Virtual beam-assisted inertial / Doppler tight integration navigation method under non-holonomic beam

Abstract

The invention provides a virtual beam-assisted inertial / Doppler tight integrated navigation method under an incomplete beam, which specifically comprises the following steps: step 1: establishing an inertial / acoustic Doppler velocimeter tight integrated system, comprising an inertial system error model, an acoustic Doppler velocimeter beam error model, a combined system state equation, and a combined system observation equation; step 2: when all four beams of the acoustic Doppler velocimeter are valid, constructing an artificial intelligence module based on a minimum support vector machine, taking the speed of each beam of the acoustic Doppler velocimeter as output, and training the artificial intelligence network; step 3: when a beam of the acoustic Doppler velocimeter fails, entering an intelligent prediction stage, using the artificial intelligence module that has completed training and learning to continuously predict the speed of the failed beam, and forming virtual beam information; step 4: combining the virtual beam information with the valid beam information as system observation, and continuously performing an inertial / Doppler tight integrated navigation algorithm.

Description

Technical Field

[0001] The invention belongs to the field of underwater combined navigation and belongs to multi-sensor information fusion technology; it relates to an artificial intelligence virtual beam (Virtual Beam) assisted inertial / Doppler tight combined navigation method under an incomplete beam. Background Art

[0002] More and more countries have made the promotion of ocean exploration and development a strategic goal, and accurate and stable underwater navigation systems are the cornerstone of ocean exploration missions. SINS has become the core navigation system for underwater navigation due to its autonomy and concealment, but the problem of error accumulation limits its independent use. Therefore, combining SINS with DVL for navigation has become the mainstream navigation method for underwater vehicles (AUVs) to achieve long-term, high-precision underwater navigation.

[0003] For the SINS / DVL tightly integrated navigation system, there are many factors that affect its navigation accuracy, such as installation angle error, arm error, and speed measurement error. After the SINS and DVL are fixed, the installation angle and arm error are relatively stable and can be calibrated before navigation. DVL is an active sonar that needs to receive reflected sound waves from the outside world. The acoustic signal it receives is closely related to the surrounding acoustic environment. Therefore, the actual speed measurement accuracy of DVL is greatly affected by the complex environment, which is the part that has the greatest impact on the accuracy of integrated navigation. When the AUV encounters seabed terrain obstruction or marine biological obstruction during navigation, part or even all of the DVL beam measurement data will fail, resulting in a decrease in long-term navigation accuracy. In order to solve the problem of high-precision SINS / DVL tightly integrated navigation when the DVL beam fails partially or completely, some scholars have proposed a DVL virtual beam construction method based on the carrier ZUPT assumption, assuming the carrier motion state and solving the failed DVL beam. Its disadvantage is that the carrier ZUPT assumption may not actually hold true, and there are special requirements for the failed DVL configuration. In view of the above drawbacks, a more versatile and robust virtual beam construction method is urgently needed.

[0004] As an improved support vector machine based on statistical theory, LS-SVM has an advanced and complete theoretical system, which can transform the solution of quadratic optimization problems into the solution of linear equations, thus simplifying the solution of the problem. Therefore, it has been successfully applied in many fields, including data regression, pattern recognition, time series prediction, etc. Summary of the invention

[0005] In order to solve the above problems, the present invention discloses a virtual beam-assisted inertial / Doppler tight integrated navigation method under non-complete beam; taking the SINS / DVL tight integrated navigation system as the research object, the minimum support vector machine is used to construct the DVL virtual beam, and combined with the effective DVL measurement value. Then the synthesized measurement data is combined with the corresponding data solved by inertial navigation for navigation, so as to obtain high-precision navigation information.

[0006] In order to achieve the above object, the present invention provides the following technical solutions:

[0007] The virtual beam-assisted inertial / Doppler tight integrated navigation method under non-holonomic beam specifically comprises the following steps:

[0008] Step 1: Establish an inertial / acoustic Doppler velocimeter compact combination system, including the inertial system error model, the acoustic Doppler velocimeter beam error model, the combined system state equation, and the combined system observation equation;

[0009] Step 2: When all four beams of the acoustic Doppler velocimeter are valid, an artificial intelligence module based on a minimum support vector machine is constructed, with the current inertial solution attitude and velocity of the carrier and the output of the gyro and accelerometer of the inertial measurement unit as input, and the velocity of each beam of the acoustic Doppler velocimeter as output, to train the artificial intelligence network;

[0010] Step 3: When the acoustic Doppler velocimeter has a beam failure, it enters the intelligent prediction stage. Based on the continuous inertial system attitude, speed, gyroscope, and accelerometer information, the trained artificial intelligence module is used to continuously predict the speed of the failed beam to form virtual beam information;

[0011] Step 4: Combine the virtual beam information with the effective beam information as system observation and continue to perform the inertial / Doppler tight integration navigation algorithm.

[0012] Furthermore, in step 1, a tightly combined inertial / acoustic Doppler velocimeter system is established, and the specific process is as follows:

[0013] Step 1.1: Establish the inertial system error model. The specific steps are as follows:

[0014] Given the initial attitude, velocity and position information, the SINS inertial system uses the gyroscope measurements to and the accelerometer measurement f b Update attitude, speed and position information:

[0015]

[0016]

[0017]

[0018]

[0019]

[0020]

[0021]

[0022]

[0023] Among them, b is the carrier coordinate system, e is the earth coordinate system, n is the navigation coordinate system, and i is the inertial coordinate system. is the transformation matrix from the b coordinate system to the n coordinate system, is the transformation matrix from the n coordinate system to the b coordinate system, V n is the velocity in the n coordinate system, and its velocity components in the east, north and sky directions are V E 、V N 、V U , L is latitude, λ is longitude, h is altitude, R M is the meridian radius, R N is the radius of the Maoyou circle, is the angular velocity of the earth's rotation in the n coordinate system, Ω is the angular velocity of the earth's rotation, is the angular velocity from the n coordinate system to the e coordinate system in the n coordinate system, is the angular velocity from the b coordinate system to the n coordinate system in the b coordinate system, G n is the gravity vector in the n coordinate system, (×) represents the skew-symmetric matrix, is the angular velocity from the b coordinate system to the i coordinate system in the b coordinate system, f b For comparison.

[0024] Accordingly, the SINS inertial system error model is established as follows:

[0025]

[0026]

[0027]

[0028]

[0029]

[0030] Among them, ε b is the gyroscope error, is the accelerometer error, φ is the misalignment angle, and δ represents the corresponding errors of each parameter.

[0031] Step 1.2: Establish the acoustic Doppler velocimeter beam error model. The specific steps are as follows:

[0032] According to the Doppler effect, we can get:

[0033] Δf=f0(V Trans +V Rec ) / (CV Trans )≈2Vf0 / C

[0034] Where Δf is the Doppler frequency shift, f0 is the transmitting beam frequency, C is the speed of sound in water, V Trans is the relative velocity in the beam direction when the beam is transmitted, V Rec is the relative speed in the beam direction when receiving the beam. Assuming that the relative speed when transmitting and receiving the beam at the same time remains unchanged, the equation can be simplified to the form on the right side of the equal sign.

[0035] Based on this, given the Doppler frequency shift, the acoustic Doppler velocimeter can measure the relative speed in the four beam directions. Solve:

[0036]

[0037] Where Δf i (i=1, 2, 3, 4) is the Doppler frequency shift measured by the i-th beam. The velocity in the d coordinate system The calculation is as follows:

[0038]

[0039] Where d is the orthogonal coordinate system aligned with the beam center of the DVL, and A is the transformation matrix. In the “+” installation mode, we have:

[0040]

[0041] Where α is the beam launch angle between the beam and the horizontal plane of the d coordinate system. In the "×" installation mode, there are:

[0042]

[0043] Wherein, β is the angle between the DVL beam 1 and the y-axis of the d-coordinate system in the “×” installation mode, and β=45°.

[0044] Therefore, the acoustic Doppler velocimeter beam error model is established as follows:

[0045]

[0046] Where K is the scale factor, δK is the scale factor error, θ is the installation error angle between the d coordinate system and the b coordinate system, is the DVL beam velocity zero bias, is the true value of velocity in the b coordinate system, and the superscript ~ indicates the measured value.

[0047] Step 1.3: Establish the state equation and observation equation of the combined system. The specific steps are as follows:

[0048] According to the SINS inertial system error model in step 1.1 and the Doppler velocimeter beam error model in step 1.2, select the state quantity to be estimated and establish the combined system state equation.

[0049] The observation equation of the combined system is established as follows:

[0050]

[0051]

[0052] The superscript ^ indicates a calculated value.

[0053] Furthermore, in step 2, an artificial intelligence module based on a minimum support vector machine is constructed, with the current inertial solution attitude and speed of the carrier and the output of the gyro and accelerometer of the inertial measurement unit as input, and the beam speed of the acoustic Doppler velocimeter as output, to train the artificial intelligence network, such as Figure 1 As shown, the specific steps include:

[0054] Step 2.1: Establish a training data set with the current inertial solution attitude and velocity of the carrier and the output of the gyro and accelerometer of the inertial measurement unit as input and the beam velocity of the acoustic Doppler velocimeter as output. The specific steps are as follows:

[0055] Given an input vector x k And the corresponding output vector yk, the training data set is established as follows:

[0056]

[0057] Among them, Att is the posture vector, and N is the number of data groups in the training data set.

[0058] Step 2.2: Establish an artificial intelligence module based on the minimum support vector machine. The specific steps are as follows:

[0059] Find a nonlinear function of the form:

[0060]

[0061] So that it meets the following optimization goals:

[0062]

[0063]

[0064] Among them, γ is the regularization parameter, s k is the weight of the kth group of training data.

[0065] Step 2.3: According to the training data set described in step 2.1, the artificial intelligence module described in step 2.2 is trained. The specific steps are as follows:

[0066] Solve the constrained optimization problem described in step 2.2 using the Lagrange multiplier method:

[0067]

[0068] Among them, α and α k is the Lagrange multiplier.

[0069] Let the partial derivative of the above formula be zero, and we get

[0070]

[0071]

[0072] Among them, l N =[1,…,1], S is the weight matrix, Ω is the kernel function matrix, and its elements can be expressed as

[0073] Select radial basis function as the kernel function, we have:

[0074]

[0075] Here, σ is the kernel width.

[0076] Therefore, the following results can be obtained:

[0077]

[0078]

[0079] That is, the training of the artificial intelligence module described in step 2.2 is completed.

[0080] Furthermore, in step 3, according to the continuous inertial system attitude, speed, gyroscope, and accelerometer information, the artificial intelligence module that has completed training and learning is used to continuously predict the speed of the failed beam to form virtual beam information. The specific process is as follows:

[0081] Step 3.1: Create an AI module input vector that includes continuous inertial system attitude, velocity, gyroscope, and accelerometer information. The specific steps are as follows:

[0082] According to the continuously autonomously running SINS inertial system and IMU measurements, the input vector x of the artificial intelligence module based on the minimum support vector machine described in step 2 can be continuously obtained. k :

[0083]

[0084] Step 3.2: Use the trained artificial intelligence module to continuously predict the failed beam velocity and form virtual beam information. The specific steps are as follows:

[0085] Based on the artificial intelligence module trained in step 2, the virtual beam information is predicted and formed by the following formula:

[0086]

[0087] Where x is the input vector of the training dataset described in step 2.

[0088] Furthermore, in step 4, the virtual beam information is combined with the effective beam information as system observation, and the inertial / Doppler tight integrated navigation algorithm is continuously performed, which specifically includes the following steps:

[0089] Step 4.1: Replace the invalid beam information with the virtual beam information described in step 3, and retain the valid beam information to form a complete Doppler velocimeter system observation.

[0090] Step 4.2: Using the complete Doppler velocimeter system observations described in step 4.1, input the inertial / acoustic Doppler velocimeter tight combination system described in step 1 for solution, and update the current attitude, velocity and position information.

[0091] Step 4.3: Repeat the above steps and continue to use the inertial / Doppler tight combination navigation algorithm until the navigation mission is completed.

[0092] Compared with the prior art, the present invention has the following advantages and beneficial effects:

[0093] 1. The minimum support vector machine is used to construct a virtual beam, which can be trained to more effectively identify the complex relationship between IMU measurements and DVL beams.

[0094] 2. There are no special requirements for the number and configuration of failed DVL beams, which improves the versatility of the invention.

[0095] 3. Make more effective use of the effective DVL beam measurement information and improve the navigation accuracy of the SINS / DVL tightly integrated navigation system when the DVL beam fails partially or completely.

[0096] 4. It is particularly suitable for the underwater navigation field of Strapdown Inertial Navigation System (SINS) / Doppler Velocity Log (DVL) tight combination in complex environments. BRIEF DESCRIPTION OF THE DRAWINGS

[0097] Figure 1 It is a principle diagram of the training process of the SINS / DVL tightly integrated navigation system provided by the present invention;

[0098] Figure 2 It is a principle diagram of the prediction process of the SINS / DVL tightly integrated navigation system provided by the present invention. DETAILED DESCRIPTION

[0099] The present invention is further explained below in conjunction with the accompanying drawings and specific embodiments. It should be understood that the following specific embodiments are only used to illustrate the present invention and are not used to limit the scope of the present invention. It should be noted that the words "front", "rear", "left", "right", "upper" and "lower" used in the following description refer to directions in the accompanying drawings, and the words "inner" and "outer" refer to directions toward or away from the geometric center of a specific component, respectively.

[0100] The application method of the virtual beam-assisted inertial / Doppler tight integrated navigation method under the non-holonomic beam of this embodiment in AUV navigation specifically includes the following steps:

[0101] Step 1: Establish an inertial / acoustic Doppler velocimeter compact combination system, including the inertial system error model, the acoustic Doppler velocimeter beam error model, the combined system state equation, and the combined system observation equation;

[0102] Step 1.1: Establish the inertial system error model. The specific steps are as follows:

[0103] Given the initial attitude, velocity and position information, the SINS inertial system uses the gyroscope measurements to and the accelerometer measurement f b Update attitude, speed and position information:

[0104]

[0105]

[0106]

[0107]

[0108]

[0109]

[0110]

[0111]

[0112] Among them, b is the carrier coordinate system, e is the earth coordinate system, n is the navigation coordinate system, and i is the inertial coordinate system. is the transformation matrix from the b coordinate system to the n coordinate system, is the transformation matrix from the n coordinate system to the b coordinate system, Vn is the velocity in the n coordinate system, and its velocity components in the east, north, and sky directions are V E 、V N 、V U , L is latitude, λ is longitude, h is altitude, R M is the meridian radius, R N is the radius of the Maoyou circle, is the angular velocity of the earth's rotation in the n coordinate system, Ω is the angular velocity of the earth's rotation, is the angular velocity from the n coordinate system to the e coordinate system in the n coordinate system, is the angular velocity from the b coordinate system to the n coordinate system in the b coordinate system, G n is the gravity vector in the n coordinate system, (×) represents the skew-symmetric matrix, is the angular velocity from the b coordinate system to the i coordinate system in the b coordinate system, f b For comparison.

[0113] Accordingly, the SINS inertial system error model is established as follows:

[0114]

[0115]

[0116]

[0117]

[0118]

[0119] Among them, ε b is the gyroscope error, is the accelerometer error, φ is the misalignment angle, and δ represents the corresponding errors of each parameter.

[0120] Step 1.2: Establish the acoustic Doppler velocimeter beam error model. The specific steps are as follows:

[0121] According to the Doppler effect, we can get:

[0122] Δf=f0(V Trans +VRec ) / (CV Trans )≈2Vf0 / C

[0123] Where Δf is the Doppler frequency shift, f0 is the transmitting beam frequency, C is the speed of sound in water, V Trans is the relative velocity in the beam direction when the beam is transmitted, V Rec is the relative speed in the beam direction when receiving the beam. Assuming that the relative speed when transmitting and receiving the beam at the same time remains unchanged, the equation can be simplified to the form on the right side of the equal sign.

[0124] Based on this, given the Doppler frequency shift, the acoustic Doppler velocimeter can measure the relative speed in the four beam directions. Solve:

[0125]

[0126] Where Δf i (i=1, 2, 3, 4) is the Doppler frequency shift measured by the i-th beam. The velocity in the d coordinate system The calculation is as follows:

[0127]

[0128] Where d is the orthogonal coordinate system aligned with the beam center of the DVL, and A is the transformation matrix. In the “+” installation mode, we have:

[0129]

[0130] Where α is the beam launch angle between the beam and the horizontal plane of the d coordinate system. In the "×" installation mode, there are:

[0131]

[0132] Wherein, β is the angle between the DVL beam 1 and the y-axis of the d-coordinate system in the “×” installation mode, and β=45°.

[0133] Therefore, the acoustic Doppler velocimeter beam error model is established as follows:

[0134]

[0135] Where K is the scale factor, δK is the scale factor error, θ is the installation error angle between the d coordinate system and the b coordinate system, is the DVL beam velocity zero bias, is the true value of velocity in the b coordinate system, and the superscript ~ indicates the measured value.

[0136] Step 1.3: Establish the state equation and observation equation of the combined system. The specific steps are as follows:

[0137] According to the SINS inertial system error model in step 1.1 and the Doppler velocimeter beam error model in step 1.2, select the state quantity to be estimated and establish the combined system state equation.

[0138] The observation equation of the combined system is established as follows:

[0139]

[0140]

[0141] The superscript ^ indicates a calculated value.

[0142] Step 2: When the four beams of the acoustic Doppler velocimeter are all valid, an artificial intelligence module based on the minimum support vector machine is constructed, with the current inertial solution attitude and speed of the carrier and the output of the gyro and accelerometer of the inertial measurement unit as input, and the speed of each beam of the acoustic Doppler velocimeter as output, and the artificial intelligence network is trained, such as Figure 1 As shown;

[0143] Step 2.1: Establish a training data set with the current inertial solution attitude and velocity of the carrier and the output of the gyro and accelerometer of the inertial measurement unit as input and the beam velocity of the acoustic Doppler velocimeter as output. The specific steps are as follows:

[0144] Given an input vector xk and the corresponding output vector yk, the training data set is established as follows:

[0145]

[0146] Among them, Att is the posture vector, and N is the number of data groups in the training data set.

[0147] Step 2.2: Establish an artificial intelligence module based on the minimum support vector machine. The specific steps are as follows:

[0148] Find a nonlinear function of the form:

[0149]

[0150] So that it meets the following optimization goals:

[0151]

[0152]

[0153] Among them, γ is the regularization parameter, s k is the weight of the kth group of training data.

[0154] Step 2.3: According to the training data set described in step 2.1, the artificial intelligence module described in step 2.2 is trained. The specific steps are as follows:

[0155] Solve the constrained optimization problem described in step 2.2 using the Lagrange multiplier method:

[0156]

[0157] Among them, α and α k is the Lagrange multiplier.

[0158] Let the partial derivative of the above formula be zero, and we get

[0159]

[0160]

[0161] Among them, l N =[1,…,1], S is the weight matrix, Ω is the kernel function matrix, and its elements can be expressed as

[0162] Select radial basis function as the kernel function, we have:

[0163]

[0164] Here, σ is the kernel width.

[0165] Therefore, the following results can be obtained:

[0166]

[0167]

[0168] That is, the training of the artificial intelligence module described in step 2.2 is completed.

[0169] Step 3: When the acoustic Doppler velocimeter has a beam failure, it enters the intelligent prediction stage. Based on the continuous inertial system attitude, speed, gyroscope, and accelerometer information, the trained artificial intelligence module is used to continuously predict the speed of the failed beam to form virtual beam information;

[0170] Step 3.1: Create an AI module input vector that includes continuous inertial system attitude, velocity, gyroscope, and accelerometer information. The specific steps are as follows:

[0171] According to the continuously autonomously running SINS inertial system and IMU measurements, the input vector x of the artificial intelligence module based on the minimum support vector machine described in step 2 can be continuously obtained. k :

[0172]

[0173] Step 3.2: Use the trained artificial intelligence module to continuously predict the failed beam velocity and form virtual beam information. The specific steps are as follows:

[0174] Based on the artificial intelligence module trained in step 2, the virtual beam information is predicted and formed by the following formula:

[0175]

[0176] Where x is the input vector of the training dataset described in step 2.

[0177] Step 4: Combine the virtual beam information with the effective beam information as system observation and continue to perform the inertial / Doppler tight integration navigation algorithm, such as Figure 2 As shown;

[0178] Step 4.1: Use the virtual beam information described in step 3 to replace the invalid beam information, and retain the valid beam information to combine into a complete Doppler velocimeter system observation, such as Figure 2 In this figure, it is assumed that beam 3 and beam 4 fail.

[0179] Step 4.2: Using the complete Doppler velocimeter system observations described in step 4.1, input the inertial / acoustic Doppler velocimeter tight combination system described in step 1 for solution, and update the current attitude, velocity and position information.

[0180] Step 4.3: Repeat the above steps and continue to use the inertial / Doppler tight combination navigation algorithm until the navigation mission is completed.

[0181] The technical means disclosed in the scheme of the present invention are not limited to the technical means disclosed in the above-mentioned implementation mode, but also include technical schemes composed of any combination of the above technical features.

Claims

1. A virtual beam-assisted inertial / Doppler tight integrated navigation method under non-holonomic beam, characterized in that: The method comprises the following steps: The method comprises the following steps: S1: Establishing a tightly combined inertial / acoustic Doppler velocimeter system, including an inertial system error model, an acoustic Doppler velocimeter beam error model, a combined system state equation, and a combined system observation equation; The tightly combined inertial / acoustic Doppler velocimeter system is established as described in S1, and the specific process is as follows: S1.1: Establish the inertial system error model. The specific steps are as follows: Given the initial attitude, velocity and position information, the SINS inertial system uses the gyroscope measurements to and the accelerometer measurement f b Update attitude, speed and position information: Among them, b is the carrier coordinate system, e is the earth coordinate system, n is the navigation coordinate system, and i is the inertial coordinate system. is the transformation matrix from the b coordinate system to the n coordinate system, is the transformation matrix from the n coordinate system to the b coordinate system, V n is the velocity in the n coordinate system, and its velocity components in the east, north and sky directions are V E 、V N 、V U , L is latitude, λ is longitude, h is altitude, R M is the meridian radius, R N is the radius of the Maoyou circle, is the angular velocity of the earth's rotation in the n coordinate system, Ω is the angular velocity of the earth's rotation, is the angular velocity from the n coordinate system to the e coordinate system in the n coordinate system, is the angular velocity from the b coordinate system to the n coordinate system in the b coordinate system, G n is the gravity vector in the n coordinate system, (×) represents the skew-symmetric matrix, is the angular velocity from the b coordinate system to the i coordinate system in the b coordinate system, f b For comparison; Accordingly, the SINS inertial system error model is established as follows: Among them, ε b is the gyroscope error, is the accelerometer error, φ is the misalignment angle, and δ represents the corresponding errors of each parameter; S1.2: Establish the acoustic Doppler velocimeter beam error model. The specific steps are as follows: According to the Doppler effect, we can get: Δf=f0(V Trans +V Rec ) / (CV Trans )≈2Vf0 / C Where Δf is the Doppler frequency shift, f0 is the transmitting beam frequency, C is the speed of sound in water, V Trans is the relative velocity in the beam direction when the beam is transmitted, V Rec is the relative speed in the beam direction when receiving the beam; assuming that the relative speed during the same transmission and reception of the beam remains unchanged, the equation can be simplified to the form on the right side of the equal sign; Based on this, given the Doppler frequency shift, the acoustic Doppler velocimeter can measure the relative speed in the four beam directions. Solve: Where Δf i (i=1,2,3,4) is the Doppler frequency shift measured by the i-th beam; d is the velocity in the coordinate system The calculation is as follows: Where d is the orthogonal coordinate system aligned with the beam center of the DVL, and A is the transformation matrix. In the "+" installation mode, we have: Where α is the beam launch angle between the beam and the horizontal plane of the d coordinate system; in the "×" installation mode: Wherein, β is the angle between DVL beam 1 and the y-axis of the d-coordinate system in the "×" installation mode, and β = 45°; Therefore, the acoustic Doppler velocimeter beam error model is established as follows: Where K is the scale factor, δK is the scale factor error, θ is the installation error angle between the d coordinate system and the b coordinate system, is the DVL beam velocity zero bias, is the true value of velocity in the b coordinate system, and the superscript ~ indicates the measured value; S1.3: Establish the state equation of the combined system and the observation equation of the combined system. The specific steps are as follows: According to the SINS inertial system error model in S1.1 and the Doppler velocimeter beam error model in S1.2, select the state quantity to be estimated and establish the state equation of the combined system; The observation equation of the combined system is established as follows: The superscript ^ indicates the calculated value. S2: When all four beams of the acoustic Doppler velocimeter are valid, an artificial intelligence module based on a minimum support vector machine is constructed, with the carrier's current inertial solution attitude and velocity and the output of the gyro and accelerometer of the inertial measurement unit as input, and the velocity of each beam of the acoustic Doppler velocimeter as output, to train the artificial intelligence module; S3: When the acoustic Doppler velocimeter has a beam failure, it enters the intelligent prediction stage. Based on the continuous inertial system attitude, speed, gyroscope, and accelerometer information, the trained artificial intelligence module is used to continuously predict the speed of the failed beam to form virtual beam information; S4: Combine the virtual beam information with the effective beam information as system observation and continue to perform the inertial / Doppler tight integration navigation algorithm.

2. The virtual beam-assisted inertial / Doppler tight integrated navigation method under non-holonomic beam according to claim 1, characterized in that: The architecture described in S2 is based on the artificial intelligence module of the minimum support vector machine, with the current inertial solution attitude and speed of the carrier and the output of the gyro and accelerometer of the inertial measurement unit as input, and the beam speed of the acoustic Doppler velocimeter as output, and the training of the artificial intelligence module specifically includes the following steps: S2.1: Establish a training data set with the current inertial solution attitude and velocity of the carrier and the output of the gyro and accelerometer of the inertial measurement unit as input and the beam velocity of the acoustic Doppler velocimeter as output. The specific steps are as follows: Given an input vector x k And the corresponding output vector y k , the training data set is established as follows: Among them, Att is the posture vector, N is the number of data groups in the training data set; S2.2: Establish an artificial intelligence module based on the minimum support vector machine. The specific steps are as follows: Find a nonlinear function with the following form: So that it meets the following optimization goals: Among them, γ is the regularization parameter, s k is the weight of the kth group of training data; S2.3: According to the training data set described in S2.1, the artificial intelligence module described in S2.2 is trained. The specific steps are as follows: Solve the constrained optimization problem of S2.2 using the Lagrange multiplier method: Among them, α and α k is the Lagrange multiplier; Let the partial derivative of the above formula be zero, and we get Among them, l N =[1,…,1], S is the weight matrix, Ω is the kernel function matrix, and its elements can be expressed as Select radial basis function as the kernel function, we have: Where, σ is the kernel width; Therefore, the following results can be obtained: That is, the training of the artificial intelligence module described in S2.2 is completed.

3. The virtual beam-assisted inertial / Doppler tight integrated navigation method under non-holonomic beam according to claim 1, characterized in that: As described in S3, based on the continuous inertial system attitude, speed, gyroscope, and accelerometer information, the artificial intelligence module that has completed training and learning is used to continuously predict the speed of the failed beam to form virtual beam information. The specific process is as follows: S3.1: Establish the AI ​​module input vector including the continuous inertial system attitude, velocity, gyroscope, and accelerometer information. The specific steps are as follows: According to the continuously autonomously running SINS inertial system and IMU measurements, the input vector x of the artificial intelligence module based on the minimum support vector machine described in S2 can be continuously obtained. k : S3.2: Use the trained artificial intelligence module to continuously predict the failed beam velocity and form virtual beam information. The specific steps are as follows: Based on the artificial intelligence module trained in S2, the virtual beam information is predicted and formed by the following formula Where x is the input vector of the training dataset described in S2.

4. The virtual beam-assisted inertial / Doppler tight integrated navigation method under non-holonomic beam according to claim 1, characterized in that: The virtual beam information and the effective beam information are combined as system observations in S4, and the inertial / Doppler tight integrated navigation algorithm is continuously performed, which specifically includes the following steps: S4.1: Replace the failed beam information with the virtual beam information described in S3, and retain the valid beam information to form a complete Doppler velocimeter system observation, assuming that beams 3 and 4 are failed; S4.2: Observe the complete Doppler velocimeter system described in S4.1, input the inertial / acoustic Doppler velocimeter tight combination system described in S1 for solution, and update the current attitude, velocity and position information; S4.3: Repeat the above steps and continue to perform the inertial / Doppler tight combination navigation algorithm until the navigation mission is completed.

Images