the initial value of the positioning data sample and the minimum mean square error matrix of the initial value of the positioning data sample into the prediction equation to obtain the predicted value of the positioning data sample and the minimum mean square error matrix of the predicted value; S22, calculating a positioning data sample error value based on the positioning data sample prediction value and the positioning data sample concentrated positioning data sample; S23, calculating the components of the covariance matrix of the augmented noise matrix according to the positioning data sample error value and the augmented noise matrix; S24, constructing a generalized minimum error entropy Kalman filter model according to the components of the covariance matrix of the augmented noise matrix, the positioning data sample prediction value and the minimum mean square error matrix of the prediction value; S25, updating the estimated positioning data sample according to the generalized minimum error entropy Kalman filter model to obtain an updated estimated positioning data sample; S26, judging whether the updated estimated positioning data sample meets the error condition, if so, updating the estimated positioning data sample to the corrected positioning data, if not, updating the error covariance matrix, and jumping to step S25; The generalized minimum error entropy Kalman filter model in step S24 is: in, Right D k , W k and e k Solving it, we get: L=m+n Rewrite it as: Θ k Breaks down into: but: in, is the estimated positioning data sample updated at the tth iteration, is the predicted value of the kth positioning data sample, is the Kalman gain of the predicted value of the kth positioning data sample, y k is the true measurement vector of the kth positioning data sample, C k is the observation transfer matrix of the k-th positioning data sample, Wk is the transformation matrix of the covariance matrix of the predicted value of the k-th positioning data sample, is the sum of squares of the generalized Gaussian probability density function values ​​of the predicted values ​​of the k-th positioning data sample, is the weighted sum of the generalized Gaussian probability density function values ​​of the predicted values ​​of the k-th positioning data sample, is the generalized Gaussian probability density function value of the predicted value of the kth positioning data sample, for The element in the i-th row and j-th column of G α,β ( ) is the Parzen window function, e j;k is the error matrix e of the predicted value of the kth positioning data sample k The jth element of i;k is the error matrix e of the predicted value of the kth positioning data sample k The i-th element of , | | is the absolute value, α is the shape parameter, sign() is the sign function, for The element in row i and column j in is the real number field, m is the column number of the matrix, n is the row number of the matrix, L is the number of the elements in the statistical matrix, is the error matrix e k The i-th element of is the error matrix e k The gth element of i;k is the transformation matrix D of the predicted value of the kth positioning data sample k The i-th element in w i;k The transformation matrix W is the covariance matrix of the predicted value of the kth positioning data sample k The i-th row matrix of is the estimated positioning data sample updated at the t-1th iteration, D k is the transformation matrix of the predicted value of the kth positioning data sample, x k is the kth positioning data sample in the positioning data sample set, e k is the error matrix of the predicted value of the kth positioning data sample, Θ k is the component of the covariance matrix of the augmented noise matrix of the kth positioning data sample, I m is a unit vector, v k is the observation noise of the kth positioning data sample, T is the transposition operation, d 1;k is the transformation matrix D of the predicted value of the kth positioning data sample k The first element in d 2;k is the transformation matrix D of the predicted value of the kth positioning data sample k The second element in d L;k is the transformation matrix D of the predicted value of the kth positioning data sample k The Lth element in w 1;k The transformation matrix W is the covariance matrix of the predicted value of the kth positioning data sample k The first row of the matrix, w 2;k The transformation matrix W is the covariance matrix of the predicted value of the kth positioning data sample k The second row matrix, w i;k The transformation matrix W is the covariance matrix of the predicted value of the kth positioning data sample k The i-th row matrix, w L;k The transformation matrix W is the covariance matrix of the predicted value of the kth positioning data sample k The L-th row matrix of is the error matrix e of the predicted value of the kth positioning data sample k The first element of is the error matrix e of the predicted value of the kth positioning data sample k The second element of is the error matrix e of the predicted value of the kth positioning data sample k The Lth element of is the square sum matrix of the m-order Gaussian probability density function values ​​of the k-th positioning data sample prediction value, is the square sum matrix of the n-order Gaussian probability density function values ​​of the k-th positioning data sample prediction value, is the square sum matrix of the n×m dimensional generalized Gaussian probability density function values ​​of the k-th fixed 15-bit data sample prediction value, is the square sum matrix of the m×n dimensional generalized Gaussian probability density function values ​​of the k-th positioning data sample prediction value, For the matrix The positive definite matrix of For the matrix The positive definite matrix, Θ q;k is the first matrix block of Θk, Θ v;k is θ k The second matrix is ​​divided into blocks.

2. The underwater unmanned vehicle positioning method based on generalized minimum error entropy Kalman according to claim 1 is characterized in that: The prediction equation in step S21 is: in, is the predicted value of the kth positioning data sample, A k-1 is the k-1th state transfer matrix, is the predicted value of the k-1th positioning data sample. When k=0, is the initial value of the positioning data sample.

3. The underwater unmanned vehicle positioning method of generalized minimum error entropy Kalman according to claim 1 is characterized in that: The formula for calculating the positioning data sample error value in step S22 is: Among them, ε k|k-1 is the kth positioning data sample error value, xk 为 The kth positioning data sample in the positioning data sample set, is the predicted value of the kth positioning data sample.

4. The underwater unmanned vehicle positioning method based on generalized minimum error entropy Kalman according to claim 1 is characterized in that: The step S23 comprises the following sub-steps: S231, calculating an augmented noise matrix of the positioning data sample according to the positioning data sample error value; S232. Calculate components of a covariance matrix of the augmented noise matrix according to the augmented noise matrix of the positioning data samples.

5. The underwater unmanned vehicle positioning method of generalized minimum error entropy Kalman according to claim 4 is characterized in that: The formula for calculating the augmented noise matrix of the positioning data sample in step S231 is: Among them, μ k is the augmented noise matrix of the kth positioning data sample, ε k|k-1 is the kth positioning data sample error value, v k is the observation noise of the kth positioning data sample.

6. The underwater unmanned vehicle positioning method of generalized minimum error entropy Kalman according to claim 4 is characterized in that: The formula for calculating the components of the covariance matrix of the augmented noise matrix in step S232 is: Among them, Θ k is the component of the covariance matrix of the augmented noise matrix of the kth positioning data sample, is the covariance matrix of the augmented noise matrix of the kth positioning data sample, T is the transpose operation, μ k is the augmented noise matrix of the kth positioning data sample.

7. The underwater unmanned vehicle positioning method of generalized minimum error entropy Kalman according to claim 1 is characterized in that: The error condition in step S26 is: in, is the estimated positioning data sample updated at the tth iteration, is the estimated positioning data sample updated at the t-1th iteration, || || is the two-norm operation, and τ is the positive threshold.