GNSS and INS integrity monitoring method and system based on 5g and multi-layer fusion mechanism

Abstract

The present application relates to the technical field of multi-source data fusion analysis and monitoring, in particular to a GNSS and INS integrity monitoring method and system based on 5G and multi-layer fusion mechanism, the method comprising: acquiring multi-source data; preprocessing the data based on the data type of the multi-source data, performing multi-source layered fusion processing on the preprocessed multi-source data to obtain uncertainty information; calculating the protection level based on the uncertainty information using a statistical method; performing consistency evaluation on the multi-source data to obtain an evaluation result, calculating the position confidence based on the uncertainty information output by the neural network model and the evaluation result; acquiring multi-dimensional related data, and analyzing and learning the multi-dimensional related data using a neural network to output a risk prediction value. The present application proposes a three-layer fusion mechanism, from the first layer fusion, the second layer fusion to the introduction of a neural network for deep feature extraction and fusion, gradually improving the precision and reliability of the fusion and improving the positioning and navigation performance.

Description

TECHNICAL FIELD

[0001] The application relates to the technical field of multi-source data fusion analysis and monitoring, and particularly relates to a GNSS and INS integrity monitoring method and system based on a 5G and multi-layer fusion mechanism. BACKGROUND

[0002] In positioning and navigation scenarios with high safety requirements, such as automatic driving, aerospace, etc., accurate and reliable positioning information is crucial. Global GNSS navigation satellite systems and INS inertial navigation systems are commonly used positioning means, but they each have limitations. GNSS signals are susceptible to obstructions and interference, which can cause positioning accuracy to decrease or even fail. INS has the problem of error accumulation over time. The development of 5G technology brings new opportunities for positioning and navigation. It has high bandwidth, low latency, and large connectivity, and can provide high-precision positioning information.

[0003] The prior art cannot fully tap the potential value of multi-source data, and therefore, there is an urgent need for a GNSS and INS integrity monitoring scheme based on a 5G and multi-layer fusion mechanism, which aims to fully utilize the advantages of 5G, GNSS and INS, and improve the accuracy and reliability of positioning and navigation. SUMMARY

[0004] The application aims to solve the problems in the background art and proposes a GNSS and INS integrity monitoring method and system based on a 5G and multi-layer fusion mechanism.

[0005] The technical scheme of the application: the GNSS and INS integrity monitoring method based on a 5G and multi-layer fusion mechanism comprises the following steps:

[0006] Obtain GNSS position data, INS position data and 5G position data to obtain multi-source data;

[0007] Preprocess the data based on the data types of the multi-source data, and perform multi-source layered fusion processing on the preprocessed multi-source data to obtain uncertainty information;

[0008] Based on the uncertainty information, a statistical method is used to calculate the protection level;

[0009] Conduct consistency evaluation on the multi-source data to obtain an evaluation result, and calculate the position confidence based on the uncertainty information output by the neural network model and the evaluation result;

[0010] Used for obtaining GNSS signal multi-dimensional related data, INS signal multi-dimensional related data and 5G signal multi-dimensional related data, and using neural network multi-dimensional related data for analysis and learning to output risk prediction values.

[0011] Preferably, the GNSS position data at least includes GNSS position information and speed information; the INS position data at least includes INS position information and speed information; and the 5G position data at least includes initial position information.

[0012] Preferably, the method for multi-source hierarchical fusion processing of the preprocessed multi-source data comprises:

[0013] The preprocessed GNSS position and speed information and the INS position and speed information are fused by using a weighted average algorithm to obtain a first fusion result.

[0014] The initial position information is taken as observation information, and the first fusion result is taken as state prediction information, and state estimation and updating are performed according to the observation information and the state prediction information to obtain a second fusion result.

[0015] Based on the GNSS position data, the INS position data and the second fusion result, deep feature extraction and fusion are performed to output more accurate position information, speed information and attitude information, which are integrated into uncertainty information.

[0016] Preferably, the first layer fusion specifically comprises determining position vectors P GNSS and V GNSS according to the GNSS position information and the speed information respectively, determining P INS and V INS according to the INS position information and the speed information, and fusing the position vectors and the speed vectors based on the following formulas:

[0017] P fusion =w GNSS ×P GNSS +w INS ×P INS ;

[0018] V fusion =w GNSS ×V GNSS +w INS ×V INS ;

[0019] In the formulas, w GNSS and w INS are GNSS weight and INS weight respectively, which are dynamically adjusted based on the measurement accuracy of GNSS and INS.

[0020] Preferably, the second layer fusion specifically comprises taking the position and speed information after the first layer fusion as the initial value of the state vector, and setting the state vector X = [x, y, z, v1, v2, v3] T wherein (x, y, z) represents position coordinates, and (v1, v2, v3) represents corresponding speed components.

[0021] The initial position based on 5G is taken as an observation vector G = [x 5G ,y 5G ,z 5G ] T , wherein (x 5G ,y 5G ,z 5G ) is the initial position coordinate obtained by 5G positioning.

[0022] Preferably, the second layer fusion further comprises a uniform motion model based on a carrier, and a state transition matrix is constructed;

[0023] According to the transition matrix and the state vector at the last time, a state vector at the current time is predicted and marked as a predicted state vector, and the calculation formula of the prediction is as follows:

[0024] X k|k-1 = ZJ·X k-1 ; in the formula, X k|k-1 is the predicted state vector at the current time; X k-1 is the state vector at the last time; k is the time number based on the timestamp; and ZJ is the state matrix.

[0025] Preferably, the second layer fusion further comprises obtaining the state vector obtained after the first layer fusion at the last time and the predicted state vector, calculating the error covariance matrix Z k-1 at the last time, and predicting the error covariance matrix Z k|k-1 at the current time, and the calculation formula is as follows:

[0026] Z k|k-1 = ZJ·Z k-1 + Q; in the formula, Q is the process noise covariance matrix, which is calculated based on the predicted process noise data;

[0027] The Kalman gain K k is calculated according to the error covariance matrix Z k|k-1 at the current time, and the calculation formula is as follows:

[0028] K k = Z k|k-1 ·H T ·(H·Z k|k-1 ·H T + R) -1 ;

[0029] In the formula, H is an observation matrix; and R is an observation noise covariance matrix.

[0030] Preferably, the second layer fusion further comprises calculating the state vector at the current time according to the Kalman gain K k , the observation vector G = [x 5G ,y5G ,z 5G ] T and the predicted state vector X k|k-1 , the updated state vector X k , the calculation formula is as follows:

[0031] X k =X k|k-1 +K k (G k -H·X k|k-1 );

[0032] According to the Kalman gain K k and the error covariance matrix Z k|k-1 of the current moment, the error covariance matrix Z k is updated, and the calculation formula is as follows:

[0033] Z k =(I-K k ·H)·Z k|k-1 ;In the formula, I is a unit matrix;

[0034] The state vector X k of the second layer fused position and speed information is taken as the second fusion result output.

[0035] Based on the GNSS position data, the INS position data and the second fusion result, deep feature extraction and fusion are carried out, and more accurate position information, speed information and attitude information are output, and they are integrated into uncertainty information.

[0036] Preferably, the deep feature extraction and fusion include input layer setting, hidden layer architecture, output layer setting and neural network training.

[0037] The input layer setting includes obtaining the feature quantity of the GNSS position data, the INS position data and the 5G position data and the second fusion result, and performing summation calculation on the feature quantity to determine the input neuron quantity.

[0038] The hidden layer architecture adopts a multi-hidden layer deep neural network structure, each hidden layer uses a ReLU activation function, introduces a nonlinear factor, and enhances the expression ability of the network to complex data relationships.

[0039] The output layer setting includes three output neurons, and the three output neurons respectively output accurate position information, speed information and attitude information;The output layer adopts a linear activation function and directly outputs the predicted numerical result.

[0040] The application also discloses a GNSS and INS integrity monitoring system based on a 5G and multi-layer fusion mechanism, which applies the GNSS and INS integrity monitoring method based on the 5G and multi-layer fusion mechanism, and specifically comprises:

[0041] The data monitoring acquisition module is configured to acquire GNSS position data, INS position data and 5G position data to obtain multi-source data.

[0042] The data preprocessing and fusion module is configured to preprocess the data based on the data types of the multi-source data, and to perform multi-source hierarchical fusion processing on the preprocessed multi-source data to obtain uncertainty information.

[0043] The protection level definition module is configured to calculate a protection level based on the uncertainty information using a statistical method.

[0044] The fixed position reliability calculation module is configured to perform consistency evaluation on the multi-source data to obtain an evaluation result, and to calculate a fixed position reliability based on the uncertainty information output by the neural network model and the evaluation result.

[0045] The related data acquisition and risk prediction module is configured to acquire GNSS signal multi-dimensional related data, INS signal multi-dimensional related data and 5G signal multi-dimensional related data, and to analyze and learn the neural network multi-dimensional related data to output a risk prediction value.

[0046] Compared with the prior art, the above technical solutions of the present application have the following beneficial technical effects:

[0047] The present application proposes a three-layer fusion mechanism, which gradually improves the accuracy and reliability of fusion from the first layer fusion, the second layer fusion to the deep fusion. In particular, in the deep fusion layer, a neural network is introduced for deep feature extraction and fusion, which can better handle the complex relationship between multi-source data and improve the positioning and navigation performance. BRIEF DESCRIPTION OF DRAWINGS

[0048] Figure 1 The method block diagram of the embodiment one proposed by the present application is shown. DETAILED DESCRIPTION

[0049] Embodiment one, as shown in the figure, the GNSS and INS integrity monitoring method based on 5G and multi-layer fusion mechanism proposed by the present application includes the following steps: Figure 1

[0050] Acquire GNSS position data, INS position data and 5G position data to obtain multi-source data.

[0051] ​The GNSS position data at least includes GNSS position information and speed information; the INS position data at least includes INS position information and speed information; and the 5G position data at least includes initial position information; wherein the GNSS position data can adopt a multi-frequency multi-constellation GNSS receiver to receive signals from different satellite constellations and to obtain the signals through conversion and calculation; the INS data is obtained through an inertial measurement unit to collect acceleration and angular velocity information of a carrier and to perform position calculation; and the 5G data is obtained by using a 5G communication module to receive signals from a 5G base station and to obtain position information, signal strength, time of arrival, angle of arrival and other measurement values of the base station;

[0052] The data is preprocessed based on the data types of the multi-source data, including denoising the GNSS position data by using an adaptive filtering algorithm; calibrating the zero offset and the scale factor of the INS position data; and performing multipath suppression and non-line-of-sight identification on the 5G position data; it should be noted that the above data preprocessing is a prior art means, which will not be described in detail here;

[0053] The multi-source hierarchical fusion processing is performed on the preprocessed multi-source data to obtain uncertainty information, and the method includes:

[0054] The GNSS position and speed information and the INS position and speed information after preprocessing are fused by using a weighted average algorithm to obtain a first fusion result;

[0055] The first layer fusion specifically includes determining position vectors P GNSS and V GNSS according to the GNSS position information and speed information respectively, determining P INS and V INS according to the INS position information and speed information, and fusing the position vectors and the speed vectors based on the following formulas:

[0056] P fusion =w GNSS ×P GNSS +w INS ×P INS ;

[0057] V fusion =w GNSS ×V GNSS +w INS ×V INS ;

[0058] In the formulas, w GNSS and w INSGNSS weight and INS weight, respectively, are dynamically adjusted based on the measurement accuracy of GNSS and INS; for example, when the GNSS signal is good, that is, the satellite signal has high signal-to-noise ratio and small GDOP value, the weight of GNSS is increased; when the GNSS signal is blocked or interfered, such as signal-to-noise ratio is reduced and the number of visible satellites is reduced, the weight of GNSS is reduced and the weight of INS is increased;

[0059] The initial position information is used as observation information by using the extended Kalman filtering algorithm, the first fusion result is used as state prediction information, state estimation and updating are performed according to the observation information and the state prediction information, and the second fusion result is obtained;

[0060] The second layer fusion specifically includes taking the position and velocity information after the first layer fusion as the initial value of the state vector, setting the state vector X = [x, y, z, v1, v2, v3] T , wherein (x, y, z) represents the position coordinates, and (v1, v2, v3) represents the corresponding velocity components;

[0061] The initial position based on 5G is taken as the observation vector G = [x 5G ,y 5G ,z 5G ] T , wherein (x 5G ,y 5G ,z 5G ) is the initial position coordinates obtained by 5G positioning;

[0062] Based on the uniform motion model of the carrier, a state transition matrix is constructed; it should be noted that under the assumption of uniform motion, the state transition matrix can be expressed in the form of a matrix combination related to the state vector and the time interval, which is used to predict the state at the next time;

[0063] According to the transition matrix and the state vector at the last time, the state vector at the current time is predicted and marked as the predicted state vector, and the calculation formula of the prediction is as follows:

[0064] X k|k-1 = ZJ·X k-1 ; in the formula, X k|k-1 is the predicted state vector at the current time; X k-1 is the state vector at the last time; k is the time number based on the time stamp; ZJ is the state matrix;

[0065] The state vector obtained after the first layer fusion at the last time and the predicted state vector are obtained, the error covariance matrix Z k-1 at the last time is calculated, the error covariance matrix Z k|k-1 at the current time is predicted, and the calculation formula is as follows:

[0066] Zk|k-1 = ZJ·Z k-1 + Q; wherein Q is a process noise covariance matrix, which is calculated based on the predicted process noise data;

[0067] According to the error covariance matrix Z k|k-1 of the current moment, the Kalman gain K k is calculated, and the calculation formula is as follows:

[0068] K k = Z k|k-1 ·H T ·(H·Z k|k-1 ·H T + R) -1 ;

[0069] wherein H is an observation matrix; R is an observation noise covariance matrix; it should be noted that the role of the observation matrix H is to map the state vector of the system to the observation space, establish the relationship between the state and the observation data, and the observation matrix H is calculated based on the geometric relationship between the signal arrival time, the arrival angle and the position coordinates; through linearization processing of these relationships;

[0070] According to the Kalman gain K k , the observation vector G = [x 5G , y 5G , z 5G ] T and the predicted state vector X k|k-1 , the state vector X k is updated, and the calculation formula is as follows:

[0071] X k = X k|k-1 + K k (G k -H·X k|k-1 );

[0072] According to the Kalman gain K k and the error covariance matrix Z k|k-1 of the current moment, the error covariance matrix Z k is updated, and the calculation formula is as follows:

[0073] Z k = (I-K k ·H)·Z k|k-1 ; wherein I is an identity matrix;

[0074] The state vector X k of the position and velocity information of the second layer fusion is taken as the second fusion result output;

[0075] Based on GNSS location data, INS location data, and the second fusion result, deep feature extraction and fusion are performed to output more accurate location, velocity and attitude information, which are then integrated into uncertainty information.

[0076] Deep feature extraction and fusion includes input layer setup, hidden layer architecture, output layer setup, and neural network training;

[0077] The input layer settings include acquiring the number of features from GNSS location data, INS location data, and 5G location data, and the second fusion result, and summing the number of features to determine the number of input neurons;

[0078] The hidden layer architecture adopts a multi-hidden-layer deep neural network structure. Each hidden layer uses the ReLU activation function to introduce non-linear factors and enhance the network's ability to express complex data relationships.

[0079] The output layer consists of three output neurons, which output precise position, velocity, and attitude information, respectively. The output layer uses a linear activation function to directly output the predicted numerical results.

[0080] Neural network training involves collecting a large amount of multi-source data and its corresponding real position, velocity and attitude information in different scenarios as sample data, and dividing the sample data into training set, validation set and test set;

[0081] The mean squared error loss function is used to calculate the average L of the squared differences between the predicted and true values ​​for each dimension of position, velocity, and attitude. The calculation formula is as follows:

[0082] In the formula, m is the number of training samples; y c i represents the predicted value of the i-th training sample; yi represents the true value of the i-th training sample; it should be noted that the mean squared error loss function is used to measure the difference between the output of the neural network and the true value;

[0083] The Adam optimization algorithm is used to train the neural network. During the training process, the training data is input into the neural network in batches. Each batch of data is processed by forward propagation to calculate the output result. Then, the gradient of the loss function with respect to the parameters of each layer is calculated by backpropagation algorithm. The network parameters are updated by using the Adam optimization algorithm, and the network weights and biases are continuously adjusted to make the loss function converge.

[0084] The trained neural network model is evaluated using a test set, including cross-validation, to tune and optimize the hyperparameters of the neural network. The hyperparameters of the neural network include, but are not limited to, the number of hidden layers, the number of neurons per layer, and the learning rate. For example, the number of hidden layers is changed, and the changes in the loss and accuracy of the model on the validation set are observed to select the hyperparameter combination that optimizes the model performance.

[0085] The trained neural network model is evaluated using a test set, including but not limited to calculating the root mean square error and mean absolute error of the position, velocity, and attitude prediction results, to comprehensively evaluate the model's accuracy and reliability. If the evaluation results do not meet expectations, the hyperparameters can be further adjusted or the amount of training data can be increased to retrain the model.

[0086] Using a trained and optimized neural network model, precise position information, precise velocity information, and precise attitude information are output and integrated into a hierarchical fusion result;

[0087] Based on uncertain information, statistical methods are used to calculate the protection level. These methods include:

[0088] Through formula In the formula, α is a constant related to the confidence level; Z xx These are the diagonal elements of the positional components in the error covariance matrix;

[0089] Consistency assessment is performed on multi-source data to obtain assessment results. Location reliability is then calculated based on the uncertainty information output by the neural network model and the assessment results. Methods include:

[0090] Based on the localization results and uncertainty distribution output by the neural network, the entropy value Q is calculated. The entropy value Q is based on the following formula:

[0091] In the formula, p(xi) is the probability that the positioning result is in state xi; n is the number of possible positioning states;

[0092] Calculate the Euclidean distance between GNSS location information, INS location information, and initial location information, and perform normalization processing to obtain the consistency index C;

[0093] The location reliability R is obtained by weighted fusion of the combined entropy value Q and the consistency index C, as follows:

[0094] R = w1 × (1 - Q) + w2 × C; where w1 and w2 are weights; the weights w1 and w2 are adjusted according to the actual situation to balance the influence of entropy and consistency on location reliability.

[0095] By dynamically adjusting the constant related to the confidence level, the protection level can be accurately calculated according to the actual application scene and safety requirements; in the fixed position confidence calculation, combined with the uncertainty information of the neural network output and the consistency evaluation of the multi-source data, compared with the traditional fixed position confidence calculation which only relies on a single data source or a simple statistical method, the fixed position confidence calculation is more comprehensive and accurate to reflect the credibility of the positioning result;

[0096] The GNSS signal multi-dimensional correlation data, the INS signal multi-dimensional correlation data and the 5G signal multi-dimensional correlation data are acquired, the neural network multi-dimensional correlation data is analyzed and learned, and a risk prediction value is output, and the method comprises the following steps:

[0097] The signal risk value is evaluated, and the following first evaluation model and second evaluation model are used to realize the evaluation;

[0098] The expression of the first evaluation model is In the formula, Out1 is the output value of the first evaluation model, P in is an input parameter, and Pmin and Pmax are the lower limit value and the upper limit value of the standard parameter range respectively;

[0099] The expression of the second evaluation model is In the formula, Out2 is the output value of the second evaluation model; P thr is a parameter correlation threshold value;

[0100] The GNSS signal multi-dimensional correlation data comprises GNSS signal real-time intensity, real-time signal-to-noise ratio and real-time carrier phase residual; the corresponding historical data is acquired, and the signal intensity range, the signal-to-noise ratio range and the carrier phase residual threshold value are determined:

[0101] The signal intensity risk value and the signal-to-noise ratio risk value are calculated based on the first evaluation model, the carrier phase residual risk value is calculated based on the second evaluation model, and a weighted summation operation is performed to obtain the GNSS signal risk value;

[0102] The INS signal multi-dimensional correlation data comprises accelerometer zero offset drift parameters, gyroscope drift parameters and integral errors; the corresponding historical data is acquired, and the accelerometer zero offset drift threshold value and the gyroscope drift threshold value are determined:

[0103] The accelerometer zero offset drift risk value and the gyroscope drift risk value are calculated based on the first evaluation model:

[0104] The accelerometer zero offset drift risk value and the gyroscope drift risk value are calculated based on the first evaluation model:

[0105] The 5G signal multi-dimensional correlation data comprises signal arrival time sequences, arrival angle sequences and signal intensity sequences; the parameter variation rate is determined based on the sequence data; the corresponding historical data is acquired, and the corresponding parameter variation rate threshold value is determined;

[0106] The signal arrival time variation risk value, the arrival angle variation risk value and the signal intensity variation risk value are calculated based on the second evaluation model and are weighted and summed to obtain a 5G signal risk value;

[0107] The input risk value is normalized, a multi-layer fully connected neural network structure is adopted, the input risk value is extracted and transformed through a ReLU function, a Sigmoid activation function is adopted, and a risk prediction value is output;

[0108] By simultaneously processing GNSS, INS, 5G data and environmental information and other multi-source risk factors, the complex relationship between the risk factors is automatically learned to perform risk prediction.

[0109] In embodiment two, the GNSS and INS integrity monitoring system based on the 5G and multi-layer fusion mechanism is applied to the GNSS and INS integrity monitoring method based on the 5G and multi-layer fusion mechanism in embodiment one, and specifically includes:

[0110] A data monitoring acquisition module is configured to acquire GNSS position data, INS position data and 5G position data to obtain multi-source data.

[0111] A data preprocessing and fusion module is configured to preprocess the data based on the data types of the multi-source data, to perform multi-source layered fusion processing on the preprocessed multi-source data, and to obtain uncertainty information.

[0112] A protection level definition module is configured to calculate a protection level based on the uncertainty information by using a statistical method.

[0113] A position fix confidence calculation module is configured to perform consistency evaluation on the multi-source data to obtain an evaluation result, and to calculate a position fix confidence based on the uncertainty information output by the neural network model and the evaluation result.

[0114] A related data acquisition and risk prediction module is configured to acquire GNSS signal multi-dimensional related data, INS signal multi-dimensional related data and 5G signal multi-dimensional related data, to analyze and learn the multi-dimensional related data by using a neural network, and to output a risk prediction value.

[0115] The embodiments of the present application are described in detail above in combination with the drawings, but the present application is not limited thereto, and various changes can be made within the knowledge of those skilled in the art without departing from the purpose of the present application.

Claims

1. A GNSS and INS integrity monitoring method based on 5G and multi-layer fusion mechanism, characterized in that, Includes the following steps: S1. Acquire GNSS location data, INS location data, and 5G location data to obtain multi-source data; S2. Based on the data type of multi-source data, preprocess the data, and perform multi-source hierarchical fusion processing on the preprocessed multi-source data to obtain the uncertainty information reflecting the integration of position, velocity and attitude after three-layer fusion processing of neural network. S3. Based on the aforementioned uncertainty information, a statistical method is used to calculate the protection level, wherein the protection level is calculated using a formula. The calculated protection level PL corresponding to the position error; In the formula, α is a constant related to the confidence level; These are the diagonal elements of the positional components in the error covariance matrix; S4. Perform a consistency assessment on the multi-source data to obtain the assessment results. Calculate the location reliability based on the uncertainty information output by the neural network model and the assessment results. The consistency assessment includes calculating the Euclidean distance between GNSS location information, INS location information, and 5G initial location information, and performing normalization processing to obtain a consistency index C. Based on the entropy value Q obtained from the uncertainty information and the consistency index C, the location confidence R is calculated through weighted fusion. S5. Acquire multi-dimensional correlation data of GNSS signals, multi-dimensional correlation data of INS signals and multi-dimensional correlation data of 5G signals, and use a neural network to analyze and learn the multi-dimensional correlation data to output risk prediction values. Specifically, steps S3 and S4 generate a first integrity monitoring output based on the uncertainty information, and step S5 generates a second integrity monitoring output based on the multidimensional correlation data.

2. The GNSS and INS integrity monitoring method based on 5G and multi-layer fusion mechanism according to claim 1, characterized in that, GNSS location data includes at least GNSS location information and velocity information; INS location data includes at least INS location information and velocity information; 5G location data includes at least initial location information.

3. The GNSS and INS integrity monitoring method based on 5G and multi-layer fusion mechanism according to claim 1, characterized in that, Methods for performing multi-source hierarchical fusion processing on preprocessed multi-source data include: A weighted average algorithm is used to perform a first-level fusion of the preprocessed GNSS position and velocity information with the INS position and velocity information to obtain the first fusion result. The extended Kalman filter algorithm is used to take the initial position information as the observation information and the first fusion result as the state prediction information. The state is estimated and updated based on the observation information and the state prediction information to obtain the second fusion result. Deep feature extraction and fusion are performed based on GNSS location data, INS location data, and the second fusion result to output more accurate location, velocity, and attitude information, which are then integrated into uncertainty information.

4. The GNSS and INS integrity monitoring method based on 5G and multi-layer fusion mechanism according to claim 3, characterized in that, The first layer of fusion specifically includes determining the position vector based on GNSS position information and velocity information, respectively. and Determined based on INS location and speed information and The position vector and velocity vector are fused based on the following formula: ； ； In the formula, and These are the GNSS weights and INS weights, respectively, which are dynamically adjusted based on the measurement accuracy of GNSS and INS.

5. The GNSS and INS integrity monitoring method based on 5G and multi-layer fusion mechanism according to claim 4, characterized in that, The second layer of fusion specifically involves using the position and velocity information fused from the first layer as the initial values ​​of the state vector, and setting the state vector... Where (x, y, z) represent position coordinates, and (v1, v2, v3) represent the corresponding velocity components; Using the initial position based on 5G as the observation vector ,in, This represents the initial location coordinates obtained from 5G positioning.

6. The GNSS and INS integrity monitoring method based on 5G and multi-layer fusion mechanism according to claim 5, characterized in that, The second layer of fusion also includes a uniform motion model based on the carrier, constructing a state transition matrix; Based on the transition matrix and the state vector from the previous time step, predict the state vector at the current time step and label it as the predicted state vector. The calculation formula for the prediction is as follows: In the formula, This is the predicted state vector at the current moment; ZJ is the state vector of the previous time step; k is the time number based on the timestamp; ZJ is the state matrix.

7. The GNSS and INS integrity monitoring method based on 5G and multi-layer fusion mechanism according to claim 6, characterized in that, The second-layer fusion also includes obtaining the state vector obtained after the first-layer fusion at the previous time step and the predicted state vector, and calculating the error covariance matrix at the previous time step. Predict the error covariance matrix at the current time. The calculation formula is as follows: In the formula, Q is the process noise covariance matrix, which is calculated based on the predicted process noise data; Based on the error covariance matrix at the current time Calculate Kalman gain The calculation formula is as follows: ； In the formula, H is the observation matrix; R is the observation noise covariance matrix.

8. The GNSS and INS integrity monitoring method based on 5G and multi-layer fusion mechanism according to claim 7, characterized in that, The second-layer fusion also includes adjustments based on Kalman gain. Observation vector and predicted state vector Update the state vector The calculation formula is as follows: ； in, This represents the observation vector, which is the actual measurement value obtained from the external sensor at time k. ; According to Kalman gain Error covariance matrix at the current time Update error covariance matrix The calculation formula is as follows: ； In the formula, I is the identity matrix; The state vector containing the position and velocity information fused from the second layer As the second fusion result output; Deep feature extraction and fusion are performed based on GNSS location data, INS location data, and the second fusion result to output more accurate location, velocity, and attitude information, which are then integrated into uncertainty information.

9. The GNSS and INS integrity monitoring method based on 5G and multi-layer fusion mechanism according to claim 8, characterized in that, Deep feature extraction and fusion includes input layer setup, hidden layer architecture, output layer setup, and neural network training; The input layer settings include acquiring the number of features from GNSS location data, INS location data, and 5G location data, and the second fusion result, and summing the number of features to determine the number of input neurons; The hidden layer architecture adopts a multi-hidden-layer deep neural network structure. Each hidden layer uses the ReLU activation function to introduce non-linear factors and enhance the network's ability to express complex data relationships. The output layer consists of three output neurons, which output precise position, velocity, and attitude information, respectively. The output layer uses a linear activation function to directly output the predicted numerical results.

10. A GNSS and INS integrity monitoring system based on 5G and multi-layer fusion mechanism, applied to the GNSS and INS integrity monitoring method based on 5G and multi-layer fusion mechanism as described in any one of claims 1 to 9, characterized in that, Specifically, it includes: The data monitoring and acquisition module is used to acquire GNSS location data, INS location data, and 5G location data to obtain multi-source data. The data preprocessing and fusion module is used to preprocess the data based on the data type of the multi-source data, and to perform multi-source hierarchical fusion processing on the preprocessed multi-source data to obtain the uncertainty information reflecting the integration of position, velocity and attitude after three-layer fusion processing of the neural network. The protection level definition module is used to calculate the protection level based on the uncertainty information using statistical methods. The protection level is calculated using a formula. The calculated protection level PL corresponding to the position error; where α is a constant related to the confidence level; These are the diagonal elements of the positional components in the error covariance matrix; The location reliability calculation module is used to perform consistency assessment on multi-source data, obtain assessment results, and calculate location reliability based on the uncertainty information output by the neural network model and the assessment results. The consistency assessment includes calculating the Euclidean distance between GNSS location information, INS location information, and 5G initial location information, and performing normalization processing to obtain a consistency index C. Based on the entropy value Q obtained from the uncertainty information and the consistency index C, the location reliability R is calculated through a weighted fusion method. The relevant data acquisition and risk prediction module is used to acquire multi-dimensional related data of GNSS signals, INS signals, and 5G signals, and to use neural networks to analyze and learn the multi-dimensional related data to output risk prediction values.

Images