CN109059909A - Satellite based on neural network aiding/inertial navigation train locating method and system - Google Patents

Abstract

The present invention provides a satellite/inertial navigation train positioning method and system based on neural network assistance, wherein the method includes: detecting the collected satellite data, if the positioning accuracy is good, then transmitting to the data fusion module for data fusion, if the positioning If the accuracy does not meet the requirements, the set of satellite data is discarded; then, the inertial navigation calculation module performs calculation on the collected data of the inertial measurement unit, and fuses the calculation result with the positioning result of the satellite. Finally, output the result after fusion processing and integrate it into the on-board computer. The invention can solve the problems of low data reliability and low precision in the existing train positioning system.

Description

Satellite/inertial navigation train positioning method and system based on neural network

technical field

The present invention relates to the technical field of vehicle positioning, and more specifically, relates to a satellite/inertial-guided train positioning method and system assisted by a neural network.

Background technique

At present, the existing train positioning methods mainly contain the following: odometer, query transponder, Doppler radar, track circuit. Among them, the cost of the odometer is low, and the position is obtained by speed integration, and there is error accumulation; Doppler radar is mainly used for speed measurement, and the accuracy is high, but there is also error accumulation and high cost; a large number of transponders and track circuits are laid on the ground. In order to eliminate the cumulative error of positioning, it is necessary to add a corresponding receiving device on the train, but it requires a lot of cost, and the ground equipment needs regular maintenance, which requires a large amount of maintenance work and low efficiency. The train speed and position information is the main research object of the train control system. The measurement error will directly affect the train safety protection distance, tracking interval, block control mode, etc. If the error is too large, it will directly touch the train brake, affecting the train efficiency and passenger comfort. In severe cases, it even endangers the safety of train driving.

In recent years, satellite navigation and positioning technology has gradually been researched and applied in the field of train positioning. Satellite navigation and positioning technology has the advantages of all-weather, continuous, real-time, and long-term high precision, and can provide precise positioning of trains. However, satellite signals are easily interfered by the environment and cause The positioning accuracy drops rapidly, so only using satellites as the information source for positioning has insufficient security and reliability. Inertial navigation has the characteristics of complete autonomous work independent of the external environment, complete navigation information (attitude, speed, position), good dynamics and continuity, etc., and can provide high-precision positioning information in a short time, but inertial navigation requires trains. Initial information and positioning errors diverge over time, and the combination of the two navigation methods can achieve complementary advantages and disadvantages. Therefore, satellite/inertial navigation is one of the most widely used integrated navigation methods at present.

The accuracy of the satellite/inertial navigation train positioning system mainly depends on the satellite signal. When the satellite is good, the system can output high-precision positioning data for a long time without interruption, providing reliable position information for the safe driving of the train. When the satellite signal is not Well, for example, when a train passes through cities, forests and other places covered by tall buildings or trees, the multipath effect will cause large errors in the satellite positioning effect and affect the system output. Even when the satellite is completely out of lock in caves, tunnels and other environments, then The output of the system is calculated by the inertial measurement unit after collecting data. Due to the accumulation of errors in the inertial measurement unit, the output accuracy of the system gradually decreases, and the environment where satellites are out of lock, such as caves and tunnels, is the area where accidents occur frequently. Therefore, in these In this environment, reliable and accurate train location information is more important for the safe driving of trains. Therefore, improving the positioning accuracy of the train in the case of satellite loss of lock also means improving the reliability and safety of the train positioning system.

In order to solve the problems of low data reliability and low precision in the existing train positioning system, the present invention provides a satellite/inertial navigation train positioning method and system based on neural network assistance.

Contents of the invention

In view of the above problems, the object of the present invention is to provide a satellite/inertial navigation train positioning method and system based on neural network assistance, to solve the problems of low data reliability and low precision in the existing train positioning system.

The present invention provides a satellite/inertial navigation train positioning method based on neural network assistance, comprising:

Obtain sensor real-time data, wherein the sensor real-time data includes inertial measurement unit data and satellite data;

The inertial measurement unit data is calculated by the inertial navigation calculation module to obtain the attitude, speed and position of the train;

Detecting the satellite data, wherein, detecting whether the satellite ephemeris observation data and the satellite positioning precision factor value meet the set condition, if the condition is met, the satellite data is available, and if the condition is not met, the satellite data is unavailable ;

When the satellite data is available, the satellite data and the attitude, speed and position information of the train obtained after the calculation by the inertial navigation calculation module are fused to obtain the corrected speed and position of the train;

Simultaneously, the attitude, speed and position information of the train obtained after the satellite data and the calculation by the inertial navigation module are input as the neural network, and the system calculation error value is output as the neural network, and the neural network is trained;

When the satellite data is not available, the output value after the training of the neural network and the attitude, speed and position information of the train obtained after the calculation of the inertial navigation calculation module are fused to obtain the corrected speed and position of the train.

The present invention also provides a satellite/inertial navigation train positioning system based on neural network assistance, including:

The inertial navigation calculation module is used to obtain the inertial measurement unit data in real time, and obtain the attitude, speed and position of the train;

The satellite signal receiving module is used to obtain satellite data in real time, and detect the satellite data, wherein, detect whether the satellite ephemeris observation data and the satellite positioning precision factor value meet the set conditions, and if the conditions are met, the satellite data Available, if the condition is not met, this satellite data is not available;

The data fusion module is used to fuse the satellite data with the attitude, speed and position information of the train obtained after the calculation by the inertial navigation calculation module when the satellite data is available, and obtain the corrected speed and position information of the train. Location;

Simultaneously, the attitude, speed and position information of the train obtained after the satellite data and the calculation by the inertial navigation module are input as the neural network, and the system calculation error value is output as the neural network, and the neural network is trained;

When the satellite data is not available, the output value after the training of the neural network and the attitude, speed and position information of the train obtained after the calculation of the inertial navigation calculation module are fused to obtain the corrected speed and position of the train.

As can be seen from the above technical scheme, the satellite/inertial navigation train positioning method and system based on neural network assistance provided by the present invention can obtain the following beneficial effects:

1, the train positioning method that the present invention proposes can not only improve the precision of train under the good situation of satellite signal, and can provide the train positioning result of higher precision under the abnormal condition of satellite signal.

2. The train positioning method proposed by the present invention adopts neural network-assisted data fusion, which can effectively suppress the divergence of system positioning errors under the abnormal situation of satellites, and can also improve the reliability and safety of the satellite/inertial navigation train positioning system.

3. Each module of the train positioning system proposed by the present invention works independently without affecting each other, which can improve the stability and anti-interference of the system.

To the accomplishment of the above and related ends, one or more aspects of the invention include the features hereinafter described in detail. The following description and accompanying drawings detail certain exemplary aspects of the invention. These aspects are indicative, however, of but a few of the various ways in which the principles of the invention may be employed. Furthermore, the invention is intended to include all such aspects and their equivalents.

Description of drawings

By referring to the following description in conjunction with the accompanying drawings, and with a more comprehensive understanding of the present invention, other objects and results of the present invention will be more clear and easy to understand. In the attached picture:

Fig. 1 is the schematic flow chart of satellite/inertial navigation train positioning method based on neural network assistance according to an embodiment of the present invention;

Fig. 2 is the satellite/inertial navigation train positioning system structural representation based on neural network assistance according to an embodiment of the present invention;

Fig. 3 is a block diagram of algorithmic block diagram based on neural network assisted satellite/inertial navigation train positioning solution according to an embodiment of the present invention;

4 is a schematic diagram of an inertial navigation update algorithm according to an embodiment of the present invention;

5 is a schematic flow diagram of a data fusion algorithm according to an embodiment of the present invention;

6 is a schematic diagram of a neural network model according to an embodiment of the present invention;

7 is a schematic diagram of neural network training logic according to an embodiment of the present invention;

Fig. 8 is a logical schematic diagram of a neural network predicting a filtering measurement value in a system according to an embodiment of the present invention.

The same reference numerals indicate similar or corresponding features or functions throughout the drawings.

Detailed ways

In the following description, for purposes of explanation, numerous specific details are set forth in order to provide a thorough understanding of one or more embodiments. It is evident, however, that these embodiments may be practiced without these specific details.

In view of the problems of low data reliability and low precision in the existing train positioning system proposed above, the present invention provides a satellite/inertial navigation train positioning method and system based on neural network assistance.

Specific embodiments of the present invention will be described in detail below in conjunction with the accompanying drawings.

In order to illustrate the neural network-based satellite/inertial navigation train positioning method provided by the present invention, Fig. 1 shows the process of the neural network-based satellite/inertial navigation train positioning method according to an embodiment of the present invention.

As shown in Figure 1, the satellite/inertial navigation train positioning method based on neural network assistance provided by the present invention comprises: S110: obtain sensor real-time data, wherein, sensor real-time data comprises inertial measurement unit data and satellite data;

S120: through the inertial navigation calculation module, the inertial measurement unit data is calculated to obtain the attitude, speed and position of the train;

S130: Detect the satellite data, wherein, detect whether the satellite ephemeris observation data and the satellite positioning precision factor value meet the set condition, if the condition is met, the satellite data is available, and if the condition is not met, the satellite data is unavailable ;

S140: When the satellite data is available, the satellite data is fused with the attitude, speed and position information of the train obtained through the inertial navigation calculation module to obtain the corrected speed and position of the train;

S150: At the same time, the satellite data and the attitude, speed and position information of the train acquired after the inertial navigation calculation module are used as the neural network input, and the system calculation error value is used as the neural network output to train the neural network;

S160: When the satellite data is not available, the output value after the training of the neural network is fused with the attitude, speed and position information of the train obtained through the calculation of the inertial navigation calculation module, and the corrected speed and position of the train are obtained.

Fig. 1 in conjunction with Fig. 3, the detailed steps of the satellite/inertial navigation train positioning method based on neural network assistance provided by the present invention are as follows: Step 1: IMU data collection S101, satellite data reception S102;

Step 2: Satellite signal detection. The satellite signal detection algorithm refers to detecting whether the satellite ephemeris observation data and the satellite positioning precision factor value meet the set conditions. If the conditions are met, the satellite data is available. If the conditions are not met, the satellite data The data is unavailable; if the condition is not met, the output of the neural network is used as the measured value of data fusion S106, and if the condition is met, the set of satellite data is output S105.

Among them, it should be noted that, according to the actual requirements of the application, the conditions set in the present invention are: the number of satellites > 6 or HDOP < 10, that is, when either or both of the conditions are satisfied , the satellite data can be used in S105, if one of the two conditions is not satisfied, then the satellite data cannot be used in S106. When the number of satellites is 6, the accuracy of satellite data detection can be satisfied, and when it is greater than 6, the accuracy of satellite data detection can be better satisfied, so one of the conditions set in the present invention is that the number of satellites>6.

Step 3: inertial navigation calculation S103, the inertial navigation calculation refers to the use of the data of the collected inertial measurement unit for calculation to obtain the attitude, speed and position of the carrier, and the result of the inertial navigation calculation is output S104;

Step 4: data fusion S107. The information output by the satellite receiver (when the satellite data is available), the neural network prediction value (when the satellite data is not available) and the result of the inertial navigation solution are fused to obtain a corrected train position estimate;

Step 5: output the result of step 4 to the vehicle-mounted computer S108;

Step 6: neural network training S109. The sensor information and calculated information in the system are used as the input of the neural network, and the calculated error value of the system is used as the expected output value, and the neural network is trained until the required precision is reached.

Among them, the acquisition of real-time sensor data further includes: the three-axis (x, y, z) angular rate output by the gyroscope, the three-axis (x, y, z) acceleration output by the accelerometer, and the longitude, latitude, and altitude output by the satellite receiver. , eastward velocity, northward velocity, and skyward velocity.

As shown in Figure 3, a kind of satellite/inertial navigation train positioning method based on neural network assistance provided by the present invention, the specific process of inertial navigation solution algorithm is as follows:

(1) Speed update

The speed update equation is:

in,

Δv _m is the specific force increment output by the accelerometer in the time period [t _m-1 , t _m ], which is approximated by multiplying the specific force output by the sampling interval in practice. Δθ _m is the angular increment output by the gyroscope in the time period [t _m-1 ,t _m ], which can be approximately transformed into an angular increment by multiplying it by the sampling interval T _s .

(2) Earth parameter calculation

To calculate the principal radius of curvature of the meridian circle and the principal radius of curvature of the Maoyou circle, the calculation formula is as follows:

Among them, e is the eccentricity of the ellipse, f is the oblateness of the ellipse, f=1/298.257, R _e is the radius of the earth, R _M is the main curvature radius of the meridian circle, and R _N is the main curvature radius of the meridian circle.

(3) Location update

The position update equation is:

Among them, T is the sampling period, λ is the longitude of the train location, L is the latitude of the train location, and h is the height of the train location.

(4) Posture update

The angular velocity measured by the gyroscope is used as the quaternion update, and the gravity acceleration is used as the observation of the quaternion to calculate the attitude angle in real time.

The matrix form of the attitude quaternion differential equation is:

Among them, T is the sampling interval, q ₀ , q ₁ , q ₂ , and q ₃ represent the quaternion of the attitude;

ω _x , ω _y , and ω _z represent the three-axis (x, y, z) output of the gyroscope.

Normalize the updated attitude quaternion:

Wherein, the subscripts i=0, 1, 2, 3 respectively represent the values in the attitude quaternion.

The attitude angle of the carrier can be calculated according to the normalized quaternion.

As shown in Figure 5, a kind of satellite/inertial navigation train positioning method based on neural network assistance provided by the present invention, the specific flow of data fusion algorithm is as follows:

(1) Establish the satellite/inertial navigation train positioning system mathematical model:

The state equation of the system is:

X _k ＝Φ _k/k-1 X _k-1 +Γ _k/k-1 W _k-1

Among them, X _k represents the state vector at time k, Φ _k/k-1 represents the 15×15-dimensional state one-step transition matrix, Γ _k/k-1 represents the 15×15-dimensional system noise allocation matrix, W _k-1 represents A 15×1-dimensional system noise vector, which is a zero-mean Gaussian white noise vector sequence.

The measurement equation of the system is:

Z _k ＝H _k X _k +V _k

Among them, Z _k represents the system measurement vector, H _k represents the 6×15-dimensional measurement matrix, and V _k represents the 6×1-dimensional measurement noise vector, which is a zero-mean Gaussian white noise vector sequence.

(2) Establish the system error equation

The equation for establishing a simplified system error based on the system state matrix is:

in, and Respectively represent the angular velocity and error of the n system relative to the i system, and Respectively represent the angular velocity and error of the n system relative to the e system, and Respectively represent the angular velocity and error of the e system relative to the i system, is the direction cosine matrix of the b system relative to the n system. The state one-step transition matrix Φ _k/k-1 can be obtained from the above formula.

(3) Data fusion:

The data fusion algorithm is divided into two parts, prediction update and measurement update. Prediction update refers to updating the system state at this moment through the established system model and the system state at the previous moment. Measurement update refers to the use of The measured value updates the system state to obtain the final system state value. After the forecast update, check whether there is a new measurement value, if there is, perform the measurement update, if not, directly use the forecast update value to correct the inertial navigation calculation value.

(4) Use the system state vector after data fusion to correct the speed and position value of the system:

in, are the estimated velocity errors in the three directions (east-north-day) of the system, respectively, are the system latitude, longitude, and altitude error estimates, respectively.

(5) Output the corrected data.

As shown in Fig. 6, a neural network-based satellite/inertial navigation train positioning method provided by the present invention has a neural network topology of three layers, namely an input layer, a hidden layer, and an output layer. In Figure 6, φ ₁ , φ ₂ ... φ _k represent the radial basis function values of each node in the hidden layer, and ω ₁ , ω ₂ , ω ₃ ... ω _k represent the weight information from each node in the hidden layer to the output layer . x is the input of the network and y is the output of the neural network.

As shown in Figure 7 and Figure 8, a neural network-assisted satellite/inertial navigation train positioning method provided by the present invention, the neural network-assisted train positioning scheme is as follows: when the satellite signal is normal, use satellite and inertial navigation to solve The output speed and position information is fused to estimate the position and speed information of the train, and the selected neural network input and output are used to train the neural network; when the satellite signal is not available, the output value of the neural network is used as the measurement value of the data fusion. Make the filter continue to provide correction values for the inertial navigation system to obtain satisfactory positioning data.

In summary, the present invention provides a neural network-assisted satellite/inertial navigation train positioning method. The neural network-assisted train positioning scheme is: when the satellite signal is normal, use the satellite and inertial navigation to calculate the output speed and position After the information is fused, the position and speed information of the train is estimated, and the selected neural network input and output are used to train the neural network; when the satellite data is not available, the output value of the neural network is used as the measurement value of the data fusion, so that the inertial navigation system Continue to get corrections.

Corresponding to the above method, the present invention also provides a neural network-assisted satellite/inertial navigation train positioning system, and Fig. 2 shows the logical structure of the neural network-based satellite/inertial navigation train positioning system according to an embodiment of the present invention .

As shown in Figure 2, a kind of neural network-assisted satellite/inertial navigation train positioning system provided by the present invention includes a safety power supply module 1, an inertial navigation calculation module 2, a satellite signal receiving module 3, a data fusion module 4, and a data output module Module 5, on-board computer 6.

Among them, the inertial navigation calculation module 2 is composed of two sub-modules: IMU data collection 21 and data calculation 22. It is responsible for solving the collected IMU data, obtaining the speed, position and attitude information of the carrier, and establishing the carrier coordinate system to navigate. The pose transition matrix of the coordinate system.

Satellite signal receiving module 3 is made up of two submodules of satellite data receiving 31 and data detection 32, and it is responsible for receiving the data of satellite, and the data of satellite is detected, and output meets the satellite data of detection condition.

The data fusion module 4 is composed of three sub-modules: data fusion 41, data correction 42, and neural network training and prediction 43. The calculated data is corrected to obtain the estimated value of the train's position.

The onboard computer includes data display 61 and data storage 62 . The data display sub-module 61 is responsible for real-time display of the data transmitted to the on-board computer. The displayed data includes: satellite/inertial navigation train positioning data, train trajectory curve, satellite ephemeris observation data and precision factor, and real-time calculation of train attitude angle Curve, train three-axis speed real-time solution curve. The data storage sub-module saves the data transmitted to the on-board computer in the on-board computer for future reference.

The connection relationship of the above modules is as follows: the inertial navigation calculation module 2 is electrically connected to the data fusion module 4; the satellite signal receiving module 3 is electrically connected to the data fusion module 4; the data fusion module 4 is electrically connected to the data output module 5; the data output module 5 is electrically connected to the on-board computer 6. The safety power supply module 1 is electrically connected to the inertial navigation calculation module 2, the satellite signal receiving module 3, the data fusion module 4 and the data output module 5 respectively.

As shown in Figure 2, the core of a kind of satellite/inertial navigation train positioning method based on neural network assistance provided by the present invention is a neural network-assisted train combination positioning algorithm, and this train positioning method can be roughly divided into inertial navigation calculation algorithm, There are three modules of satellite signal detection algorithm and data fusion algorithm. These three modules are independent in structure, but are interrelated in each calculation cycle. The input of the system includes: The three-axis acceleration information output from the inertial measurement unit Together with the three-axis angular rate information and the satellite ephemeris observation data, horizontal precision factor, longitude, latitude, altitude, eastward speed, northward speed, and skyward speed of the satellite receiver, the system outputs the estimated train speed and position.

Among them, the inertial navigation calculation module includes inertial measurement unit (IMU) signal acquisition and inertial navigation data calculation. The inertial measurement unit (IMU) sensor model is 3DM-IMU200A, and the microprocessor model used for inertial navigation calculation is STM32F103C6T6.

The satellite signal receiving module includes satellite data acquisition and data detection, and the satellite data acquisition sensor uses GNSS receiver K700.

The data fusion module includes data fusion, neural network training and prediction module, data correction, and the microprocessor model used by the data fusion module is STM32F103C6T6.

The on-board computer includes data display and data storage. The displayed data are: satellite/inertial navigation train positioning data, train trajectory curve, satellite ephemeris observation data and precision factor, real-time solution curve of train attitude angle, train three-axis speed (xyz Three-axis) solve the curve in real time. The saved data include: estimated train position, position data output by satellite receiver, IMU original signal and corresponding time information. The electrical connection between the data output module and the vehicle-mounted computer adopts RS422 serial communication connection.

It can be seen from the above embodiments that the neural network-assisted satellite/inertial navigation train positioning method and system provided by the present invention can not only improve the accuracy of the train when the satellite signal is good, but also can improve the accuracy of the train when the satellite signal is abnormal. Provide higher-precision train positioning results; the use of neural network-assisted data fusion can effectively suppress the divergence of system positioning errors in the case of satellite abnormalities, and can also improve the reliability and safety of satellite/inertial navigation train positioning systems; train positioning Each module of the system works independently without affecting each other, which can improve the stability and anti-interference of the system.

The satellite/inertial navigation train positioning method and system based on the neural network assistance proposed according to the present invention have been described by way of example with reference to the accompanying drawings. However, those skilled in the art should understand that various improvements can also be made on the basis of not departing from the content of the present invention for the neural network-assisted satellite/inertial train positioning method and system proposed by the present invention. Therefore, the protection scope of the present invention should be determined by the contents of the appended claims.

Claims (8)

1. A satellite/inertial navigation train positioning method based on neural network assistance, comprising: Acquiring real-time sensor data, wherein the real-time sensor data includes inertial measurement unit data and satellite data; The inertial measurement unit data is calculated by the inertial navigation calculation module to obtain the attitude, speed and position of the train; Detecting the satellite data, wherein, detecting whether the satellite ephemeris observation data and the satellite positioning precision factor value meet the set conditions, if the set conditions are met, the satellite data is available, if the set conditions are not met, then This satellite data is not available; When the satellite data is available, the satellite data is fused with the attitude, speed and position information of the train obtained after the calculation of the inertial navigation calculation module to obtain the corrected speed and position of the train; At the same time, the satellite data and the attitude, speed and position information of the train acquired after the calculation by the inertial navigation module are input as a neural network, and the system calculation error value is output as a neural network to train the neural network; When the satellite data is not available, the output value after the training of the neural network is fused with the attitude, speed and position information of the train obtained through the calculation of the inertial navigation calculation module to obtain the corrected speed and position of the train. 2. The satellite/inertial navigation train positioning method based on neural network assistance as claimed in claim 1, wherein, said inertial measurement unit data comprises the three-axis (x, y, z) angular rate of gyroscope output, accelerometer output The three-axis (x, y, z) acceleration; The satellite data includes longitude, latitude, altitude, eastward speed, northward speed and skyward speed output by the satellite receiver. 3. The satellite/inertial navigation train positioning method based on neural network assistance as claimed in claim 1, wherein, in the process of solving the inertial measurement unit data by the inertial navigation calculation module, (1) Speed update The speed update equation is: in, Among them, Δv _m is the specific force increment output by the accelerometer in the time period [t _m-1 ,t _m ]; Δθ _m is the angular increment output by the gyroscope in the time period [t _m-1 ,t _m ], It can be approximately transformed into an angular increment by multiplying it by the sampling interval T _s ; (2) Earth parameter calculation To calculate the principal radius of curvature of the meridian circle and the principal radius of curvature of the Maoyou circle, the calculation formula is as follows: Among them, e is the eccentricity of the ellipse, f is the oblateness of the ellipse, f=1/298.257, _Re is the radius of the earth, R _M is the main curvature radius of the meridian circle, and R _N is the main curvature radius of the Maoyou circle; (3) Location update The position update equation is: Wherein, T is the sampling period, λ is the longitude of the train location, L is the latitude of the train location, and h is the height of the train location; (4) Posture update Use the angular velocity measured by the gyroscope as the quaternion update, and use the acceleration of gravity as the observation of the quaternion to calculate the attitude angle in real time; Among them, the matrix form of the attitude quaternion differential equation is: Among them, T is the sampling interval; q ₀ , q ₁ , q ₂ , and q ₃ represent the quaternion of the attitude; ω _x , ω _y , ω _z represent the three-axis (x, y, z) output of the gyroscope; Normalize the updated attitude quaternion: Wherein, the subscript i=0,1,2,3 represents each value in the attitude quaternion respectively; The attitude angle of the carrier can be calculated according to the normalized quaternion. 4. The satellite/inertial navigation train positioning method based on neural network assistance as claimed in claim 1, wherein, when the satellite data is available, the satellite data and the obtained after solving by the inertial navigation solution module The attitude, speed and position information of the train are fused, and in the process of obtaining the corrected speed and position of the train, (1) Establish the satellite/inertial navigation train positioning system mathematical model: The state equation of the system is: X _k ＝Φ _k/k-1 X _k-1 +Γ _k / _k-1 W _k-1 Among them, X _k represents the state vector at time k, Φ _k/k-1 represents the 15×15-dimensional state one-step transition matrix, Γ _k/k-1 represents the 15×15-dimensional system noise distribution matrix, W _k-1 represents 15×1-dimensional system noise vector, which is a Gaussian white noise vector sequence with zero mean value; The measurement equation of the system is: Z _k ＝H _k X _k +V _k Among them, Z _k represents the system measurement vector, H _k represents the 6×15-dimensional measurement matrix, and V _k represents the 6×1-dimensional measurement noise vector, which is a Gaussian white noise vector sequence with zero mean value; (2) Establish the system error equation The equation for establishing a simplified system error based on the system state matrix is: in, and Respectively represent the angular velocity and error of the n system relative to the i system, and Respectively represent the angular velocity and error of the n system relative to the e system, and Respectively represent the angular velocity and error of the e system relative to the i system, is the direction cosine matrix of the b system relative to the n system; The state one-step transition matrix Φ _k/k-1 can be obtained from the above formula; (3) Data Fusion The data fusion algorithm is divided into two parts, prediction update and measurement update. Prediction update refers to updating the system state at this moment through the established system model and the system state at the previous moment. Measurement update refers to the use of The measurement value updates the system state to obtain the final system state value; after the prediction update, check whether there is a new measurement value, if there is, perform measurement update, if not, directly use the predicted update value Correct the solution value of inertial navigation; (4) Use the system state vector after data fusion to correct the speed and position value of the system: in, are the estimated velocity errors in the three directions of the system, respectively, are the system latitude, longitude, and altitude error estimates, respectively. 5. The satellite/inertial navigation train positioning method based on neural network assistance as claimed in claim 1, wherein, after said satellite data and the attitude, speed and The position information is used as the input of the neural network, and the system calculation error value is used as the output of the neural network. During the training of the neural network, The input value and expected output value of the neural network are respectively: enter: Output: [δv δp] = [δv ^E δv ^N δv ^U δl δλ δh] The input of the neural network is respectively: the output increment of accelerometer eastward and northward, the speed increment of system eastward and northward, pitch angle, roll angle, heading angle error, pitch angle and roll angle; The outputs of the neural network are respectively: the speed error in three directions of the system, the latitude error, the longitude error and the height error. 6. A satellite/inertial navigation train positioning system based on neural network assistance, comprising: The inertial navigation calculation module is used to obtain the data of the inertial measurement unit in real time, and obtain the attitude, speed and position of the train; The satellite signal receiving module is used to obtain satellite data in real time, and detect the satellite data, wherein, detect whether the satellite ephemeris observation data and the satellite positioning precision factor value meet the set conditions, and if the conditions are met, the satellite data Available, if the condition is not met, this satellite data is not available; The data fusion module is used to fuse the satellite data with the attitude, speed and position information of the train obtained after the calculation by the inertial navigation calculation module when the satellite data is available, and obtain the corrected speed and position information of the train. Location; At the same time, the satellite data and the attitude, speed and position information of the train acquired after the calculation by the inertial navigation module are input as a neural network, and the system calculation error value is output as a neural network to train the neural network; When the satellite data is not available, the output value after the training of the neural network is fused with the attitude, speed and position information of the train obtained through the calculation of the inertial navigation calculation module to obtain the corrected speed and position of the train. 7. Auxiliary satellite/inertial navigation train positioning system based on neural network as claimed in claim 6, wherein, also comprising safety voltage module, data output module and on-board computer, The safety voltage module is used to provide power to the inertial navigation calculation module, satellite signal receiving module, data fusion module, and data output module respectively; The data output module is used to transmit the data fused by the data fusion module to the on-board computer. 8. the satellite/inertial navigation train positioning system based on neural network assistance as claimed in claim 6, wherein, the inertial measurement unit sensor model in the described inertial solution module is 3DM-IMU200A, and the used microprocessing of inertial navigation solution The device model is STM32F103C6T6; The GNSS receiver K700 used by the sensor for collecting satellite data; The microprocessor model used in the data fusion module is STM32F103C6T6.