CN111290007A - A BDS/SINS integrated navigation method and system based on neural network assistance - Google Patents

Abstract

The invention provides a BDS/SINS integrated navigation method and system based on neural network assistance, including: establishing a SINS navigation system error model, using the SINS navigation system error model to obtain error data; establishing a BDS/SINS integrated navigation model, and using the error data as The initial value of the BDS/SINS integrated navigation model, and the original value and the estimated value of the target receiver are obtained by using the BDS/SINS integrated navigation model, and the difference between the two is used as a measurement vector; The measured vector is input into the unscented Kalman filter, and the unscented Kalman filter is used to filter the BDS/SINS integrated navigation model; the neural network is used to optimize the training of the filtered BDS/SINS integrated navigation model, and the optimized The BDS/SINS combined navigation model; for tracking navigation. It can effectively suppress the phenomenon that the BDS signal often loses lock, which is conducive to the positioning of vehicles in tunnels, viaducts, mountainous areas and dense high-rise environments.

Description

A BDS/SINS integrated navigation method and system based on neural network assistance

technical field

The invention mainly relates to the technical field of navigation, in particular to a BDS/SINS combined navigation method and system based on neural network assistance.

Background technique

At present, the main disadvantage of inertial navigation is that the positioning error accumulates over time, so it is difficult to work independently for a long time. There are two main ways to solve this problem: to improve the accuracy of the inertial navigation system itself, and the other is to use integrated navigation technology . To improve the accuracy of INS, rely on new materials and new processes to improve the accuracy of inertial sensors, which requires a lot of manpower and material resources, and the improvement of inertial sensor accuracy is limited.

BeiDou Navigation Satellite System (BDS) is my country's national strategy. On December 16 this year (2019), my country used the Long March 3B carrier rocket at the Xichang Satellite Launch Center to "one arrow and two stars". The 52nd and 53rd Beidou navigation satellites were successfully launched. With the successful launch of these two satellites, China's Beidou satellite navigation program has officially entered a milestone - the completion of the launch of all satellites in medium-circular orbits of the earth means that the deployment of the core constellation of the Beidou global system has been completed, and the Beidou-3 will be fully completed in 2020. The system has laid a solid foundation. Some current navigation systems have the problems of losing signal lock and low accuracy.

SUMMARY OF THE INVENTION

The technical problem to be solved by the present invention is to provide a BDS/SINS integrated navigation method and system based on neural network assistance, aiming at the deficiencies of the prior art.

The technical solution of the present invention to solve the above-mentioned technical problems is as follows: a neural network-assisted BDS/SINS combined navigation method, comprising:

establishing a SINS navigation system error model, and using the SINS navigation system error model to obtain error data;

The BDS navigation model and the SINS navigation system error model are combined in a loose combination to obtain a BDS/SINS integrated navigation model, the error data is used as the initial value of the BDS/SINS integrated navigation model, and the BDS/SINS integrated navigation model is used. Obtain the original value and the estimated value of the target receiver, and calculate the difference between the original value and the estimated value, and use the difference as a measurement vector;

Establishing an unscented Kalman filter, inputting the measurement vector into the unscented Kalman filter, and using the unscented Kalman filter to filter the BDS/SINS integrated navigation model;

The Back Propagation neural network is used to optimize the training of the filtered BDS/SINS integrated navigation model, and the optimized BDS/SINS integrated navigation model is obtained;

The target receiver is tracked and navigated by using the optimized BDS/SINS integrated navigation model.

Another technical solution of the present invention to solve the above-mentioned technical problems is as follows: a neural network-assisted BDS/SINS integrated navigation system, comprising:

The error value calculation module is used to establish the error model of the SINS navigation system, and obtain the error data by using the error model of the SINS navigation system;

The integrated navigation model building module is used to combine the BDS navigation model and the SINS navigation system error model through a loose combination to obtain a BDS/SINS integrated navigation model BDS/SINS integrated navigation model, and use the error data as the BDS/SINS combination. The initial value of the navigation model, and the original value and the estimated value of the target receiver are obtained by using the BDS/SINS combined navigation model, and the difference between the original value and the estimated value is calculated, and the difference is used as a measurement vector;

A filter processing module, used for establishing an unscented Kalman filter, inputting the measurement vector into the unscented Kalman filter, and using the unscented Kalman filter to filter the BDS/SINS integrated navigation model;

The optimization module is used to optimize the training of the filtered BDS/SINS integrated navigation model by using the Back Propagation neural network to obtain the optimized BDS/SINS integrated navigation model;

The tracking module is used for tracking and navigating the target receiver by using the optimized BDS/SINS integrated navigation model.

Another technical solution of the present invention to solve the above technical problem is as follows: a neural network-assisted BDS/SINS integrated navigation system, comprising a memory, a processor, and a computer stored in the memory and running on the processor A program, when the processor executes the computer program, implements the navigation method as described above.

Another technical solution of the present invention to solve the above technical problem is as follows: a computer-readable storage medium, the computer-readable storage medium stores a computer program, when the computer program is executed by a processor, the above-mentioned based on A neural network-assisted BDS/SINS integrated navigation method.

The beneficial effects of the invention are: effectively combine the error model of the BDS navigation system and the SINS navigation system, take the error value as the initial value of the combined navigation model, obtain the original value and the estimated value of the target receiver, and calculate the difference between the two. Calculation, filtering through the difference value, and optimizing the process can effectively suppress the phenomenon that the BDS signal often loses lock, which is conducive to the positioning of vehicles in tunnels, viaducts, mountainous areas and high-rise dense environments.

Description of drawings

FIG. 1 is a schematic flowchart of a navigation method provided by an embodiment of the present invention;

2 is a module block diagram of a navigation system provided by an embodiment of the present invention;

3 is a schematic diagram of an overall scheme of a loose combination of BDS/SINS provided by an embodiment of the present invention;

FIG. 4 is a schematic diagram of solving an error model of a SINS navigation system provided by an embodiment of the present invention;

FIG. 5 is a schematic diagram of training a Back Propagation neural network according to an embodiment of the present invention.

Detailed ways

The principles and features of the present invention will be described below with reference to the accompanying drawings. The examples are only used to explain the present invention, but not to limit the scope of the present invention.

FIG. 1 is a schematic flowchart of a navigation method provided by an embodiment of the present invention;

As shown in Figure 1, a neural network-assisted BDS/SINS combined navigation method includes:

Establish a SINS navigation system error model, and use the SINS navigation system error model to obtain error values;

The BDS navigation model and the SINS navigation system error model are combined in a loose combination to obtain a BDS/SINS integrated navigation model, the error data is used as the initial value of the BDS/SINS integrated navigation model, and the BDS/SINS integrated navigation model is used. Obtain the original value and the estimated value of the target receiver, and calculate the difference between the original value and the estimated value, and use the difference as a measurement vector;

Establishing an unscented Kalman filter, inputting the measurement vector into the unscented Kalman filter, and using the unscented Kalman filter to filter the BDS/SINS integrated navigation model;

The Back Propagation neural network is used to optimize the training of the filtered BDS/SINS integrated navigation model, and the optimized BDS/SINS integrated navigation model is obtained;

The target receiver is tracked and navigated by using the optimized BDS/SINS integrated navigation model.

It should be understood that the BDS/SINS combined navigation model is represented as a combination of the BDS navigation model and the SINS navigation system error model.

In the above embodiment, the BDS navigation system and the SINS navigation system error model are effectively combined, the error value is used as the initial value of the combined navigation model, and the original value and the estimated value of the target receiver are obtained, and the difference is calculated between the two, Filtering and optimizing the difference value can effectively suppress the phenomenon that the BDS signal often loses lock, which is beneficial to the positioning of vehicles in tunnels, viaducts, mountainous areas and high-rise dense environments.

而导航计算机中解算给出的姿态矩阵为

两者之间存在偏差。对于变换矩阵

和

一般认为它们的b系是重合的，而将与

对应的导航坐标系称为计算导航坐标系，简记为n'系，所以也常将计算姿态阵记为

因此，

与

之间的偏差在于n'系与n系与之间的偏差。以n系作为参考坐标系，记从n系至n'的等效旋转矢量(失准角误差)记为φ，陀螺测量误差记为

导航系计算误差记为

Optionally, as an embodiment of the present invention, it is assumed that the ideal SINS attitude matrix from the navigation coordinate system (n system) to the carrier coordinate system (b system) is:

And the attitude matrix given by the solution in the navigation computer is

There is a deviation between the two. For the transformation matrix

and

It is generally believed that their b series are coincident, and will be the same as

The corresponding navigation coordinate system is called the calculated navigation coordinate system, abbreviated as n' system, so the calculated attitude array is often recorded as

therefore,

and

The deviation between is the deviation between the n' series and the n series and . Taking the n system as the reference coordinate system, the equivalent rotation vector (misalignment angle error) from the n system to n' is recorded as φ, and the gyro measurement error is recorded as

The calculation error of the navigation system is recorded as

The process of obtaining the error value using the SINS navigation system error model includes:

The misalignment angle error value of the SINS navigation system error model is obtained by the first formula. The first formula is:

为n系相对于i系的旋转，

为b系相对于i系的角速度，

为失准角误差值，

为失准角误差的微分形式，记δ为误差符号，

为n系计算误差，

陀螺测量误差值，

为姿态矩阵；Among them, n is the navigation reference coordinate system, b is the carrier coordinate system, i is the inertial coordinate system,

is the rotation of the n system relative to the i system,

is the angular velocity of the b system relative to the i system,

is the misalignment angle error value,

is the differential form of the misalignment angle error, and denote δ as the error symbol,

Calculate the error for the n-series,

gyro measurement error value,

is the attitude matrix;

The velocity error of the SINS navigation system error model is obtained by the second formula, which is:

为速度误差的微分，

为姿态矩阵，

为加速度计测量值，

为加速度计测量误差，

为地球自转角速度，

为n系转动角速度，

为地球自转角速度计算误差，

为导航系旋转计算误差；Among them, v ⁿ is the velocity value, δ is the error symbol, δv ⁿ is the velocity error,

is the differential of the velocity error,

is the attitude matrix,

is the accelerometer measurement,

is the accelerometer measurement error,

is the angular velocity of the Earth's rotation,

is the rotational angular velocity of the n-series,

Calculate the error for the angular velocity of the Earth's rotation,

Calculate the error for the rotation of the navigation system;

The longitude error, latitude error and altitude error of the SINS navigation system error model are calculated by the longitude error formula, latitude error formula and altitude error formula. The latitude error formula is:

为纬度误差，R_M为子午圈主曲率半径，h为高度，δh为高度误差，v_N为北向速度，δv_N为北向速度误差；in,

is the latitude error, R _M is the main curvature radius of the meridian circle, h is the height, δh is the height error, v _N is the northing velocity, and δv _N is the northing velocity error;

The longitude error formula is:

为经度的微分，

为经度误差的微分，R_Nh为卯酉圈曲率半径，L为纬度值，

为东向速度，

为东向速度误差；in,

is the differential of longitude,

is the differential of the longitude error, R _Nh is the radius of curvature of the unitary circle, L is the latitude value,

is the eastward speed,

is the eastward speed error;

The height error formula is:

为高度的微分，

为高度误差的微分，

为天向速度，

为天向速度误差。in,

is the differential of height,

is the differential of the height error,

is the sky speed,

is the skyward velocity error.

陀螺仪随机常值漂移记为ε_b，通过误差模型的建立，为下一步卡尔曼滤波的系统方程建立做准备。In the SINS navigation system error model, the inertial measurement element consists of a gyroscope and an accelerometer. Both sensors have corresponding device errors. We will record the zero offset for the accelerometer measurement as

The random constant drift of the gyroscope is recorded as ε _b , and the establishment of the error model is used to prepare for the establishment of the system equation of the next Kalman filter.

In the above embodiment, the Kalman filter system equation and measurement equation of the BDS/SINS integrated navigation model are constructed, and the combination method adopts a loose combination, the BDS outputs the original speed and position of the receiver, and the SINS outputs the speed and position calculated by the inertial navigation algorithm. , the difference between the two is used as the measurement input of the Kalman filter. The error of the SINS navigation system error model is input as the initial state of the system.

Optionally, as an embodiment of the present invention, as shown in Figures 3-4, the BDS receiver model in the BDS navigation model includes an antenna, a radio frequency front end, and the radio frequency front end performs signal capture and signal tracking, and outputs data.

The SINS navigation system error model includes a three-axis gyroscope and a three-axis accelerometer. The SINS solution is performed, the solution data is fused with the data in the BDS navigation model, and the SINS navigation system error model outputs the solution value. The BDS/SINS combined navigation model outputs fused values.

Specifically, the process of using the BDS/SINS integrated navigation model to obtain the original value and the estimated value of the target receiver includes:

Calculate the system state of the BDS/SINS integrated navigation model according to the state equation, and the state equation is:

F_w为SINS导航系统误差模型的误差值的状态转移矩阵，

为姿态矩阵,

ω_gx,ω_gy,ω_gz为X、Y、Z轴对应的随机漂移，ω_ax,ω_ay,ω_az为X、Y、Z轴对应的加速度随机误差；Among them, F(t)·X(t) is the original value, G(t)·W(t) is the estimated value, X(t) is the state quantity equation of the integrated navigation model, F(t) is the state transition matrix, G(t) is the noise allocation matrix, W(t) is the system noise vector,

F _w is the state transition matrix of the error value of the SINS navigation system error model,

is the attitude matrix,

ω _gx , ω _gy , ω _gz are the random drifts corresponding to the X, Y, and Z axes, and ω _ax , ω _ay , and ω _az are the acceleration random errors corresponding to the X, Y, and Z axes;

The process of obtaining the state quantity is:

The northeast sky coordinate system is selected as the navigation coordinate system, and the state quantity equation is established according to the 15-dimensional state parameters, and the state quantity equation is:

为组合导航模型的速度误差值，

分别为组合导航模型的经度误差值、纬度误差值和高度误差值，ε_bx,ε_by,ε_bz分别为x、y和z三轴坐标的陀螺随机常值漂移值，

分别为x、y和z三轴坐标的加速度计随机常值零漂。Among them, φ _E , φ _N , φ _U are the three-axis misalignment angle errors of the northeast sky,

is the velocity error value of the integrated navigation model,

are the longitude error value, latitude error value and altitude error value of the integrated navigation model, respectively, ε _bx , ε _by , ε _bz are the random constant drift values of the gyro in the x, y and z three-axis coordinates, respectively,

The random constant zero drift of the accelerometer for the x, y, and z three-axis coordinates, respectively.

Optionally, as an embodiment of the present invention, the process of calculating the difference between the original value and the estimated value and using the difference as a measurement vector includes:

The difference between the original value and the estimated value is calculated using a measurement equation, and the measurement equation is:

H_p和H_v分别为位置和速度的量测矩阵，

V_p和V_v分别为位置测量白噪声和速度测量白噪声。Among them, Z(t) is the measurement vector of the integrated navigation model, H(t) is the measurement matrix of the integrated navigation model, X(t) is the state quantity equation of the integrated navigation model, and V(t) is the integrated navigation model. measurement noise vector,

H _p and H _v are the measurement matrices of position and velocity, respectively,

_Vp and _Vv are white noise for position measurement and white noise for velocity measurement, respectively.

Optionally, as an embodiment of the present invention, the process of performing filtering processing on the BDS/SINS integrated navigation model by using the unscented Kalman filter includes:

The initial state of the combined navigation model is set according to the third formula, which is:

为初始状态值，E(x₀)为期望，T为转置矩阵，P₀为方差；in,

is the initial state value, E(x ₀ ) is the expectation, T is the transposed matrix, and P ₀ is the variance;

The BDS/SINS integrated navigation model is sampled through Sigma qualitative sampling points and a sampling formula, and the sampling formula is:

Among them, ξ _i,k-1 (i=0,1,...2w) is the set of Sigma qualitative sampling points, and w is the dimension of the random vector x;

The first-order weight and the second-order weight are calculated for the Sigma qualitative sampling points by using the first-order weight calculation formula and the second-order weight calculation formula, and the first-order weight calculation formula is:

为一阶权值，λ＝α²(w+K)-w，w为随机向量x的维数，K为比例参数；in,

is the first-order weight, λ=α ² (w+K)-w, w is the dimension of the random vector x, and K is the scale parameter;

The second-order weight calculation formula is:

为二阶权值，λ＝α²(w+K)-w，α为正值的比例缩放因子，α用于调整采样后Sigma点与

的距离，α的取值范围为[0,1]，w为随机向量x的维数，β为非负权系数。in,

is the second-order weight, λ=α ² (w+K)-w, α is a positive scaling factor, α is used to adjust the Sigma point and the

The distance of α is in the range of [0,1], w is the dimension of the random vector x, and β is the non-negative weight coefficient.

Optionally, as an embodiment of the present invention, the process of performing filtering processing on the BDS/SINS integrated navigation model by using the unscented Kalman filter further includes the step of updating:

The Sigma qualitative sampling points are updated according to a first update formula, where the first update formula is:

γ _{i, k/k-1} =f(ξ _i,k-1 ,u _k-1 )+q _k-1 ,

Among them, γ _{i, k/k-1} is the updated point set, f(ξ _{i, k-1} , u _k-1 ) is the nonlinear transformation, q _k-1 is the mean vector of the system process noise; according to the updated The post Sigma qualitative sampling point and the second update formula update the predicted mean, and the second update formula is:

为预测均值，

为一阶权值，γ_i，k/k-1为更新后的点集，k、k-1分别为时刻；in,

is the predicted mean,

is the first-order weight, γ _{i, k/k-1} is the updated point set, and k and k-1 are the moments respectively;

Update the covariance according to the updated Sigma qualitative sampling points and the third update formula, where the third update formula is:

为二阶权值，γ_i，k/k-1为更新后的点集，Q_k-1为非负定方差矩阵；Among them, P _k/k-1 is the covariance,

is the second-order weight, γ _{i, k/k-1} is the updated point set, and Q _k-1 is the non-negative definite variance matrix;

The measurement equation is updated according to the fourth update formula, and the fourth update formula is:

为二阶权值，

为一阶权值，x_i,k/k-1为更新后的预测均值，k、k-1分别为时刻，R_k为量测噪声的正定方差矩阵。Among them, P _z,k is the measurement data,

is the second-order weight,

is the first-order weight, x _{i, k/k-1} is the updated prediction mean value, k and k-1 are the time respectively, and R _k is the positive definite variance matrix of the measurement noise.

In the above embodiment, unscented Kalman filtering is used for Kalman filtering, because in actual target tracking, the state model and measurement model of the tracking system are mostly nonlinear, so unscented Kalman filtering is used to improve data accuracy.

Optionally, as an embodiment of the present invention, as shown in FIG. 5 , the process of using the Back Propagation neural network to optimize the training of the filtered BDS/SINS combined navigation model includes:

Iteratively process the Back Propagation neural network according to the conjugate gradient method;

Correct the weights of the Back Propagation neural network after iterative processing;

The Back Propagation neural network is predicted and corrected by using the modified Back Propagation neural network.

When the BDS signal is locked, the BDS and SINS use the unscented Kalman filter for combined navigation, and the BP neural network is trained online based on the conjugate gradient optimization. When the BDS signal loses lock, the model trained when the signal is locked is called for online training It is predicted to reduce the problem of rapid decline of SINS accuracy in the case of BDS signal loss of lock.

Specifically, the BP (Back Propagation) neural network, that is, the learning process of the error back propagation algorithm, is composed of two processes: forward propagation of information and back propagation of errors. The traditional BP neural network algorithm adopts the direction of steepest descent, that is, the direction of negative gradient, but it only reflects the local nature of the error function at a certain point, not necessarily the direction of global steepest descent. The conjugate gradient method is an algorithm to improve the search direction. The conjugate direction is obtained by adding a correction term to the original negative gradient direction of the BP standard algorithm, which makes the steepest descent method conjugate. In short, the search direction of the conjugate gradient method is a conjugate direction. The algorithm mainly uses the conjugate gradient direction to modify the weight w _k , which makes the determination of w _k faster, speeds up the training speed of the BP neural network, and avoids the network falling into the local minimum.

The number of layers in the Back Propagation neural network is three layers, namely the input layer, the hidden layer, and the output layer.

Let the objective function of Back Propagation neural network be:

Among them, the hidden layer is a nonlinear layer, using the sigmoid function:

The output of the hidden layer node is:

The output of the output node is:

In the formula, x _i , y _i , and zl are the input, hidden layer, and output nodes, respectively, w _ji is the network weight between x _i and y _i , and v _ij is the network weight between y _j and z _i . θ _j and θ _l are the thresholds between x _i and _yi and between y _j and _zi , respectively.

Specifically, the modified Back Propagation neural network is used to predict and correct the Back Propagation neural network

In the above embodiment, the neural network adopts the improved BP neural network algorithm based on the conjugate gradient method. The conjugate gradient method is an algorithm for improving the search direction. It multiplies the gradient of the previous point by an appropriate coefficient and adds it to the modified point. On the gradient, a new search direction is obtained. Speed up the training speed of BP neural network and avoid the network falling into local minimum;

The Back Propagation neural network is iteratively processed according to the conjugate gradient method, and the iterative equation is:

P _k+1 =-g _k+1 +β _k P _k ,

where:

P_k为搜索方向，g _k is the gradient of E to w _k

P _k is the search direction,

β _k is the conjugation factor whose value is

The weights of the BP neural network are modified by the above iterative equation:

Δw(k+1)=w(k)+λ(k)P(k), where λ(k) is the optimal step size.

2 is a module block diagram of a navigation system provided by an embodiment of the present invention;

Optionally, as another embodiment of the present invention, as shown in FIG. 2 , a neural network-assisted BDS/SINS integrated navigation system includes:

The error value calculation module is used to establish the error model of the SINS navigation system, and obtain the error data by using the error model of the SINS navigation system;

An integrated navigation model establishment module is used to combine the BDS navigation model and the SINS navigation system error model through a loose combination to obtain a BDS/SINS integrated navigation model, and the error data is used as the initial value of the BDS/SINS integrated navigation model, And utilize the BDS/SINS combined navigation model to obtain the original value and the estimated value of the target receiver respectively, and calculate the difference between the original value and the estimated value, and use the difference as a measurement vector;

A filter processing module, used for establishing an unscented Kalman filter, inputting the measurement vector into the unscented Kalman filter, and using the unscented Kalman filter to filter the BDS/SINS integrated navigation model;

The optimization module is used to optimize the training of the filtered BDS/SINS integrated navigation model by using the Back Propagation neural network to obtain the optimized BDS/SINS integrated navigation model;

The tracking module is used for tracking and navigating the target receiver by using the optimized BDS/SINS integrated navigation model.

Optionally, as another embodiment of the present invention, a neural network-assisted BDS/SINS integrated navigation system includes a memory, a processor, and a computer stored in the memory and running on the processor. A program, when the processor executes the computer program, implements the neural network-assisted BDS/SINS integrated navigation method as described above.

Optionally, as another embodiment of the present invention, a computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the above-mentioned based A neural network-assisted BDS/SINS integrated navigation method.

Those skilled in the art can clearly understand that, for the convenience and brevity of description, for the specific working process of the system and unit described above, reference may be made to the corresponding process in the foregoing method embodiments, which will not be repeated here.

In the several embodiments provided in this application, it should be understood that the disclosed system and method may be implemented in other manners. For example, the system embodiments described above are only illustrative. For example, the division of units is only a logical function division. In actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated. to another system, or some features can be ignored, or not implemented.

Units described as separate components may or may not be physically separated, and components shown as units may or may not be physical units, that is, may be located in one place, or may be distributed to multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solutions of the embodiments of the present invention.

In addition, each functional unit in each embodiment of the present invention may be integrated into one processing unit, or each unit may exist physically alone, or two or more units may be integrated into one unit. The above-mentioned integrated units may be implemented in the form of hardware, or may be implemented in the form of software functional units.

The integrated unit, if implemented as a software functional unit and sold or used as a stand-alone product, may be stored in a computer-readable storage medium. For such understanding, the technical solution of the present invention is essentially or the part that contributes to the prior art, or all or part of the technical solution can be embodied in the form of a software product, and the computer software product is stored in a storage device. The medium includes several instructions for causing a computer device (which may be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods of the various embodiments of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, Read-Only Memory (ROM, Read-Only Memory), Random Access Memory (RAM, Random Access Memory), magnetic disk or optical disk and other media that can store program codes .

The above are only specific embodiments of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art can easily think of various equivalent modifications or modifications within the technical scope disclosed by the present invention. Replacement, these modifications or replacements should all be covered within the protection scope of the present invention. Therefore, the protection scope of the present invention should be subject to the protection scope of the claims.

Claims (9)

1. a BDS/SINS combined navigation method based on neural network assistance, is characterized in that, comprises: establishing a SINS navigation system error model, and using the SINS navigation system error model to obtain error data; The BDS navigation model and the SINS navigation system error model are combined in a loose combination to obtain a BDS/SINS integrated navigation model, the error data is used as the initial value of the BDS/SINS integrated navigation model, and the BDS/SINS integrated navigation model is used. Obtain the original value and the estimated value of the target receiver, and calculate the difference between the original value and the estimated value, and use the difference as a measurement vector; Establishing an unscented Kalman filter, inputting the measurement vector into the unscented Kalman filter, and using the unscented Kalman filter to filter the BDS/SINS integrated navigation model; The Back Propagation neural network is used to optimize the training of the filtered BDS/SINS integrated navigation model, and the optimized BDS/SINS integrated navigation model is obtained; The target receiver is tracked and navigated by using the optimized BDS/SINS integrated navigation model. 2. navigation method according to claim 1, is characterized in that, the described process that utilizes SINS navigation system error model to obtain error data comprises: The misalignment angle error value of the SINS navigation system error model is obtained by the first formula. The first formula is:

Among them, n is the navigation reference coordinate system, b is the carrier coordinate system, i is the inertial coordinate system,

is the rotation of the n system relative to the i system,

is the angular velocity of the b system relative to the i system,

is the misalignment angle error value,

is the differential form of the misalignment angle error, and denote δ as the error symbol,

Calculate the error for the n-series,

gyro measurement error value,

is the attitude matrix; The velocity error of the SINS navigation system error model is obtained by the second formula, which is:

Among them, v ⁿ is the velocity value, δ is the error symbol, δv ⁿ is the velocity error,

is the differential of the velocity error,

is the attitude matrix,

is the accelerometer measurement,

is the accelerometer measurement error,

is the angular velocity of the Earth's rotation,

is the rotational angular velocity of the n-series,

Calculate the error for the angular velocity of the Earth's rotation,

Calculate the error for the rotation of the navigation system; The longitude error, latitude error and altitude error of the SINS navigation system error model are calculated by the longitude error formula, latitude error formula and altitude error formula. The latitude error formula is:

in,

is the latitude error, R _M is the main curvature radius of the meridian circle, h is the height, δh is the height error, v _N is the northing velocity, and δv _N is the northing velocity error; The longitude error formula is:

in,

is the differential of longitude,

is the differential of the longitude error, R _Nh is the radius of curvature of the unitary circle, L is the latitude value,

is the eastward speed,

is the eastward speed error; The height error formula is:

in,

is the differential of height,

is the differential of the height error,

is the sky speed,

is the skyward velocity error. 3. navigation method according to claim 1 is characterized in that, the process that utilizes BDS/SINS integrated navigation model to obtain the original value of target receiver and estimated value comprises: Calculate the system state of the BDS/SINS integrated navigation model according to the state equation, and the state equation is:

Among them, F(t)·X(t) is the original value, G(t)·W(t) is the estimated value, X(t) is the state quantity equation of the integrated navigation model, F(t) is the state transition matrix, G(t) is the noise allocation matrix, W(t) is the system noise vector,

F _w is the state transition matrix of the error value of the SINS navigation system error model,

is the attitude matrix,

ω _gx , ω _gy , ω _gz are the random drifts corresponding to the X, Y, and Z axes, and ω _ax , ω _ay , and ω _az are the acceleration random errors corresponding to the X, Y, and Z axes; The process of obtaining the state quantity is: The northeast sky coordinate system is selected as the navigation coordinate system, and the state quantity equation is established according to the 15-dimensional state parameters, and the state quantity equation is:

Among them, φ _E , φ _N , φ _U are the three-axis misalignment angle errors of the northeast sky,

is the velocity error value of the integrated navigation model,

are respectively the longitude error value, latitude error value and altitude error value of the integrated navigation model, ε _bx , ε _by , ε _bz are the random constant drift values of the gyro in the three-axis coordinates of x, y and z, respectively, ▽ _x , ▽ _y , ▽ _z is the random constant zero drift of the accelerometer of the three-axis coordinates of x, y and z, respectively. 4. The navigation method according to claim 3, wherein the process of calculating the difference between the original value and the estimated value and using the difference as a measurement vector comprises: The difference between the original value and the estimated value is calculated using a measurement equation, and the measurement equation is: Z(t)=H(t)·X(t)+V(t), Among them, Z(t) is the measurement vector of the integrated navigation model, H(t) is the measurement matrix of the integrated navigation model, X(t) is the state quantity equation of the integrated navigation model, and V(t) is the integrated navigation model. measurement noise vector,

H _p and H _v are the measurement matrices of position and velocity, respectively,

_Vp and _Vv are white noise for position measurement and white noise for velocity measurement, respectively. 5. navigation method according to claim 4 is characterized in that, the described process that utilizes described unscented Kalman filter to carry out filtering processing to BDS/SINS integrated navigation model comprises: The initial state of the combined navigation model is set according to the third formula, which is:

in,

is the initial state value, E(x ₀ ) is the expectation, T is the transposed matrix, and P ₀ is the variance; The BDS/SINS integrated navigation model is sampled through Sigma qualitative sampling points and a sampling formula, and the sampling formula is:

Among them, ξ _i,k-1 (i=0,1,...2w) is the set of Sigma qualitative sampling points, and w is the dimension of the random vector x; The first-order weight and the second-order weight are calculated for the Sigma qualitative sampling points by using the first-order weight calculation formula and the second-order weight calculation formula, and the first-order weight calculation formula is:

Wherein, Wi ^m is the first-order weight, λ=α ² ( _w +K)-w, w is the dimension of the random vector x, and K is the scale parameter; The second-order weight calculation formula is:

Among them, W _i ^c is the second-order weight, λ=α ² (w+K)-w, α is a positive scaling factor, α is used to adjust the Sigma point after sampling and the

The distance of α is in the range of [0,1], w is the dimension of the random vector x, and β is the non-negative weight coefficient. 6. navigation method according to claim 5, is characterized in that, the described process that utilizes described unscented Kalman filter to carry out filtering process to BDS/SINS integrated navigation model also comprises the step of updating: The Sigma qualitative sampling points are updated according to a first update formula, where the first update formula is: γ _{i, k/k-1} =f(ξ _i,k-1 ,u _k-1 )+q _k-1 , Among them, γ _{i, k/k-1} is the updated point set, f(ξ _{i, k-1} , u _k-1 ) is the nonlinear transformation, q _k-1 is the mean vector of the system process noise; according to the updated The post Sigma qualitative sampling point and the second update formula update the predicted mean, and the second update formula is:

in,

is the predicted mean value, W _i ^m is the first-order weight, γ _{i, k/k-1} is the updated point set, and k and k-1 are the moments respectively; Update the covariance according to the updated Sigma qualitative sampling points and the third update formula, where the third update formula is:

Among them, P _k/k-1 is the covariance, W _i ^c is the second-order weight, γ _{i, k/k-1} is the updated point set, and Q _k-1 is the non-negative definite variance matrix; The measurement equation is updated according to the fourth update formula, and the fourth update formula is:

Among them, P _z,k is the measurement data, W _i ^c is the second-order weight,

Wi ^m is the first-order weight, x _{i ,k/k-1} is the updated prediction mean value, k and k-1 are the moments respectively, and R _k is the positive definite variance matrix of the measurement noise. 7. navigation method according to claim 5, is characterized in that, the described process that utilizes Back Propagation neural network to carry out optimization training to the BDS/SINS combined navigation model after filtering processing comprises: Iteratively process the Back Propagation neural network according to the conjugate gradient method; Correct the weights of the Back Propagation neural network after iterative processing; The Back Propagation neural network is predicted and corrected by using the modified Back Propagation neural network. 8. A BDS/SINS integrated navigation system based on neural network assistance is characterized in that, comprising: The error value calculation module is used to establish the error model of the SINS navigation system, and obtain the error data by using the error model of the SINS navigation system; An integrated navigation model building module is used to combine the BDS navigation model and the SINS navigation system error model through a loose combination to obtain a BDS/SINS integrated navigation model, and the error data is used as the initial value of the BDS/SINS integrated navigation model, And utilize the BDS/SINS combined navigation model to obtain the original value and the estimated value of the target receiver, and calculate the difference between the original value and the estimated value, and use the difference as a measurement vector; A filtering processing module, used for establishing an unscented Kalman filter, inputting the measurement vector into the unscented Kalman filter, and using the unscented Kalman filter to filter the BDS/SINS integrated navigation model; The optimization module is used to optimize and train the filtered BDS/SINS integrated navigation model by using the Back Propagation neural network to obtain the optimized BDS/SINS integrated navigation model; The tracking module is used for tracking and navigating the target receiver by using the optimized BDS/SINS integrated navigation model. 9. A neural network-assisted BDS/SINS integrated navigation system, comprising a memory, a processor, and a computer program stored in the memory and running on the processor, characterized in that when the processor When the computer program is executed, the neural network-assisted BDS/SINS integrated navigation method according to any one of claims 1 to 7 is realized.

Images