CN110597273A - Dead reckoning method based on motor propulsion model - Google Patents

Abstract

The invention relates to a positioning navigation method, in particular to a dead reckoning method based on motor propulsion model assistance. According to the method, data of a GPS, a DVL, an ADCP and an attitude sensor are collected, a model of the rotating speed and the navigational speed of the AUV relative to seawater is established by using a motor propulsion model, the model precision is improved based on the training of a fuzzy support vector machine, when the DVL data fails, the AUV speed is calculated by depending on the rotating speed of the motor and the motor propulsion model, the AUV is assisted to continue to carry out dead reckoning navigation, and the environment adaptability of the AUV and the navigation robustness are improved.

Description

Dead reckoning method based on motor propulsion model

Technical Field

The invention relates to a positioning navigation method, in particular to a dead reckoning method based on motor propulsion model assistance.

Background

While the development of human beings and society, three difficult topics of population, resources and environment are in front of all human beings. Due to abundant biological resources and mineral resources in the ocean, people turn attention to the development of the ocean in order to further develop their living space. Because the noise is small, the target can be conveniently approached, and the AUV is an ideal measuring platform. As an important means for human exploration and use of the ocean, people have paid more and more attention to the application and development of autonomous underwater robots (AUVs).

One of the key factors for determining whether the AUV can safely operate and return is the accuracy of the AUV navigation system. The application of Inertial Navigation Systems (INS) to AUVs has the following disadvantages: firstly, INS has the problem that the error constantly increases along with time, is difficult to satisfy the requirement of its precision to the AUV of long-time work, secondly with high costs, bulky. In view of the above, the use of INS in small AUVs is limited, whereas the conventional dead reckoning method is more suitable for AUVs. Dead reckoning methods rely on attitude data and velocity data collected by a Doppler Velocimeter (DVL). The depth range of the AUV exceeding the DVL can occur in the deep sea navigation process, and the measured data is invalid, so that great errors are generated in the dead reckoning method.

An Acoustic Doppler Current Profiler (ADCP) is an advanced Acoustic Current Profiler. The instrument can accurately measure the flow velocity and the flow direction of water flow according to the Doppler frequency shift effect. Therefore, according to the motor propulsion model and the ADCP, the dead reckoning method based on motor propulsion model assistance is designed, and the AUV navigation precision and robustness are improved.

Disclosure of Invention

The invention aims to provide a dead reckoning method based on motor propulsion model assistance.

In order to realize the purpose of the invention, the technical scheme is as follows:

a dead reckoning method based on motor propulsion model assistance comprises the following specific steps:

(1) obtaining and processing AUV position, speed, attitude information and propeller propulsion speed;

(2) establishing a motor propulsion model of the water flow relative to the AUV speed relative to the propeller rotation speed;

(3) optimizing a motor propulsion model based on a fuzzy support vector machine;

(4) and calculating the position of the AUV according to the speed information and the attitude information.

AUV position, speed, attitude information and propeller propulsion speed obtain and process:

acquiring initial position (x) of AUV navigation process through GPS₀，y₀) Wherein x is₀Is the initial position longitude, y₀Is the initial position latitude; when the AUV moves horizontally and the data is judged to be valid through the DVL message information in the DVL measuring range, a group of time series { (u) } is formed by using the AUV navigation process heading speed u acquired by the DVL₁，t₁)，(u₂，t₂)，…，(u_n，t_n) Acquiring a yawing angle phi in the AUV navigation process through an attitude sensor, and calculating through a traditional dead reckoning method;

meanwhile, recording the current propeller propelling rotation speed n of the AUV by using an electromechanical state monitoring system of the AUV body; measuring the velocity of water flow relative to AUV by a flow profiler { (u)_w1，t₁)，(u_w2，t₂)，…，(u_wn，t_n) J, the absolute velocity u of the current water flow_{w is to the bottom}Can be represented as u_{w is to the bottom}＝E(u_i-u_wi)。

The method is characterized in that a motor propulsion model of water flow relative to AUV speed relative to propeller rotation speed is established:

assuming a flow velocity u_{w is to the bottom}The value and the direction of the AUV do not change along with the position of a time point and a space point, the change of the movement speed of the AUV is mainly influenced by the thrust of the propeller, and when the AUV sails stably, the effective thrust provided by the propeller and the total resistance of the ship reach balance, namely:

where ρ isIs the density of water, u_wThe navigational speed of AUV relative to sea water, omega is AUV wet surface area, zeta is total resistance coefficient, which is a constant in general_pIs the thrust derating coefficient, ρ is the density of water, w_pTo wake factor, D_pIs the diameter of the propeller; tau is_pThe method is related to factors such as the appearance, the size and the load of the AUV, the installation position of the AUV and the like, and is generally determined according to the self-propulsion test or the empirical formula of the AUV; k₀、K₁And K₂Dimensionless thrust K to describe a propeller_PAnd dimensionless resistive torque K_MAccording to the propeller test result, determining through curve fitting;

in the formula

A training model of AUV electromechanical propulsion is provided:

the optimized motor propulsion model based on the fuzzy support vector machine is as follows:

for sequence { (u)_w1，n₁)，(u_w2，n₂)，…，(u_wn，n_n) H, data u is divided into_wiGenerating a symmetric triangular blur number is marked as A_i＝(a_i-δ_i，a_i，a_i+δ_i)；

Mapping the speed features to a high-order feature space, and then performing approximate linear regression in the high-dimensional feature space, wherein the training set is as follows:

Tr：{(Φ(n₁)，A₁)，(Φ(n₂)，A₂)，…，(Φ(n_l)，A_n)}

wherein y is_i＝i，(i＝1，2…n)；

Respectively increasing epsilon and decreasing epsilon (0 < epsilon) to the y value of each training point in the training set Tr to obtain two sets of positive class points and negative class points, and respectively recording the two sets as D⁺And D^-

D⁺：{((Φ(n₁)，A₁+ε)；1)，((Φ(n₂)，A₂+ε)；1)，…，((Φ(n_n)，A_n+ε)；1)}

D^-：{((Φ(n₁)，A₁-ε)；-1)，((Φ(n₂)，A₂-ε)；-1)，…，((Φ(n_n)，A_n-ε)；-1)}

Carrying out classification training of a support vector machine on the processed data to obtain a fuzzy optimal hyperplane (omega phi (n)) + eta A + b which is 0 in the fuzzy classification problem, wherein omega is (omega)₁，…，ω_n)^TIs a blur vector, b is a blur number, i＝1，2，…，n；the fuzzy optimal classification hyperplane (omega. phi (n)) + eta A + b ═ 0 problem is converted into the solution of the opportunistic constraint programming with fuzzy decision:

solving the opportunity constraint planning with fuzzy decision to obtain a fuzzy optimal solution (omega, eta, b), wherein omega is a fuzzy vector formed by triangular fuzzy numbers, eta is a real number, and b is the triangular fuzzy numbers;

(ω · Φ (n)) + η a + b ═ 0Then obtain A + (omega)^*·Φ(n))+b^*0, whereinWherein ω is^*A blur vector formed by triangular blur numbers, b^*Is a triangular fuzzy number, A is a triangular fuzzy number; make the triangle fuzzy number (omega)^*·Φ(x))+b^*Is C thenThe speed can be obtained by solving A by C + A as 0Taking the fuzzy center delta a as the speed u_w；

At the moment, the speed u of the water flow relative to the AUV measured by the rotating speed n and the flow velocity profiler is established_wThe model of (2), the model is described as:

combining:

the following can be obtained:

and the AUV position estimation is carried out according to the speed information and the attitude information:

when the DVL information is invalid, an AUV motor state monitoring system is used for obtaining propeller propelling rotation speed n, and the speed u 'of water flow relative to the AUV corresponding to the rotation speed n is obtained through calculation of a support vector machine model'_wThe ground speed of the AUV is: u '═ u'_w+u_{w is to the bottom}The AUV position updated last before the DVL failure is (x'₀，y′₀) (x ') if the system sampling period T is constant and the period is high enough, and the AUV makes uniform motion in the sampling period'₀，y′₀) The coordinate positions of (a) are:

by analogy to this, (x'_k，y′_k) The coordinate positions of (a) are:

the motor propulsion model and the fuzzy support vector machine are applied to the dead reckoning method, when the DVL fails or breaks down, the AUV can be assisted to continue dead reckoning navigation, the influence of isolated points or outliers in a data set on a dead reckoning algorithm is reduced, and the method has the advantages of high learning rate and strong generalization capability of the support vector machine and the robustness advantage of fuzzy regression.

Drawings

FIG. 1 is a flow chart of the present invention;

FIG. 2 is a flow chart of fuzzy support vector machine modeling.

Detailed description of the invention

The invention is further described below with reference to the accompanying drawings:

the invention relates to a positioning navigation method, in particular to a dead reckoning method based on motor propulsion model assistance.

The invention aims to provide a dead reckoning method based on motor propulsion model assistance.

In order to realize the purpose of the invention, the technical scheme is as follows:

1. acquiring initial position (x) of AUV navigation process through GPS₀，y₀) Wherein x is₀Is the initial position longitude, y₀The initial position latitude. When the AUV moves horizontally and the data is judged to be valid through the DVL message information in the DVL measuring range, a group of time series { (u) } is formed by using the AUV navigation process heading speed u acquired by the DVL₁，t₁)，(u₂，t₂)，…，(u_n，t_n) And acquiring a yawing angle phi in the AUV navigation process through an attitude sensor, and calculating through a traditional dead reckoning method.

And meanwhile, recording the current propeller propelling rotation speed n of the AUV by using an electromechanical state monitoring system of the AUV body. Measuring the velocity of water flow relative to AUV by a flow profiler { (u)_w1，t₁)，(u_w2，t₂)，…，(u_wn，t_n) J, the absolute velocity u of the current water flow_{w is to the bottom}Can be represented as u_{w is to the bottom}＝E(u_i-u_wi) I is 1, 2 … n and input to the electromechanical propulsion model of the AUV.

2. Using speed information u_wThe propeller propulsion speed n is used for establishing an AUV speed model based on a fuzzy support vector machine;

(1) using speed information u_wEstablishing AUV electromechanical propulsion training model by propeller propulsion speed n

The AUV needs to overcome the resistance of water when navigating forward, including frictional resistance, shape resistance and wave making resistance, and the total resistance can be expressed as:

where ρ is the density of water, u_wThe navigation speed of AUV relative to seawater, omega is AUV wet surface area, zeta is total resistance coefficient, and the total resistance coefficient zeta is a constant under general conditions.

Assuming a flow velocity u_{w is to the bottom}The value and direction of (a) do not change with the position of the point in time and space. The change of the motion speed of the AUV is mainly influenced by the thrust of the propeller, and when the AUV sails stably, the effective thrust provided by the propeller and the total resistance of the ship reach balance, namely:

in the formula tau_pIs the thrust derating coefficient, rho is the seawater density, w_pTo wake factor, D_pIs the diameter of the propeller. Tau is_pExternal to AUVThe shape, propeller size, load and mounting position of the AUV are related, and are generally determined according to the self-propulsion test or empirical formula of the AUV. K₀、K₁And K₂Dimensionless thrust K to describe a propeller_PAnd dimensionless resistive torque K_MAnd the test result can be determined by curve fitting according to the propeller test result. Therefore, the above formula can be simplified to

In the formula

Because the right formula in the above formula contains the approximate values obtained by a plurality of fitting curves, the coefficient calculated by directly adopting the approximate values is suitable for the health monitoring of the AUV, but is not suitable for the navigation process needing to be used as calibration information, the invention provides a training model for the electromechanical propulsion of the AUV, which comprises the following steps:

(2) by (u)_wN) sequence, optimizing motor propulsion model based on fuzzy support vector machine

For sequence { (u)_w1，n₁)，(u_w2，n₂)，…，(u_wn，n_n) H, data u is divided into_wiGenerating a symmetric triangular blur number is marked as A_i＝(a_i-δ_i，a_i，a_i+δ_i)。

Due to the velocity u_wThe relative speed n is characterized byNonlinear regression, selecting proper mapping relation to map the time characteristic to high-order characteristic space, and performing approximate linear regression in the high-dimensional characteristic space, wherein the training set is Tr_：{(Φ(n₁)，A₁)，(Φ(n₂)，A₂)，…，(Φ(n_l)，A_n) In which y is_iI, (i 1, 2 … n). Constructing two kinds of points from a training set Tr in a high-dimensional space, increasing epsilon and reducing epsilon for the y value of each training point in the training set Tr respectively to obtain two sets of positive class points and negative class points, and respectively recording the two sets as D⁺And D^-Then, then

D⁺：{((Φ(n₁)，A₁+ε)；1)，((Φ(n₂)，A₂+ε)；1)，…，((Φ(n_n)，A_n+ε)；1)}

D^-：{((Φ(n₁)，A₁-ε)；-1)，((Φ(n₂)，A₂-ε)；-1)，…，((Φ(n_n)，A_n-ε)；-1)}

At this time, the training set is updated to Tr: { D⁺，D^-And the training and data points are twice of the original training set, and the regression problem is converted into a classification problem.

And carrying out classification training on the processed data by a support vector machine to obtain a fuzzy optimal hyperplane (omega. phi (n)) + eta A + b which is 0 in the fuzzy classification problem, wherein omega is a fuzzy vector, and b is a fuzzy number. Normally, ω ═ ω (ω ═ ω₁，…，ω_n)^TMiddle omega_iI is 1, …, n and b are all triangular fuzzy numbers, andr_i，Δr_i，is omega_iI 1, 2, …, n;α，Δα，the mean value and the left-right spread of b are obtained, so that the problem of solving the fuzzy optimal classification hyperplane (omega. phi (n)) + eta A + b ═ 0 is converted into the problem of solving the opportunistic constraint programming with fuzzy decision

And solving the opportunity constraint planning with fuzzy decision to obtain a fuzzy optimal solution (omega, eta, b), wherein omega is a fuzzy vector formed by triangular fuzzy numbers, eta is a real number, and b is the triangular fuzzy numbers.

Arranging (omega. phi (n)) + eta A + b as 0 to obtain A + (omega)^*·Φ(n))+b^*0, whereinThen omega^*A blur vector formed by triangular blur numbers, b^*Is the triangular blur number, and A is the triangular blur number. Make the triangle fuzzy number (omega)^*·Φ(x))+b^*Is C thenThe speed can be obtained by solving A by C + A as 0Taking the fuzzy center delta a as the speed u_w。

At the moment, the speed u of the water flow relative to the AUV measured by the rotating speed n and the flow velocity profiler is established_wThe model of (1). The model can be described as

Bonding of

Can find out

3. And carrying out dead reckoning according to the speed information and the attitude information phi.

When the DVL information is invalid, an AUV electromechanical state monitoring system is used for obtaining propeller propulsion rotating speed n, and the speed u 'of water flow relative to the AUV corresponding to the rotating speed n is obtained through calculation of a support vector machine model'_wThe ground speed of the AUV is: u '═ u'_w+ u_{w is to the bottom}Assume that the AUV position last updated before DVL failure is (x'₀，y′₀) (x ') if the system sampling period T is constant and the period is high enough, and the AUV makes uniform motion in the sampling period'₀，y′₀) Is at a coordinate position of

By analogy to this, (x'_k，y′_k) Is at a coordinate position of

The motor propulsion model and the fuzzy support vector machine are applied to the dead reckoning method, when the DVL fails or breaks down, the AUV can be assisted to continue dead reckoning navigation, the influence of isolated points or outliers in a data set on a dead reckoning algorithm is reduced, and the method has the advantages of high learning rate and strong generalization capability of the support vector machine and the robustness advantage of fuzzy regression.

The invention is further described as follows:

a dead reckoning method based on motor propulsion model assistance comprises the following specific steps: (1) obtaining and processing AUV position, speed, attitude information and propeller propulsion speed; (2) establishing a motor propulsion model of the water flow relative to the AUV speed relative to the propeller rotation speed; (3) optimizing a motor propulsion model based on a fuzzy support vector machine; (4) and calculating the position of the AUV according to the speed information and the attitude information.

Obtaining and processing AUV position, speed, attitude information and propeller propulsion speed in the step (1):

acquiring initial position (x) of AUV navigation process through GPS₀，y₀) Wherein x is₀Is the initial position longitude, y₀The initial position latitude. When the AUV moves horizontally and the data is judged to be valid through the DVL message information in the DVL measuring range, a group of time series { (u) } is formed by using the AUV navigation process heading speed u acquired by the DVL₁，t₁)，(u₂，t₂)，…，(u_n，t_n) And acquiring a yawing angle phi in the AUV navigation process through an attitude sensor, and calculating through a traditional dead reckoning method.

And meanwhile, recording the current propeller propelling rotation speed n of the AUV by using an electromechanical state monitoring system of the AUV body. Measuring the velocity of water flow relative to AUV by a flow profiler { (u)_w1，t₁)，(u_w2，t₂)，…，(u_wn，t_n) J, the absolute velocity u of the current water flow_{w is to the bottom}Can be represented as u_{w is to the bottom}＝E(u_i-u_wi)。

Establishing a motor propulsion model of the water flow relative to the AUV speed and the propeller rotating speed:

assuming a flow velocity u_{w is to the bottom}The value and the direction of the AUV do not change along with the position of a time point and a space point, the change of the movement speed of the AUV is mainly influenced by the thrust of the propeller, and when the AUV sails stably, the effective thrust provided by the propeller and the total resistance of the ship reach balance, namely:

where ρ is the density of water, u_wThe navigational speed of AUV relative to sea water, omega is AUV wet surface area, zeta is total resistance coefficient, which is a constant in general_pIs the thrust derating coefficient, rho is the seawater density, w_pTo wake factor, D_pIs the diameter of the propeller. Tau is_pThe reasons of the shape, the size and the load of the AUV, the installation position of the AUV and the likeIs related and is usually determined by the AUV's self-contained test or empirical formula. K₀、K₁And K₂Dimensionless thrust K to describe a propeller_PAnd dimensionless resistive torque K_MAnd determining by curve fitting according to the propeller test result.

In the formula

The invention provides an AUV electromechanical propulsion training model which comprises the following steps:

optimizing a motor propulsion model based on the fuzzy support vector machine in the step (3):

for sequence { (u)_w1，n₁)，(u_w2，n₂)，…，(u_wn，n_n) H, data u is divided into_wiGenerating a symmetric triangular blur number is marked as A_i＝(a_i-δ_i，a_i，a_i+δ_i)。

Mapping the speed feature to a high-level feature space, and then performing approximate linear regression in the high-dimensional feature space, wherein the training set is

Tr_：{(Φ(n₁)，A₁)，(Φ(n₂)，A₂)，…，(Φ(n_l)，A_n)}. Wherein y is_i＝i，(i＝1，2…n)

Respectively increasing epsilon and decreasing epsilon (0 < epsilon) to the y value of each training point in the training set Tr to obtain two sets of positive class points and negative class points, and respectively recording the two sets as D⁺And D^-

D⁺：{((Φ(n₁)，A₁+ε)；1)，((Φ(n₂)，A₂+ε)；1)，…，((Φ(n_n)，A_n+ε)；1)}

D^-：{((Φ(n₁)，A₁-ε)；-1)，((Φ(n₂)，A₂-ε)；-1)，…，((Φ(n_n)，A_n-ε)；-1)}

Carrying out classification training of a support vector machine on the processed data to obtain a fuzzy optimal hyperplane (omega phi (n)) + eta A + b which is 0 in the fuzzy classification problem, wherein omega is (omega)₁，…，ω_n)^TIs a blur vector, b is a blur number, i＝1，2，…，n；the problem of fuzzy optimal classification hyperplane (omega. phi (n)) + eta A + b ═ 0 is converted into the problem of solving the opportunity constraint programming with fuzzy decision

And solving the opportunity constraint planning with fuzzy decision to obtain a fuzzy optimal solution (omega, eta, b), wherein omega is a fuzzy vector formed by triangular fuzzy numbers, eta is a real number, and b is the triangular fuzzy numbers.

Arranging (omega. phi (n)) + eta A + b as 0 to obtain A + (omega)^*·Φ(n))+b^*0, whereinWherein ω is^*A blur vector formed by triangular blur numbers, b^*Is the triangular blur number, and A is the triangular blur number. Make the triangle fuzzy number (omega)^*·Φ(x))+b^*Is C, then C ═(c，Δc，) From C + A ═ 0, A can be solved to obtain velocityTaking the fuzzy center delta a as the speed u_w。

At the moment, the speed u of the water flow relative to the AUV measured by the rotating speed n and the flow velocity profiler is established_wThe model of (1). The model is described as

Bonding of

Can find out

And (4) carrying out AUV position estimation according to the speed information and the attitude information:

when the DVL information is invalid, an AUV motor state monitoring system is used for obtaining propeller propelling rotation speed n, and the speed u 'of water flow relative to the AUV corresponding to the rotation speed n is obtained through calculation of a support vector machine model'_wThe ground speed of the AUV is: u '═ u'_w+ u_{w is to the bottom}The AUV position updated last before the DVL failure is (x'₀，y′₀) (x ') if the system sampling period T is constant and the period is high enough, and the AUV makes uniform motion in the sampling period'₀，y′₀) Is at a coordinate position of

By analogy to this, (x'_k，y′_k) Is at a coordinate position of

The method comprises the following steps:

step 1: acquiring initial position (x) of AUV navigation process through GPS₀，y₀) Wherein x is₀Is the initial position longitude, y₀The initial position latitude. When the AUV moves horizontally and the data is judged to be valid through the DVL message information in the DVL measuring range, a group of time series { (u) } is formed by using the AUV navigation process heading speed u acquired by the DVL₁，t₁)，(u₂，t₂)，…，(u_n，t_n) Acquiring a bow angle phi in the AUV navigation process through an attitude sensor, and calculating through a traditional dead reckoning method

Step 2: and recording the current propeller propelling rotation speed n of the AUV by using an electromechanical state monitoring system of the AUV body. Measuring the velocity of water flow relative to AUV by a flow profiler { (u)_w1，t₁)，(u_w2，t₂)，…，(u_wn，t_n) J, the absolute velocity u of the current water flow_{w is to the bottom}Can be represented as u_{w is to the bottom}＝E(u_i-u_wi) I is 1, 2 … n and input to the electromechanical propulsion model of the AUV.

Step 3: AUV needs to overcome water resistance when navigating forwardWhere ρ is the density of water, u_wThe navigational speed of AUV relative to sea water, omega is the wet surface area of AUV, and zeta is the total resistance coefficient.

When the AUV is stably sailing, the effective thrust provided by the propeller and the total resistance of the ship reach balance, namely:

in the formula tau_pIs the thrust derating coefficient, rho is the seawater density, w_pTo wake factor, D_pIs a helixThe diameter of the paddle. Tau is_pUsually determined from self-contained tests or empirical formulas of the AUV. K₀、K₁And K₂Dimensionless thrust K to describe a propeller_PAnd dimensionless resistive torque K_MAnd the test result can be determined by curve fitting according to the propeller test result. Thus, the above equation can be simplified as:

step 4: for sequence { (u)_w1，n₁)，(u_w2，n₂)，…，(u_wn，n_n) Let mu_wi＝max{u_w(i-1)，u_wi，u_w(i+1)}，υ_wi＝min{u_w(i-1)，u_wi，u_w(i+1)}，(i＝2，3，…，n-1)，μ_w1＝max{u_w1，u_w2}，υ_w1＝ min{u_w1，u_w2)，μ_wn＝max{u_w(n-1)，u_wn)，υ_wn＝min{u_w(n-1)，u_wn) Then data u_wi(i-1, 2, …, n) has a center value after blurringMagnitude of blurData u_wiGenerating a symmetric triangular blur number is marked as A_i＝(a_i-δ_i，a_i，a_i+δ_i). Having a membership function of

Step 5: mapping the time characteristic to a high-order characteristic space by taking a proper mapping relation, and then performing approximate linear regression in the high-order characteristic space, wherein the training set is Tr_：{(Φ(n₁)，A₁)，(Φ(n₂)，A₂)，…，(Φ(n_l)，A_n)}。

Step 6: constructing two kinds of points from a training set Tr in a high-dimensional space, specifically, increasing epsilon and decreasing epsilon respectively for y value of each training point in the training set Tr to obtain two sets of positive points and negative points, which are respectively marked as D⁺And D^-And then:

D⁺：{((Φ(n₁)，A₁+ε)；1)，((Φ(n₂)，A₂+ε)；1)，…，((Φ(n_n)，A_n+ε)；1)}

D^-：{((Φ(n₁)，A₁-ε)；-1)，((Φ(n₂)，A₂-ε)；-1)，…，((Φ(n_n)，A_n-ε)；-1)}

step 7: and carrying out support vector machine classification training on the processed data to obtain a fuzzy optimal hyperplane (omega. phi (n)) + eta A + b ═ 0 of a fuzzy classification problem, and converting the problem of solving the fuzzy optimal classification hyperplane (omega. phi (n)) + eta A + b ═ 0 into the problem of solving the opportunity constraint programming with fuzzy decision

And solving the opportunity constraint planning with fuzzy decision to obtain a fuzzy optimal solution (omega, eta, b), wherein omega is a fuzzy vector formed by triangular fuzzy numbers, eta is a real number, and b is the triangular fuzzy numbers.

Arranging (omega. phi (n)) + eta A + b as 0 to obtain A + (omega)^*·Φ(n))+b^*0, whereinThen omega^*A blur vector formed by triangular blur numbers, b^*Is the triangular blur number, and A is the triangular blur number. Make the triangle fuzzy number (omega)^*·Φ(x))+b^*Is C thenThe speed can be obtained by solving A by C + A as 0Taking the fuzzy center delta a as the speed u_w。

Step 8: the speed u of the water flow relative to the AUV measured by the rotating speed n and the flow velocity profiler is established_wThe model of (1). Simplifying the model into

In a combined form

Can find out

Step 9: when the DVL information is invalid, an AUV electromechanical state monitoring system is used for obtaining propeller propulsion rotating speed n, and the speed u 'of water flow relative to the AUV corresponding to the rotating speed n is obtained through calculation of a support vector machine model'_wThe ground speed of the AUV is: u '═ u'_w+u_{w is to the bottom}Assume that the AUV position last updated before DVL failure is (x'₀，y′₀) The sampling period T of the system is constant and the period is high enough, and the AUV makes uniform motion in the sampling period, (x'_k，y′_k) Is at a coordinate position of

In summary, the following steps: the invention provides a dead reckoning method based on motor propulsion model assistance, which is characterized in that GPS, DVL, ADCP and attitude sensor data are collected, a model of the rotating speed n and the AUV relative to the sea water is established by using the motor propulsion model, model precision is improved based on fuzzy support vector machine training, when the DVL data fails, the AUV speed is calculated by depending on the motor rotating speed and the motor propulsion model, the AUV is assisted to continue dead reckoning navigation, and the environment adaptability and the navigation robustness of the AUV are improved.

Claims (5)

1. A dead reckoning method based on motor propulsion model assistance is characterized by comprising the following steps: the method comprises the following specific steps:

(1) obtaining and processing AUV position, speed, attitude information and propeller propulsion speed;

(2) establishing a motor propulsion model of the water flow relative to the AUV speed relative to the propeller rotation speed;

(3) optimizing a motor propulsion model based on a fuzzy support vector machine;

(4) and calculating the position of the AUV according to the speed information and the attitude information.

2. The motor propulsion model-assisted dead reckoning method of claim 1, wherein: AUV position, speed, attitude information and propeller propulsion speed obtain and process:

acquiring initial position (x) of AUV navigation process through GPS₀,y₀) Wherein x is₀Is the initial position longitude, y₀Is the initial position latitude; when the AUV moves horizontally and the data is judged to be valid through the DVL message information in the DVL measuring range, a group of time series { (u) } is formed by using the AUV navigation process heading speed u acquired by the DVL₁,t₁),(u₂,t₂),…,(u_n,t_n) Acquiring a yawing angle phi in the AUV navigation process through an attitude sensor, and calculating through a traditional dead reckoning method;

meanwhile, recording the current propeller propelling rotation speed n of the AUV by using an electromechanical state monitoring system of the AUV body; measuring the velocity of water flow relative to AUV by a flow profiler { (u)_w1,t₁),(u_w2,t₂),…,(u_wn,t_n) J, the absolute velocity u of the current water flow_{w is to the bottom}Can be represented as u_{w is to the bottom}＝E(u_i-u_wi)。

3. The motor propulsion model-assisted dead reckoning method of claim 1, wherein: the method is characterized in that a motor propulsion model of water flow relative to AUV speed relative to propeller rotation speed is established:

assuming a flow velocity u_{w is to the bottom}The value and the direction of the AUV do not change along with the position of a time point and a space point, the change of the movement speed of the AUV is mainly influenced by the thrust of the propeller, and when the AUV sails stably, the effective thrust provided by the propeller and the total resistance of the ship reach balance, namely:

where ρ is the density of water, u_wThe navigational speed of AUV relative to sea water, omega is AUV wet surface area, zeta is total resistance coefficient, which is a constant in general_pIs the thrust derating coefficient, ρ is the density of water, w_pTo wake factor, D_pIs the diameter of the propeller; tau is_pThe method is related to factors such as the appearance, the size and the load of the AUV, the installation position of the AUV and the like, and is generally determined according to the self-propulsion test or the empirical formula of the AUV; k₀、K₁And K₂Dimensionless thrust K to describe a propeller_PAnd dimensionless resistive torque K_MAccording to the propeller test result, determining through curve fitting;

in the formula

A training model of AUV electromechanical propulsion is provided:

4. the motor propulsion model-assisted dead reckoning method of claim 1, wherein: the optimized motor propulsion model based on the fuzzy support vector machine is as follows:

for sequence { (u)_w1,n₁),(u_w2,n₂),…,(u_wn,n_n) H, data u is divided into_wiGenerating a symmetric triangular blur number is marked as A_i＝(a_i-δ_i,a_i，a_i+δ_i)；

Mapping the speed features to a high-order feature space, and then performing approximate linear regression in the high-dimensional feature space, wherein the training set is as follows:

Tr：{(Φ(n₁),A₁),(Φ(n₂),A₂),…,(Φ(n_l),A_n)}

wherein y is_i＝i,(i＝1,2…n)；

Respectively increasing epsilon and decreasing epsilon (0 < epsilon) to the y value of each training point in the training set Tr to obtain two sets of positive class points and negative class points, and respectively recording the two sets as D⁺And D^-

D⁺：{((Φ(n₁),A₁+ε)；1),((Φ(n₂),A₂+ε)；1),…,((Φ(n_n),A_n+ε)；1)}

D^-：{((Φ(n₁),A₁-ε)；-1),((Φ(n₂),A₂-ε)；-1),…,((Φ(n_n),A_n-ε)；-1)}

Carrying out classification training of a support vector machine on the processed data to obtain a fuzzy optimal hyperplane (omega phi (n)) + eta A + b which is 0 in the fuzzy classification problem, wherein omega is (omega)₁，…，ω_n)^TIs a blur vector, b is a blur number,i＝1,2,…,n；the fuzzy optimal classification hyperplane (omega. phi (n)) + eta A + b ═ 0 problem is converted into the solution of the opportunistic constraint programming with fuzzy decision:

solving the opportunity constraint planning with fuzzy decision to obtain a fuzzy optimal solution (omega, eta, b), wherein omega is a fuzzy vector formed by triangular fuzzy numbers, eta is a real number, and b is the triangular fuzzy numbers;

arranging (omega. phi (n)) + eta A + b as 0 to obtain A + (omega)^*·Φ(n))+b^*0, whereinWherein ω is^*A blur vector formed by triangular blur numbers, b^*Is a triangular fuzzy number, A is a triangular fuzzy number; make the triangle fuzzy number (omega)^*·Φ(x))+b^*Is C thenThe speed can be obtained by solving A by C + A as 0Taking the fuzzy center delta a as the speed u_w；

At the moment, the speed u of the water flow relative to the AUV measured by the rotating speed n and the flow velocity profiler is established_wThe model of (2), the model is described as:

combining:

the following can be obtained:

5. the motor propulsion model-assisted dead reckoning method of claim 1, wherein: and the AUV position estimation is carried out according to the speed information and the attitude information:

when the DVL information is invalid, an AUV motor state monitoring system is used for obtaining propeller propelling rotation speed n, and the speed u 'of water flow relative to the AUV corresponding to the rotation speed n is obtained through calculation of a support vector machine model'_wThe ground speed of the AUV is: u '═ u'_w+u_{w is to the bottom}The AUV position updated last before the DVL failure is (x'₀，y′₀) (x ') if the system sampling period T is constant and the period is high enough, and the AUV makes uniform motion in the sampling period'₀，y′₀) The coordinate positions of (a) are:

by analogy to this, (x'_k，y′_k) The coordinate positions of (a) are: