A train satellite positioning fraud warning method, system, device and medium

Abstract

The embodiments of this specification provide a train satellite positioning deception warning method, system, device and medium, wherein the method includes collecting the positioning data of the train autonomous positioning unit under no deception and preset deception attack conditions based on the deception scenario test, and calculating the historical reference quantity of the navigation satellite signal observation characteristics to obtain the first data set; iteratively training the optimized lightweight gradient boosting machine to obtain the train satellite positioning deception state recognition model; iteratively training the long short-term memory network to obtain the train satellite positioning deception situation prediction model; inputting the real-time positioning data of the train autonomous positioning unit into the deception state recognition model to determine the current deception state flag quantity, and when the deception state flag quantity is a deception state, calling the deception situation prediction model to determine the deception trend flag quantity, and determine the degree of deception influence of the train autonomous positioning unit. This solves the problem that the prior art does not have an active protection mechanism for preemptive prevention from the perspective of forecasting and warning.

Description

Technical Field

[0001] The present invention relates to the field of satellite navigation and positioning technology, and in particular to a train satellite positioning deception early warning method, system, equipment and medium. Background Art

[0002] The Global Navigation Satellite System (GNSS) can provide accurate location information and running speed for trains around the clock. However, satellite navigation signals have the characteristics of low signal strength, long propagation distance, and open modulation mode, which makes them susceptible to various external environmental interference and electromagnetic signal interference. In particular, there is a certain amount of artificial malicious deception interference. The deceptors use the known generation method of civil codes as a premise to autonomously transmit deception attack signals similar to the target signal. After the receiver captures the deception signal, the correlation peak of the deception signal gradually removes the correlation peak of the real signal from the tracking loop due to its power advantage, so that the receiver can lock the deception signal and obtain incorrect positioning information, resulting in the deviation of the train position based on satellite positioning, which may lead to the occurrence of safety risks.

[0003] In recent years, relevant research has been conducted on the detection and protection of satellite navigation spoofing at home and abroad, especially in-depth exploration of spoofing protection against moving carriers such as drones, smart cars, and ships. However, satellite positioning applications for rail transit trains have not received effective attention in considering the threat of satellite navigation spoofing and its protection in the preset rail transit operating environment and scenarios.

[0004] At present, domestic and foreign research has proposed related solutions to implement deception detection by adding external detection equipment or adopting highly complex system hardware and software configurations. However, there are problems such as high system implementation cost and complexity, and insufficient feasibility. In addition, the existing technology mainly focuses on the detection and confirmation of the existence of deception, and has not designed a pre-emptive active protection mechanism from the perspective of forecasting and warning. For the use of satellite navigation systems in rail transit systems over a large time and space range and the security of satellite navigation information, it is urgent to provide a train satellite positioning deception warning method that provides enhanced deception status recognition and prediction capabilities based on the existing train satellite positioning technology framework and system equipment. Summary of the invention

[0005] In order to overcome the problems existing in the related art, the present disclosure provides a train satellite positioning deception warning method and system to solve the technical problem that the detection and protection of satellite navigation deception in the related art fails to realize the deception protection technology of the train control system from the perspective of forecasting and warning.

[0006] One or more embodiments of this specification provide a train satellite positioning deception warning method, including the steps of:

[0007] Based on the deception scenario test, the positioning data of the train autonomous positioning unit under no deception and preset deception attack conditions and the calculated historical reference quantity of the navigation satellite signal observation characteristics are collected to obtain a first data set; the data with deception labels in the first data set are screened to obtain a second data set, wherein the historical reference quantity of the navigation satellite signal observation characteristics includes pseudorange residuals, carrier smoothing pseudorange residuals, carrier-to-noise ratio, speed deviation between satellite positioning and axle speed and ranging, deviation between satellite observed pseudorange and axle speed and ranging estimated pseudorange, and satellite positioning precision factor;

[0008] Iteratively training an optimized lightweight gradient boosting machine using a first training data set constructed through the first data set to obtain a train satellite positioning deception state recognition model;

[0009] Iteratively training the long short-term memory network with the first training data set constructed by the second data set obtains a train satellite positioning deception situation prediction model;

[0010] The real-time acquired positioning data of the train autonomous positioning unit is used to calculate the real-time reference quantity of the navigation satellite signal observation characteristics, and input into the deception state recognition model to determine the current deception state flag quantity. When the deception state flag quantity is a deception state, the deception situation prediction model is called based on the calculated real-time reference quantity of the navigation satellite signal observation characteristics to determine the deception trend flag quantity. According to the deception trend flag quantity, the degree of deception impact of the train autonomous positioning unit is determined and deception warning information is generated.

[0011] One or more embodiments of this specification provide a train satellite positioning deception warning system, characterized in that it includes a vehicle-mounted subsystem and a central subsystem;

[0012] The onboard subsystem extracts the information required for modeling, monitoring and prediction from the train autonomous positioning unit and sends the corresponding information to the central subsystem. The onboard subsystem includes an information acquisition and calculation module connected to the train autonomous positioning unit and the track electronic map database, a model training module, a deception status monitoring module and a deception situation prediction module;

[0013] The information acquisition and calculation module is used to collect the positioning data of the train autonomous positioning unit under the conditions of no deception and preset deception attack, and the historical reference quantity of the navigation satellite signal observation characteristics calculated according to the data of the track electronic map database to obtain a first data set; the data with deception labels in the first data set are screened to obtain a second data set, wherein the historical reference quantity of the navigation satellite signal observation characteristics includes pseudorange residual, carrier smoothing pseudorange residual, carrier-to-noise ratio, satellite positioning and axle speed and ranging speed deviation, satellite observation pseudorange and axle speed and ranging estimated pseudorange deviation and satellite positioning precision factor;

[0014] A model training module is used to iteratively train an optimized lightweight gradient boosting machine through a first training data set constructed by a first data set to obtain a train satellite positioning deception state recognition model, and iteratively train a long short-term memory network through a first training data set constructed by a second data set to obtain a train satellite positioning deception situation prediction model;

[0015] The deception state monitoring module is used to determine the current deception state flag quantity based on the train satellite positioning deception state identification model, using the observation sample feature set of each observation satellite in the current period, and send it to the train satellite positioning deception situation prediction model;

[0016] A deception situation prediction module is used to receive information sent by the deception status monitoring module, and based on the train satellite positioning deception situation prediction model, use the observation sample feature subset of each observation satellite in the current period to determine the deception trend mark quantity, and determine the deception impact degree of the train autonomous positioning unit according to the deception trend mark quantity;

[0017] Wireless communication module, used to realize information transmission between the onboard subsystem and the central subsystem of the train satellite positioning deception warning system based on axle speed and distance measurement enhancement;

[0018] Central subsystem: used to receive the train autonomous positioning unit information sent by the on-board subsystem. When the running line is modified or the train autonomous positioning unit runs on a non-original modeled line, it extracts relevant information, recompiles the test script data required for the train satellite positioning deception scenario test, and updates the set train satellite positioning deception state recognition model, train satellite positioning deception situation prediction model or the train satellite positioning deception state recognition model and train satellite positioning deception situation prediction model of the newly built target line, and sends the model information to the on-board subsystem for its call on the corresponding line.

[0019] One or more embodiments of the present specification provide a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the train satellite positioning deception warning method as described above is implemented.

[0020] One or more embodiments of the present specification provide a computer-readable storage medium, which stores a computer program. When the computer program is executed by a processor, the steps of the train satellite positioning deception warning method as described above are implemented.

[0021] The present invention provides a train satellite positioning deception warning method, system, equipment and medium, which have the advantage of aiming to solve the problem of lack of in-transit detection, prediction and warning functions for train satellite positioning deception interference in current applications, relying on navigation satellite observation information in satellite positioning of moving carriers such as railway trains, introducing axle speed and distance measurement information independent of the satellite positioning sensor channel, supplemented by a high-precision track electronic map database, to form a historical reference of navigation satellite signal observation characteristics of different deception attack scenarios, characterize the working state of the satellite positioning receiver under the influence of deception attacks from multiple angles, use data-driven methods to construct a deception state recognition model and a deception situation prediction model, form a train satellite positioning deception warning method and a corresponding system covering the combination of offline modeling, online recognition and active warning, and enhance the active prevention capability and positioning flexibility level of the satellite navigation-based train positioning system against navigation deception.

[0022] The present invention closely follows the existing non-satellite navigation information conditions of the train positioning device, makes accurate and effective judgments on the satellite navigation deception status with relatively low implementation complexity, and further adds active deception risk prediction and deception warning capabilities. It will help provide enhanced deception status recognition capabilities for the existing train satellite positioning technology framework and system equipment, and add active warning and preventive deception protection capabilities, thereby significantly improving the flexibility of train satellite positioning and its comprehensive adaptability to complex operating observation environments, thereby providing solid satellite navigation information security assurance conditions for the next generation of Beidou-based new train control systems, train integrity detection systems and other key equipment. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] In order to more clearly illustrate one or more embodiments of this specification or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in this specification. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.

[0024] Figure 1 A train satellite positioning technology framework and system schematic diagram provided for one or more embodiments of this specification;

[0025] Figure 2 A flow chart of a train satellite positioning deception warning method provided for one or more embodiments of this specification;

[0026] Figure 3 ROC curves obtained by training the optimized lightweight gradient boosting machine for five test sets provided in one or more embodiments of this specification;

[0027] Figure 4A confusion matrix diagram obtained by training an optimized lightweight gradient boosting machine for five test sets provided for one or more embodiments of this specification;

[0028] Figure 5 A schematic block diagram of a train satellite positioning deception warning system provided for one or more embodiments of this specification;

[0029] Figure 6 A schematic diagram of the structure of a computer device provided for one or more embodiments of this specification. DETAILED DESCRIPTION

[0030] In order to enable those skilled in the art to better understand the technical solutions in one or more embodiments of this specification, the technical solutions in one or more embodiments of this specification will be clearly and completely described below in conjunction with the drawings in one or more embodiments of this specification. Obviously, the described embodiments are only part of the embodiments of this specification, not all of the embodiments. Based on one or more embodiments of this specification, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the protection scope of the present invention.

[0031] The embodiment of the present invention is a train satellite positioning deception warning method based on the existing train satellite positioning technology framework and system equipment to provide enhanced deception state recognition and prediction capabilities, wherein reference Figure 1 As shown, the train satellite positioning technology framework and system schematic diagram include: the train is equipped with a train autonomous positioning unit, including an antenna, a satellite positioning receiver, an axle speed and distance measurement interface unit, a positioning computer, a positioning interface unit, etc., wherein the satellite positioning receiver can output positioning data in standard RINEX, NMEA0183 and other protocol formats, and the axle speed and distance measurement interface unit can extract pulse counting information from the axle sensor end and use it to calculate the relative running distance of the train, the speed along the track and other state quantities; the preset deception source sends a deception signal according to the requirements of the deception scenario; the present invention is described in detail below in conjunction with the specific implementation methods and the drawings in the specification.

[0032] Method Embodiment

[0033] According to an embodiment of the present invention, a train satellite positioning fraud warning method is provided, such as Figure 2 As shown, it is a flow chart of the train satellite positioning deception warning method provided in this embodiment. The train satellite positioning deception warning method according to the embodiment of the present invention comprises the steps of:

[0034] Step S1, based on the deception scenario test, collect the positioning data of the train autonomous positioning unit under no deception and preset deception attack conditions, and calculate the historical reference quantity of navigation satellite signal observation characteristics to obtain a first data set; filter the data with deception labels in the first data set to obtain a second data set, wherein the historical reference quantity of navigation satellite signal observation characteristics includes but is not limited to the satellite number of visible satellites, pseudorange residuals, carrier-smoothed pseudorange residuals, carrier-to-noise ratio, speed deviation between satellite positioning and axle speed and ranging, satellite observation pseudorange and axle speed and ranging estimated pseudorange deviation and satellite positioning (spatial, horizontal, vertical) precision factor.

[0035] Step S2: Iteratively train an optimized lightweight gradient boosting machine using a first training data set constructed using the first data set to obtain a train satellite positioning deception state recognition model.

[0036] Step S3, iteratively training the long short-term memory network through the first training data set constructed by the second data set to obtain a train satellite positioning deception situation prediction model.

[0037] Step S4, the real-time acquired positioning data of the train autonomous positioning unit and the real-time reference quantity of the navigation satellite signal observation characteristics are calculated, and the deception state recognition model is input to determine the current deception state flag quantity. When the deception state flag quantity is a deceiving state, the deception situation prediction model is called based on the calculated real-time reference quantity of the navigation satellite signal observation characteristics to determine the deception trend flag quantity. The degree of deception impact of the train autonomous positioning unit is determined according to the deception trend flag quantity and deception warning information is generated.

[0038] The method of this embodiment is based on the navigation satellite observation information in the satellite positioning of moving carriers such as railway trains, introduces the wheel axle speed and distance measurement information independent of the satellite positioning sensor channel, and is supplemented by a high-precision orbit electronic map database to form a historical reference of navigation satellite signal observation characteristics in different deception attack scenarios, so as to characterize the working state of the satellite positioning receiver under the influence of deception attacks from multiple angles, and use data-driven methods to construct a deception state recognition model and a deception situation prediction model, forming a train satellite positioning deception warning process that combines offline modeling, online recognition, and active warning. This method is based on the existing non-satellite navigation information conditions of the train positioning device, with a relatively low The implementation complexity can make accurate and effective judgments on the satellite navigation deception status, and further add active deception risk prediction and deception warning capabilities. This will help provide enhanced deception status identification capabilities for the existing train satellite positioning technology framework and system equipment, and add active warning and preventive deception protection capabilities, thereby significantly improving the flexibility of train satellite positioning and its comprehensive adaptability to complex operating observation environments. It will provide solid satellite navigation information security assurance conditions for the next generation of Beidou-based new train control systems, train integrity detection systems and other key equipment, and solve the problem of lack of in-transit detection, prediction and warning functions for train satellite positioning deception interference in current applications.

[0039] In this embodiment, in step S1, based on Figure 1 The deception scenario test system constructed as shown realizes the collection of positioning data of the train autonomous positioning unit under no deception and preset deception attack conditions. The system includes an on-board subsystem, a central subsystem and a set deception source. The specific functions are:

[0040] On-board subsystem: It runs in parallel with the train's autonomous positioning unit, extracts the information required for modeling, monitoring, and prediction from the train's autonomous positioning unit, and sends the corresponding information to the central subsystem for generating, calculating, or updating the test script data required for the train satellite positioning deception scenario test, and for constructing and updating the deception state recognition model and deception situation prediction model. On the other hand, it sends the train's autonomous positioning unit information to the relevant units of the on-board subsystem for real-time calculation of the real-time reference value of the navigation satellite signal observation characteristics and calling the loaded deception state recognition model and deception situation prediction model to determine the current deception state, and predict the deception situation when confirming that the current state is being deceitful, and generate and output deception warning information.

[0041] Central subsystem: used to receive the train autonomous positioning unit information sent by the on-board subsystem. When the running line is modified or the train autonomous positioning unit runs on a non-original modeling line, it extracts relevant information, recompiles the test script data required for the train satellite positioning deception scenario test, and updates the existing deception state recognition model, deception situation prediction model, or the deception state recognition model and deception situation prediction model of the new target line, and sends the updated or newly created model information to the on-board subsystem for its call in the corresponding line.

[0042] Deception source: Generate corresponding deception signals according to test requirements, including but not limited to no deception, sinusoidal deception, and ramp deception.

[0043] In this embodiment, based on the above system, obtaining the first data set includes the following steps:

[0044] Step S11: During the operation of the train, the high-precision train position reference system carried by the train is used to collect the original scene data of the train satellite positioning, and the train autonomous positioning unit carried by the train is used to simultaneously extract the pulse count information output by the axle speed and distance measurement interface unit to compile the test data required for the train satellite positioning deception scenario test.

[0045] Step S12, call the test script data required for the train satellite positioning deception scenario test, drive the satellite signal generation terminal under the preset deception scenario (including no deception, sinusoidal deception, and ramp deception), start the deception attack at the preset time, and generate and transmit the preset deception scenario navigation satellite test signal by setting different deception powers and deception pseudoranges.

[0046] Step S13: receiving navigation satellite test signals in different scenarios through the train satellite positioning receiver and recording them in the positioning information recording unit.

[0047] In this embodiment, the calculation of the historical reference of the observation characteristics of the navigation satellite signal is mainly achieved by collecting the data of the train autonomous positioning unit under the conditions of no deception and preset deception attack. The main calculation process includes:

[0048] (1) The satellite positioning receiver contained in the train autonomous positioning unit receives the navigation satellite test signal s(t) at time t, which is expressed as:

[0049] s(t)=s re (t)+s sp (t)+n(t) (1);

[0050]

[0051]

[0052] Among them, s re(t) and s sp (t) represents the real satellite signal and the spoofed satellite signal under the preset spoofing attack conditions. If there is no spoofing, it does not contain component s sp (t), n(t) represents the observation noise, M and N represent the number of real signals and spoofing signals received at time t, respectively, and P R and P S denote the average received power of the i-th and j-th satellites, respectively, and α i and α j denote the ranging code contained in the i-th real satellite signal and the ranging code contained in the j-th spoofed satellite signal under the preset spoofing attack condition, τ R and τ S Denote the propagation delay of the real signal and the spoofing signal respectively, D R and D S The data code representing the real signal and the spoofing signal, f1 is the carrier frequency, f dR and f dS are the Doppler shifts of the real signal and the spoofing signal, θ R and θ S Represent the initial carrier phases of the real signal and the spoofing signal respectively.

[0053] The above-mentioned signals received by the satellite positioning receiver contained in the train's autonomous positioning unit undergo multiple processing stages, including capture, tracking and navigation solution, and finally form a positioning solution. The solution results are recorded by collecting data from the train's autonomous positioning unit.

[0054] In this embodiment, during the processing of the satellite positioning receiver, the navigation solution involves different types of observation state information and original observation quantities, such as carrier-to-noise ratio (the ratio of carrier power to noise power of the mixed signal), pseudorange, carrier phase, etc. If the receiver is affected by deception, the relevant characteristics of the above-mentioned observation state information and original observation quantities will change to a certain extent. Based on this, this change can be used to construct historical reference quantity information of navigation satellite signal observation characteristics for deception identification by using historical data of the train autonomous positioning unit collected under no deception and preset deception attack conditions, including but not limited to using GNSS to mark satellite positioning related information, ODO to mark axle speed and distance measurement related information, and selecting pseudorange residuals. Carrier smoothed pseudorange residual Carrier to Noise Ratio (C / N0) t , satellite positioning and wheel axle speed measurement speed deviation Deviation between pseudo-ranges observed by satellite and pseudo-ranges calculated by wheel axle speed measurement Satellite positioning spatial precision factor Satellite positioning horizontal precision factor Satellite positioning vertical precision factor As historical reference quantities of navigation satellite signal observation characteristics, the calculation methods of the above-mentioned reference quantities are as follows.

[0055] 1) Pseudorange residual

[0056] After receiving the satellite signal, the satellite positioning receiver calculates the difference between the corrected pseudorange and the estimated pseudorange to obtain the pseudorange residual. When the satellite positioning receiver is deceived, the corrected pseudorange changes with the pseudorange observation value, thus affecting The value of can be used to characterize whether the receiver is affected by deception, that is,

[0057]

[0058] Among them, ρ p,t To correct the pseudorange, is the corrected pseudorange deviation component due to the spoofing attack, is the estimated pseudorange.

[0059] 2) Carrier smoothed pseudorange residual

[0060] The carrier phase is a measure of the distance between the navigation satellite and the receiver antenna in units of carrier frequency, and is used to evaluate the behavior changes of the navigation satellite signal carrier waveform. The carrier phase observation can be used to smooth the pseudorange and then used as a historical reference for the observation characteristics of the navigation satellite signal, that is,

[0061]

[0062] Among them, ρ s,t is the carrier smoothed pseudorange at time t, φ t is the carrier phase at time t, M is the smoothing time constant, and λ is the wavelength of the corresponding frequency.

[0063] Substituting formula (5) into the residual calculation framework shown in formula (4) yields:

[0064]

[0065] in, is the carrier smoothed pseudorange equivalent deviation component caused by spoofing attack.

[0066] 3) Carrier-to-noise ratio (C / N0) t ;

[0067] Carrier to Noise Ratio (C / N0) tIt is the ratio of baseband signal power to noise power within a unit bandwidth, and can be used to characterize the interference level in the near future. When a satellite positioning receiver is deceived, the deception source device will force the receiver to receive the deception signal, resulting in a stronger baseband signal. In the case of a strong baseband signal, a higher (C / N0) will be obtained. t value, which can be evaluated by (C / N0) t The change degree of the value constructs the observation feature reference, that is,

[0068]

[0069] in, and are the carrier power and noise power in the spoofed scenario, respectively.

[0070] 4) Speed ​​deviation between satellite positioning and wheel axle speed measurement

[0071] In the scenario of deception attack, the speed performance of satellite positioning speed solution may be degraded, while the axle speed and distance measurement information is not affected by the navigation deception signal, and has significant anti-navigation deception capability and insensitivity. Therefore, the speed observation information of the two independent sensor channels of satellite positioning and non-satellite positioning is subtracted, which can be used to quantitatively evaluate the impact of satellite navigation deception in the speed measurement information domain, that is,

[0072] λ S VODO,t =v ODO,t -v s r,t (8);

[0073] Among them, v ODO,t is the train running speed along the track obtained by axle speed measurement at time t, v S r,t It is the ground speed of the train at time t calculated by the on-board satellite positioning receiver in the deception scenario.

[0074] 5) Deviation between pseudo-ranges observed by satellite and pseudo-ranges calculated by wheel axle speed measurement

[0075] The ground speed of the train at time t is calculated by using the axle speed measurement and distance measurement. By integrating the time, the incremental distance traveled by the train within a sampling period can be obtained, which is recorded as δl t , specifically:

[0076]

[0077] From the starting position P in the track electronic map segment determined in the previous cycle t-1 Departure, using the train travel distance increment δl tPredict the train position in this cycle, where the track segment (TrackPiece) involved consists of two adjacent endpoints b m :(N P ,E q ) and b m+1 :(N P+1 ,E q+1 ), where N and E represent latitude and longitude respectively. The specific location prediction process includes the following steps:

[0078] Step A1, according to the train starting position coordinate P1, find the key point b1 with the shortest distance to P1 among the key points of each track contained in the track electronic map;

[0079] Step A2: According to the vector and Calculating vectors exist The projection on L1;

[0080]

[0081] Step A3, calculating the projection position of P1 in the current track segment;

[0082]

[0083] Step A4: Make a judgment based on the moving distance |P1P2| obtained by the wheel axle speed measurement in the first epoch:

[0084] If |P1P2| is less than the current track segment length but

[0085]

[0086] otherwise,

[0087]

[0088] Step A5: Repeat step A4 according to the projection P2' of the final coordinate of the travel distance |P2P3| of the next cycle on the track electronic map until the positioning task is completed.

[0089] Finally, the distance between the wheel axle speed and distance estimation position and the navigation satellite is solved by using the projection of each period position on the orbit electronic map, and the pseudo-range is estimated as follows:

[0090]

[0091] Among them, (x o ,y o ,z o) is the estimated position of the wheel axle speed and distance measurement in the Earth-Centered, Earth-Fixed (ECEF) coordinate system, (x s ,y s ,z s ) is the satellite coordinate position calculated using the ephemeris.

[0092] Since the position and speed estimation of wheel axle speed ranging does not need to receive and use satellite signals, it is not affected by satellite navigation deception. However, the actual observed pseudorange of the receiver will deviate or change to a certain extent due to deception. Therefore, the deviation between the actual satellite observed pseudorange and the wheel axle speed ranging pseudorange can be used to quantitatively evaluate the impact of satellite navigation deception in the ranging observation information domain, that is,

[0093]

[0094] 6) Satellite positioning (spatial, horizontal, vertical) precision factor;

[0095] Satellite positioning precision factor is a quantitative indicator to measure positioning accuracy. It has multiple definitions. Among the different satellite positioning precision factors, satellite positioning spatial precision factor Related to the distribution of navigation satellite azimuth angles, the horizontal precision factor of satellite positioning and vertical precision factor of satellite positioning They are respectively related to the satellite positioning accuracy levels in the horizontal and vertical directions under the current visible distribution conditions of space satellites.

[0096] Combining the above information, a historical reference feature set of navigation satellite signal observation characteristics can be constructed for each observation satellite in each observation period.

[0097] Step S14: collect the positioning data of the train autonomous positioning unit under no-deception and preset deception attack conditions, and calculate the historical reference quantity of the navigation satellite signal observation characteristics, associate them with the corresponding no-deception and deception conditions to form a label quantity set, and aggregate them to form the original data set of the train autonomous positioning unit.

[0098] Specifically, the label quantity set generation process is as follows:

[0099] The historical reference quantities of the navigation satellite signal observation characteristics under no-deception conditions and deception conditions are labeled, for example, no-deception is marked as "0" and deception is "1", and set as the label quantity. The historical reference quantities of the navigation satellite signal observation characteristics are "associated" with the corresponding label quantities to form a "label quantity set". The purpose is that since both no-deception and deception conditions are labeled, whether the output result is consistent with the "label quantity" during the model training process can verify whether the model meets the standard.

[0100] In this embodiment, in order to make the trained model practical for various deception scenarios, the training data needs to include data sets under different deception interference implementation modes, that is, different types of deception scenarios, namely non-deception scenarios, are selected to act on the train respectively, and run on the same track to obtain the original data of the train autonomous positioning unit.

[0101] Based on the above historical data, a first data set is obtained, and the data with deceptive labels (or for example, the label quantity is 1) in the first data set is extracted to obtain a second data set, and the first data set is divided into a first training set and a first test set according to a preset ratio, and the second data set is divided into a second training set and a second test set according to a preset ratio.

[0102] In this embodiment, step S2 iteratively trains the optimized lightweight gradient boosting machine through the first training data set constructed by the first data set to obtain a train satellite positioning deception state recognition model, wherein the training process is the lightweight gradient boosting machine integrated into the Bayesian optimization process - automatically searching for the optimal hyperparameters, which is defined as an optimized lightweight gradient boosting machine, and the training process specifically includes the following steps.

[0103] In one embodiment, the optimized lightweight gradient boosting machine setup and training includes the following steps:

[0104] Step S21, defining the hyperparameter range and the objective function;

[0105] Define the hyperparameter range of the optimized lightweight gradient boosting machine, including the learning rate, the maximum number of leaf nodes in the tree, the maximum tree depth, and select the corresponding loss function to determine the corresponding objective function.

[0106] Set the optimized lightweight gradient boosting machine to train M regression trees, with the goal of learning an additive model

[0107]

[0108] Where, T m (x) is the mth decision tree model.

[0109] Step S22, initializing a sample set;

[0110] Establish a random seed and use the deception state recognition model training set D = {(x i ,y i )}, i = 1, 2, ..., N, train the optimized lightweight gradient boosting machine and calculate the objective function value; where x i is the historical reference feature set of observation features without deception and under preset deception attack. It is a vector, that is, the 8 feature quantities in K1 above, y iis the label quantity, that is, the label quantity that characterizes each reference quantity. No deception is marked as "0" and deception is marked as "1".

[0111] Step S23, establishing an agent model;

[0112] Using the initial sample points, a surrogate model is established using the Gaussian process to predict the relationship between the hyperparameters and the objective function.

[0113] Step S24, selecting the next set of sample points;

[0114] Based on the predictions of the surrogate model and the expected improvement in the sampling function, select the hyperparameter settings to be evaluated next.

[0115] Step S25, iteratively training the model according to the selected hyperparameters;

[0116] Iteratively train a basic optimized lightweight gradient boosting machine with the selected hyperparameters. The training process includes:

[0117] (a) In each iteration, a new decision tree T is trained m (x), so that the model F(x) loss function value is minimized. Assume that there are m-1 decision trees, that is

[0118]

[0119] In one embodiment, a new decision tree T is trained m The process of (x) is as follows:

[0120] ① Calculate the sample gradient G on the current leaf node i and the Hessian matrix H i :

[0121]

[0122]

[0123] ② Find the best split point, that is, for each eigenvalue and each histogram bucket, calculate the split gain and find the maximum split point gain Gain:

[0124]

[0125] Among them, λ is the regularization parameter, G L and G R are the sum of the sample gradients on the left child node and the right child node, respectively, H L and H R are the sample Hessian matrices on the left child node and the right child node respectively.

[0126] ③ Execute splitting, that is, at each split, the leaf node with higher gain is preferentially selected for splitting, splitting it into two new leaf nodes, and obtaining a new decision tree T m (x).

[0127] (b) Update the loss function, that is, according to the new decision tree T m (x), using Taylor expansion to approximate the loss function, we can get the loss function as follows:

[0128]

[0129] G m and H m Substituting into the above formula (20), we get:

[0130]

[0131] Step S26, updating the proxy model;

[0132] Add new hyperparameters and their corresponding objective function values ​​to the existing samples and update the surrogate model.

[0133] Step S27, determining the optimal hyperparameter combination;

[0134] Repeat steps S24-S26 until the iteration number threshold is reached and the objective function converges, and select the hyperparameter combination that maximizes the objective function as the optimal hyperparameter combination.

[0135] In this embodiment, a first test set consisting of a first data set is also included, and the performance of the trained lightweight gradient boosting machine is tested and tuned by the first test set, that is, the optimized lightweight gradient boosting machine is trained using the determined optimal hyperparameter combination, and the performance of the optimized lightweight gradient boosting machine is evaluated using the deception state recognition model test set. When the evaluation model performance indicators obtained from all test sets, such as AUC values, meet the threshold requirements, the optimized lightweight gradient boosting machine obtained by the current training is used as the final deception state recognition model.

[0136] The following is a specific example to illustrate the specific training process of the optimized lightweight gradient boosting machine.

[0137] The specific references for the training set and test set in this example are as follows:

[0138] 1) Training set mode configuration;

[0139] Training set 1 (TRAIN1): no deception scenes.

[0140] Training set 2 (TRAIN2): A spoof signal with an amplitude of 5m, a frequency of 0.01Hz, and a sinusoidal wave power varying at 0.1dB / s.

[0141] Training set three (TRAIN3): a ramp deception scenario with a deception signal slope of 2m / s and a power change of 0.1dB / s.

[0142] 2) Test set mode configuration;

[0143] Test set 1 (TEST1): A spoof signal with an amplitude of 7m, a frequency of 0.01Hz, and a sine wave power varying at 0.3dB / s.

[0144] Test set 2 (TEST2): A spoof signal with an amplitude of 8m, a frequency of 0.01Hz, and a sinusoidal wave power varying at 0.3dB / s.

[0145] Test set 3 (TEST3): A sine wave deception scenario with a spoofing signal amplitude of 8m, a frequency of 0.01Hz, and a power varying at 0.5dB / s.

[0146] Test set 4 (TEST4): A ramp-type deception scenario with a deception signal slope of 2m / s and a deception power varying at 0.3dB / s.

[0147] Test set 5 (TEST5): The deception signal slope is 3m / s, and the deception power changes at a ramp-type deception scene at 0.1dB / s.

[0148] refer to Figure 3 and Figure 4 , respectively, are the ROC curves and confusion matrix diagrams obtained by the trained optimized lightweight gradient boosting machine for the above five test sets. According to the ROC curve diagram, it can be seen that the AUC of each test set accounts for a large proportion, all close to the upper left "1", and in the confusion matrix, the true positive class and true negative class of each test set account for a large proportion, while the false negative class and false positive class account for a small proportion, indicating that the established deception state recognition model has effective classification and recognition capabilities.

[0149] In this embodiment, in step S3, iteratively training the long short-term memory network through the first training data set constructed by the second data set to obtain the train satellite positioning deception situation prediction model includes the following steps.

[0150] Step S31: Filter the data with deception labels in the first data set to obtain the second data set, and reorganize the feature quantity of each sample to obtain the feature subset of the historical reference quantity of the navigation satellite signal observation feature. Construct the first training data set.

[0151] In one embodiment, training set 2 (TRAIN2) and training set 3 (TRAIN3) are used to form a deception situation prediction model training set, wherein the feature quantity of each sample in the deception situation prediction model training set is reorganized, and the satellite number PRN, pseudorange residual, and Carrier smoothed pseudorange residual Deviation between pseudo-ranges observed by satellite and pseudo-ranges calculated by wheel-axle speed measurement Satellite positioning (spatial, horizontal, vertical) precision factor The seven observation feature reference quantities constitute the navigation satellite signal observation feature historical reference quantity feature subset Using the deception scene information of the constructed sample set as the benchmark, the mean pseudo-range error of each visible satellite corresponding to each sample in the test set is calculated. ρ,t , maximum value As the label quantity of each sample, thus forming a complete sample sequence.

[0152] Step S32: Initialize the long short-term memory network model, iteratively train the model to generate a train satellite positioning deception situation prediction model, wherein the model settings are as follows:

[0153] (1) Model initialization: The input layer uses 200 memory units and the hidden layer uses 100 memory units. The dropout layer randomly resets the memory unit weights to 0 with a probability of 0.2 in each training. The output layer is set to a fully connected layer containing 42 neurons. The model uses mean square error (MSE) as the loss function, and the Adam optimizer is selected to optimize the model.

[0154] (2) Forgetting operation: selectively forget the input transmitted by the previous node. The input and the output of the previous step are integrated into an independent vector, and the activation function (the Sigmoid function is used in this embodiment) is used for dot product operation. The calculation process can be expressed as:

[0155] f t =σ(W f ·[h t-1 ,x t ]+b f ) (twenty three);

[0156] Among them, f t is the output of the forget gate, W f is the weight matrix, h t-1 is the hidden state at the previous time step, x t-1 is the current time step input, b f is the bias term and σ is the selected activation function.

[0157] (3) Memory operation: Control whether to incorporate the current time t data into the unit state, use the preset activation function (the sigmoid function is used in this embodiment) to determine the value to be memorized, and use the preset extraction operation function (the tanh function is used in this embodiment) to extract the valid information in the current vector, that is:

[0158] i t =σ(W i ·[h t-1 ,x t ]+b i ) (twenty four);

[0159]

[0160] Among them, W i , W C is the weight matrix, b i 、b C is the bias term.

[0161] (4) Unit status update: The status of each unit is updated to the “memorized” memory information and the newly acquired new information, that is:

[0162]

[0163] Among them, C t , C t-1 They are the current and previous unit states respectively.

[0164] (5) Output operation: Use a preset activation function (the sigmoid function is used in this embodiment) to extract information from the vector obtained by integrating the current value with the output value at the previous moment, and use a preset extraction operation function (the tanh function is used in this embodiment) to compress and map the current unit state to the interval (-1, 1):

[0165] o t =σ(W o ·[h t-1 ,x t ]+b o ) (27);

[0166] h t =o t *tanh(C t ) (28);

[0167] Among them, t is the output of the output gate, h t is the hidden state at the current moment, W o and b o are the weight matrices and bias terms.

[0168] (6) Model training: Based on the above-set model, multiple rounds of iterative training of the model are performed using the historical reference feature set of the second navigation satellite signal observation feature in step S31 until a threshold number of iterations is reached or the objective function converges, thereby obtaining a trained train satellite positioning deception situation prediction model.

[0169] Step S33: using the test set to perform performance inspection and optimization of the train satellite positioning deception situation prediction model. The main process includes:

[0170] (1) Test set 1 (TEST1), test set 2 (TEST2), test set 3 (TEST3), test set 4 (TEST4), and test set 5 (TEST5) are used to form a test set of the second navigation satellite signal observation feature reference quantity feature set and sample label quantity.

[0171] (2) The performance of the trained train satellite positioning deception situation prediction model is evaluated using the test set. When the performance indicators of the evaluation models obtained from all test sets meet the threshold requirements, the candidate deception situation prediction model obtained from the current training is used as the final deception situation prediction model, where the performance indicators include the mean pseudorange error, the root mean square error (RMSE) of the maximum prediction solution, the mean absolute error (MAE) and the determination coefficient (R 2 ); If the performance index of the evaluation model obtained from any test set does not meet the threshold requirement, repeat steps S31-S33 until all conditions are met to determine the final train satellite positioning deception situation prediction model.

[0172] In one embodiment, the implementation process of step S4 is as follows:

[0173] Step S41, obtaining the positioning data of the train autonomous positioning unit in real time, and calculating the real-time reference quantity of the navigation satellite signal observation characteristics;

[0174] Specifically, when the train is in operation, it receives data from the on-board train autonomous positioning unit in real time, parses navigation satellite ephemeris information, navigation satellite observation information, etc. according to the raw data in the corresponding format output by the satellite positioning receiver, and simultaneously extracts the speed and distance information provided by the axle speed and distance measurement interface unit. Based on the track electronic map data, it performs synchronous processing of key information, extracts the required feature quantity information, and calculates the real-time reference quantity of the navigation satellite signal observation feature to form the observation sample feature set of each observation satellite in the current period.

[0175] Step S42: Input the observed sample feature set obtained in step S41 into the trained train satellite positioning deception state recognition model to determine the current deception state flag value F spoof,t , and according to F spoof,t The value determines whether the current train autonomous positioning unit is affected by deception; where Fspoof,t The value settings are as follows:

[0176] (1) If the deception state recognition model outputs F spoof,t =0, the identification tag quantity is "no deception", indicating that the current train autonomous positioning unit based on satellite positioning is not affected by deception;

[0177] (2) If the deception state recognition model outputs F spoof,t =1, the identification tag quantity is the preset deception condition, indicating that the current train autonomous positioning unit based on satellite positioning is affected by deception.

[0178] Step S43: When the current fraud state flag is set to the fraud state, it is determined that F spoof,t = 1, based on the observation sample information at the current moment, the current reference feature subset of the observation feature of the navigation satellite signal is constructed. And use the Min-Max principle to normalize the feature values ​​contained in the current feature subset so that the feature values ​​contained in it are mapped to [0,1], that is,

[0179]

[0180] Among them, x' and x represent any feature quantity belonging to the current feature quantity set after normalization and before normalization respectively. By using this method, the normalized current feature quantity set can be obtained.

[0181] Step S44, input the current reference feature subset of the observation features obtained in step S43 into the train satellite positioning deception situation prediction model to obtain the pseudorange error mean prediction solution and the maximum prediction solution, determine the deception trend marker, and predict the degree to which the current train autonomous positioning unit will be affected by deception.

[0182] Specifically, according to the deception trend indicator P spoof,t The following settings are included:

[0183] (1) If the deception situation prediction model outputs the pseudo-range error mean prediction solution And the maximum prediction solution and are the first-level decision thresholds of the preset pseudorange error mean and maximum values, respectively. Then P spoof,t =1, indicating that it can be predicted that the train autonomous positioning unit based on satellite positioning will be slightly affected by deception in the future period;

[0184] (2) If the pseudo-range error mean prediction solution and the maximum prediction solution output by the deception situation prediction model meet one of the following conditions:

[0185] ① and or

[0186] ② and

[0187] in, and are the preset secondary decision thresholds of the mean and maximum values ​​of pseudorange errors, respectively. Then P spoof,t =2, indicating that it can be predicted that in the future, the train autonomous positioning unit based on satellite positioning will be significantly affected by deception;

[0188] (3) If the deception situation prediction model outputs the pseudorange error mean prediction solution or maximum prediction Then P spoof,t =3, indicating that it can be predicted that in the future, the train autonomous positioning unit based on satellite positioning will be seriously affected by deception.

[0189] Step S44: Generate and output deception warning information based on the deception trend indicator, including secondary warning, major warning, and emergency warning; specifically:

[0190] (1) When the impact of deception in the future is predicted to be minor, that is, P spoof,t =1, the system issues a minor warning sign, indicating that the current deception attack has limited effect or is still in the early stage of being deceived, prompting the train's autonomous positioning unit to trigger the deception effect elimination logic;

[0191] (2) When the impact of deception in the future is predicted to be significant, that is, P spoof,t =2, the system issues a major warning sign, indicating that the current deception attack is obvious or the influence of deception has developed to a significant degradation stage, requiring the train autonomous positioning unit to trigger the anti-deception processing logic;

[0192] (3) When the impact of deception is predicted to be more serious in the future, that is, P spoof,t =3, the system issues an emergency warning sign, indicating that the current deception attack is strong or the impact of the deception has developed to a serious deterioration stage, driving the train autonomous positioning unit to trigger the satellite positioning channel isolation logic. The positioning computer contained in the train autonomous positioning unit does not use the current output information of the satellite positioning receiver for train positioning decisions, and is reconstructed into a mode that uses non-satellite positioning information to implement positioning decisions.

[0193] System Example

[0194] According to an embodiment of the present invention, a train satellite positioning fraud warning system is provided. Figure 1 and Figure 5 As shown, Figure 5The train satellite positioning deception warning system provided in this embodiment is a schematic block diagram. The train satellite positioning deception warning system according to the embodiment of the present invention includes a vehicle-mounted subsystem and a central subsystem, and the specific functions are:

[0195] On-board subsystem: Extracts the required information for modeling, monitoring, and prediction from the train's autonomous positioning unit. On the one hand, the corresponding information is sent to the central subsystem for generating or updating the test script data required for the train satellite positioning deception scenario test, and for building or updating the deception state recognition model and deception situation prediction model. On the other hand, the train's autonomous positioning unit information is sent to the relevant units of the on-board subsystem for real-time calculation of the real-time reference quantity of the navigation satellite signal observation characteristics and calling the loaded deception state recognition model and deception situation prediction model to determine the current deception state. When confirming that the current state is being deceitful, the deception situation is predicted and deception warning information is generated and output.

[0196] The on-board subsystem is equipped with an antenna, a satellite positioning receiver, an axle speed and distance measurement interface unit, a positioning computer and a positioning interface unit arranged on the train autonomous positioning unit. Among them, the satellite positioning receiver can output positioning data in standard RINEX, NMEA0183 and other protocol formats, and the axle speed and distance measurement interface unit can extract pulse counting information from the axle sensor end and use it for the positioning computer to calculate the relative running distance of the train, the running speed along the track and other state quantities.

[0197] Central subsystem: used to receive the train autonomous positioning unit information sent by the on-board subsystem. When the running line is modified or the train autonomous positioning unit runs on a non-original modeled line, it extracts relevant information, recompiles the test script data required for the train satellite positioning deception scenario test, and updates the set train satellite positioning deception state recognition model, train satellite positioning deception situation prediction model or the train satellite positioning deception state recognition model and train satellite positioning deception situation prediction model of the newly built target line, and sends the model information to the on-board subsystem for its call on the corresponding line.

[0198] In one embodiment, the vehicle-mounted subsystem includes an information acquisition and calculation module 10 connected to the train autonomous positioning unit and the track electronic map database, a model training module 20, a deception status monitoring module 30 and a deception situation prediction module 40.

[0199] The information acquisition and calculation module 10 is used to collect the positioning data of the train autonomous positioning unit under no deception and preset deception attack conditions, and the historical reference of the navigation satellite signal observation characteristics calculated according to the data of the track electronic map database to obtain a first data set; the data with deception labels in the first data set are screened to obtain a second data set, wherein the historical reference of the navigation satellite signal observation characteristics includes but is not limited to the satellite number of the visible satellite, the pseudorange residual, the carrier-smoothed pseudorange residual, the carrier-to-noise ratio, the speed deviation between satellite positioning and axle speed and ranging, the satellite observation pseudorange and the axle speed and ranging estimated pseudorange deviation and the satellite positioning (spatial, horizontal, vertical) precision factor.

[0200] The model training module 20 is used to iteratively train the optimized lightweight gradient boosting machine through the first training data set constructed by the first data set to obtain a train satellite positioning deception state recognition model, and iteratively train the long short-term memory network through the first training data set constructed by the second data set to obtain a train satellite positioning deception situation prediction model.

[0201] The deception status monitoring module 30 is used to determine the current deception status flag based on the train satellite positioning deception status recognition model, using the observation sample feature set of each observation satellite in the current period, and send it to the train satellite positioning deception situation prediction model.

[0202] The deception situation prediction module 40 is used to receive the information sent by the deception status monitoring module 30, and based on the train satellite positioning deception situation prediction model, use the observation sample feature subset of each observation satellite in the current period to determine the deception trend mark quantity, and determine the degree of deception impact of the train autonomous positioning unit according to the deception trend mark quantity.

[0203] The wireless communication module 50 is used to realize the information transmission between the on-board subsystem and the central subsystem of the train satellite positioning deception warning system based on the enhanced wheel axle speed and distance measurement.

[0204] The system of this embodiment is based on the navigation satellite observation information in the satellite positioning of moving carriers such as railway trains, introduces the wheel axle speed and distance measurement information independent of the satellite positioning sensor channel, and is supplemented by a high-precision orbit electronic map database to form a historical reference of navigation satellite signal observation characteristics in different deception attack scenarios, so as to characterize the working state of the satellite positioning receiver under the influence of deception attacks from multiple angles, and use data-driven methods to construct a deception state recognition model and a deception situation prediction model, forming a train satellite positioning deception warning process that combines offline modeling, online recognition, and active warning. This method is based on the existing non-satellite navigation information conditions of the train positioning device, with a relatively low The implementation complexity can make accurate and effective judgments on the satellite navigation deception status, and further add active deception risk prediction and deception warning capabilities. This will help provide enhanced deception status identification capabilities for the existing train satellite positioning technology framework and system equipment, and add active warning and preventive deception protection capabilities, thereby significantly improving the flexibility of train satellite positioning and its comprehensive adaptability to complex operating observation environments. It will provide solid satellite navigation information security assurance conditions for the next generation of Beidou-based new train control systems, train integrity detection systems and other key equipment, and solve the problem of lack of in-transit detection, prediction and warning functions for train satellite positioning deception interference in current applications.

[0205] In this embodiment, a satellite positioning information interface submodule, an axle speed and distance measurement information interface submodule and a track electronic map database submodule connected to the information acquisition and calculation module 10 are also provided;

[0206] The satellite positioning information interface module is used to extract satellite positioning related information from the satellite positioning receiver contained in the train autonomous positioning unit, aggregate it to form the original data set of the train autonomous positioning unit, train and update the train satellite positioning deception state recognition model and the train satellite positioning deception situation prediction model. At the same time, it is used to calculate the real-time reference quantity of the navigation satellite signal observation characteristics, constitute the current observation sample feature set, and implement deception state recognition and other functional links.

[0207] The axle speed and distance measurement information interface module is used to extract the information output by the axle speed and distance measurement interface unit from the train autonomous positioning unit, compile the test script data required for the train satellite positioning deception scenario test, and aggregate to form the original data set of the train autonomous positioning unit, train and update the train satellite positioning deception state recognition model and the train satellite positioning deception situation prediction model. At the same time, it is used to calculate the real-time reference quantity of the navigation satellite signal observation characteristics, constitute the current observation sample feature set, and implement deception state recognition and other functional links.

[0208] The track electronic map database module is used to store, update and maintain the track electronic map database of the corresponding line, and to calculate the deviation between the satellite observation pseudorange and the axle speed and distance measurement pseudorange, support the training and update of the train satellite positioning deception state recognition model and the train satellite positioning deception situation prediction model. At the same time, it is also used to support the calculation of the real-time reference quantity of the navigation satellite signal observation characteristics, constitute the current observation sample feature set, and implement deception state recognition and other functional links.

[0209] In this embodiment, a warning information interface module 60 is also provided for triggering the generation of deception warning information according to the degree of deception impact of the train autonomous positioning unit determined by the deception situation prediction module 40, and outputting the deception warning information to the positioning computer contained in the train autonomous positioning unit.

[0210] The embodiment of the present invention is a system embodiment corresponding to the above-mentioned method embodiment. The specific operations of the processing steps or calculation processes of each module or unit can be understood by referring to the description of the method embodiment, which will not be repeated here.

[0211] like Figure 6 As shown, the present invention also provides a computer-readable storage medium on which a computer program is stored. When the computer program is executed by a processor, the train satellite positioning deception warning method in the above-mentioned embodiment is implemented, or when the computer program is executed by a processor, the train satellite positioning deception warning method in the above-mentioned embodiment is implemented.

[0212] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to memory, storage, database or other media used in the embodiments provided in this application can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM) or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. As an illustration and not limitation, RAM is available in many forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM).

[0213] Each embodiment in this specification is described in a progressive manner, and the same or similar parts between the embodiments can be referred to each other, and each embodiment focuses on the differences from other embodiments. In particular, for the device or system embodiment, since it is basically similar to the method embodiment, the description is relatively simple, and the relevant parts can be referred to the partial description of the method embodiment. The device and system embodiments described above are merely schematic, wherein the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of this embodiment. Ordinary technicians in this field can understand and implement it without paying creative labor.

[0214] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the above embodiments, or replace some or all of the technical features therein by equivalents. These modifications or replacements do not make the essence of the corresponding technical solutions deviate from the scope of the technical solutions of the embodiments of the present invention, and the contents not described in detail in the specification of the present invention belong to the common knowledge of those skilled in the art.

Claims

1. A train satellite positioning deception early warning method, characterized in that: Includes steps: Based on the deception scenario test, the positioning data of the train autonomous positioning unit under no deception and preset deception attack conditions are collected, and the historical reference of the navigation satellite signal observation characteristics obtained by calculation are constructed to construct the first data set; The second data set is obtained by filtering the data with spoofing labels in the first data set, wherein the historical reference quantities of the observation characteristics of the navigation satellite signal include pseudorange residuals, carrier smoothing pseudorange residuals, carrier-to-noise ratio, speed deviation between satellite positioning and axle speed and ranging, deviation between satellite observed pseudorange and axle speed and ranging estimated pseudorange, and satellite positioning precision factor; Iteratively training an optimized lightweight gradient boosting machine using a first training data set constructed through the first data set to obtain a train satellite positioning deception state recognition model; Iteratively training the long short-term memory network with the second training data set constructed by the second data set obtains a train satellite positioning deception situation prediction model; The real-time acquired positioning data of the train autonomous positioning unit is used to calculate the real-time reference quantity of the navigation satellite signal observation feature, and is input into the deception state recognition model to determine the current deception state flag quantity. When the deception state flag quantity is a deception state, the deception situation prediction model is called based on the calculated real-time reference quantity of the navigation satellite signal observation feature to determine the deception trend flag quantity. The influence degree of deception of the train autonomous positioning unit is determined according to the deception trend flag quantity, and deception warning information is generated; The first data set construction comprises the following steps: During the operation of the train, the high-precision train position reference system on board the train is used to collect the original scene data of the train satellite positioning. The train autonomous positioning unit on board the train is used to extract the pulse count information output by the axle speed and distance measurement interface unit, and compile the test data required for the train satellite positioning deception scene test. Call the test script data required for the train satellite positioning deception scenario test, drive the satellite signal generation terminal under the preset deception scenario, start the deception attack at the preset time, and generate and transmit the preset deception scenario navigation satellite test signal by setting different deception powers and deception pseudoranges; The navigation satellite test signals in various scenarios are received by the train satellite positioning receiver and recorded in the positioning information recording unit.

2. The train satellite positioning deception early warning method according to claim 1, characterized in that: The satellite positioning receiver contained in the train autonomous positioning unit receives the navigation satellite test signal s(t) at time t as follows: s(t)=s re (t)+s sp (t)+n(t) (1); Among them, s re (t) and s sp (t) represents the real satellite signal and the spoofed satellite signal under the preset spoofing attack conditions. If there is no spoofing, it does not contain component s sp (t), n(t) represents the observation noise, M and N represent the number of real signals and spoofing signals received at time t, respectively, and P R and P S denote the average received power of the i-th and j-th satellites, respectively, and α i and α j denote the ranging code contained in the i-th real satellite signal and the ranging code contained in the j-th spoofed satellite signal under the preset spoofing attack condition, τ R and τ S Denote the propagation delay of the real signal and the spoofing signal respectively, D R and D S The data code representing the real signal and the spoofing signal, f1 is the carrier frequency, f dR and f dS are the Doppler shifts of the real signal and the spoofing signal, θ R and θ S Represent the initial carrier phases of the real signal and the spoofing signal respectively.

3. The train satellite positioning deception early warning method according to claim 1, characterized in that: The calculation of the navigation satellite signal observation characteristic historical reference quantity obtained by calculation is specifically as follows: Pseudorange residuals used to characterize whether the receiver is affected by spoofing The calculation is as follows: Among them, ρ p,t To correct the pseudorange, is the corrected pseudorange deviation component due to the spoofing attack, To estimate the pseudorange; Carrier smoothed pseudorange residual The calculation is as follows: Among them, ρ s,t is the carrier smoothed pseudorange at time t, φ t is the carrier phase at time t, M is the smoothing time constant, and λ is the wavelength of the corresponding frequency; is the carrier smoothed pseudorange equivalent deviation component caused by spoofing attack; Speed ​​deviation between satellite positioning and wheel axle speed measurement and ranging The calculation is as follows: Among them, v ODO,t is the train running speed along the track obtained by axle speed measurement at time t, v S r,t is the ground speed of the train at time t calculated by the onboard satellite positioning receiver in the deception scenario; Deviation between pseudo-ranges observed by satellite and pseudo-ranges calculated by wheel-axle speed measurement The calculation is as follows: Among them, (x o ,y o ,z o ) is the estimated position of the wheel axle speed and distance measurement in the Earth-centered Earth-fixed coordinate system, (x s ,y s ,z s ) is the satellite coordinate position calculated using the ephemeris.

4. The train satellite positioning deception early warning method according to claim 1, characterized in that: The optimized lightweight gradient boosting machine setup and training includes the following steps: 1) Define the hyperparameter range and objective function: Define the hyperparameter range of the optimized lightweight gradient boosting machine, including the learning rate, the maximum number of leaf nodes in the tree, and the maximum tree depth. Select the corresponding loss function and set the optimized lightweight gradient boosting machine to train M regression trees. The goal is to learn an additive model: Where, T m (x) is the mth decision tree model; 2) Initialize the sample set: Initialize the training sample set D = {(x i ,y i )}, i = 1, 2, ..., N, where x i is the historical reference feature set of observation features without deception and under preset deception attack, y i is the label quantity; 3) Establishing a surrogate model: Using the initial sample points, a surrogate model is established using a Gaussian process to predict the relationship between hyperparameters and the objective function; 4) Select the next set of sample points: Based on the prediction results of the surrogate model and the expected improvement sampling function, select the hyperparameter settings that need to be evaluated in the next step; 5) Model training: Iteratively train an optimized lightweight gradient boosting machine using the selected hyperparameters; 6) Update the proxy model: Add the new hyperparameters and their corresponding objective function values ​​to the existing samples and update the proxy model; 7) Determine the optimal hyperparameter combination: Repeat steps 4)-6) until the iteration number threshold is reached and the objective function converges, and select the hyperparameter combination that maximizes the objective function as the optimal hyperparameter combination.

5. The train satellite positioning deception early warning method according to claim 1, characterized in that: The LSTM network is set up as follows: Model initialization: The input layer uses N in memory units, and the hidden layer is N hi memory units, and the dropout layer is used in each training dr The probability of randomly resetting the memory unit weight to 0 is 0, and the output layer is set to contain N ou The fully connected layer has 10 neurons, the loss function uses the mean square error, and the optimizer is the Adam optimizer; Forgetting operation: selectively forget the input passed by the previous node, integrate the input and the output of the previous step into an independent vector, and use the activation function to perform a dot product operation; Memory operation: controls whether to incorporate the current t-time data into the unit state, uses the preset activation function to determine the value that needs to be memorized, and uses the preset extraction operation function to extract valid information in the current vector; Unit state update: The state of each unit is updated with the memory information and the newly acquired new information; Output operation: Use the preset activation function to extract information from the vector of the current value and the output value at the previous moment, and use the preset extraction operation function to compress and map the current unit state to the interval (-1,1); Model training: Based on the initial model, multiple rounds of iterative training of the model are performed through the feature set of historical reference quantities of the second navigation satellite signal observation characteristics until the iteration number threshold is reached or the objective function converges, and a trained train satellite positioning deception situation prediction model is obtained; Model optimization training: Use the deception situation prediction model test set to evaluate the performance of the candidate deception situation prediction model; when the evaluation model performance indicators obtained from all test sets, including the mean pseudorange error, the root mean square error of the maximum prediction solution, the mean absolute error and the determination coefficient, all meet the threshold requirements, the final deception situation prediction model is obtained; if the evaluation model performance indicators obtained from any test set do not meet the threshold requirements, the training is repeated until all conditions are met.

6. The train satellite positioning deception early warning method according to claim 1, characterized in that: The deception status flag is set as follows: If the train satellite positioning deception state recognition model outputs F spoof,t =0, the identification tag quantity is "no deception", which means that the current train autonomous positioning unit based on satellite positioning is not affected by deception; If the train satellite positioning deception state recognition model outputs F spoof,t =1, the identification tag quantity is the preset deception condition, which means that the current train autonomous positioning unit based on satellite positioning is affected by deception.

7. The train satellite positioning deception early warning method according to claim 1, characterized in that: When the deception state flag is a deception state, calling the deception situation prediction model based on the calculated navigation satellite signal observation feature real-time reference to determine the deception trend flag includes the following scenarios: (1) If the train satellite positioning deception situation prediction model outputs the pseudo-range error mean prediction solution And the maximum prediction solution and are the first-level decision thresholds of the preset pseudorange error mean and maximum values, respectively. Then P spoof,t =1, indicating that the train autonomous positioning unit based on satellite positioning is predicted to be slightly affected by deception in the future period; (2) If the train satellite positioning deception situation prediction model outputs the pseudo-range error mean prediction solution and the maximum prediction solution that meet one of the following conditions: ① and or ② and in, and are the preset secondary decision thresholds of the mean and maximum values ​​of pseudorange errors, respectively. Then P spoof,t =2, indicating that it is predicted that in the future period, the train autonomous positioning unit based on satellite positioning will be significantly affected by deception; (3) If the train satellite positioning deception situation prediction model outputs the pseudo-range error mean prediction solution or maximum prediction Then P spoof,t =3, indicating that in the future, the train autonomous positioning unit based on satellite positioning will be seriously affected by deception.

8. The train satellite positioning deception early warning method according to claim 1 or 7, characterized in that: Also includes the steps: Based on the deception trend indicator quantity, deception warning information is generated and output, including minor warnings, major warnings, and emergency warnings.

9. A train satellite positioning fraud warning system, characterized in that: Including vehicle-mounted subsystem and central subsystem; The onboard subsystem extracts the information required for modeling, monitoring and prediction from the train autonomous positioning unit and sends the corresponding information to the central subsystem. The onboard subsystem includes an information acquisition and calculation module connected to the train autonomous positioning unit and the track electronic map database, a model training module, a deception status monitoring module and a deception situation prediction module; The information acquisition and calculation module is used to collect the positioning data of the train autonomous positioning unit under the conditions of no deception and preset deception attack, and the historical reference quantity of the navigation satellite signal observation characteristics calculated according to the data of the track electronic map database to obtain a first data set; the data with deception labels in the first data set are screened to obtain a second data set, wherein the historical reference quantity of the navigation satellite signal observation characteristics includes pseudorange residuals, carrier smoothing pseudorange residuals, carrier-to-noise ratio, satellite positioning and axle speed measurement speed deviation, satellite observation pseudorange and axle speed measurement pseudorange deviation and satellite positioning precision factor; wherein the first data set construction includes the following steps: During the operation of the train, the high-precision train position reference system on board the train is used to collect the original scene data of the train satellite positioning. The train autonomous positioning unit on board the train is used to extract the pulse count information output by the axle speed and distance measurement interface unit, and compile the test data required for the train satellite positioning deception scene test. Call the test script data required for the train satellite positioning deception scenario test, drive the satellite signal generation terminal under the preset deception scenario, start the deception attack at the preset time, and generate and transmit the preset deception scenario navigation satellite test signal by setting different deception powers and deception pseudoranges; The navigation satellite test signals in various scenarios are received by the train satellite positioning receiver and recorded in the positioning information recording unit; A model training module is used to iteratively train an optimized lightweight gradient boosting machine through a first training data set constructed by a first data set to obtain a train satellite positioning deception state recognition model, and iteratively train a long short-term memory network through a second training data set constructed by a second data set to obtain a train satellite positioning deception situation prediction model; The deception state monitoring module is used to determine the current deception state flag quantity based on the train satellite positioning deception state identification model, using the observation sample feature set of each observation satellite in the current period, and send it to the train satellite positioning deception situation prediction model; A deception situation prediction module is used to receive information sent by the deception status monitoring module, and based on the train satellite positioning deception situation prediction model, use the observation sample feature subset of each observation satellite in the current period to determine the deception trend mark quantity, and determine the deception impact degree of the train autonomous positioning unit according to the deception trend mark quantity; Wireless communication module, used to realize information transmission between the onboard subsystem and the central subsystem of the train satellite positioning deception warning system based on axle speed and distance measurement enhancement; Central subsystem: used to receive the train autonomous positioning unit information sent by the on-board subsystem. When the running line is modified or the train autonomous positioning unit runs on a non-original modeled line, it extracts relevant information, recompiles the test script data required for the train satellite positioning deception scenario test, and updates the set train satellite positioning deception state recognition model, train satellite positioning deception situation prediction model or the train satellite positioning deception state recognition model and train satellite positioning deception situation prediction model of the newly built target line, and sends the model information to the on-board subsystem for its call on the corresponding line.

10. The train satellite positioning deception warning system as claimed in claim 9, characterized in that: A satellite positioning information interface submodule, an axle speed and distance measurement information interface submodule and a track electronic map database submodule are also provided to connect with the information acquisition and calculation module; A satellite positioning information interface module is used to extract satellite positioning related information from the satellite positioning receiver contained in the train autonomous positioning unit and aggregate it to form an original data set of the train autonomous positioning unit; The wheel axle speed and distance measurement information interface module is used to extract the information output by the wheel axle speed and distance measurement interface unit from the train autonomous positioning unit, and compile the test script data required for the train satellite positioning deception scenario test; The track electronic map database module is used to store, update and maintain the track electronic map database of the corresponding line, and to calculate the deviation of the pseudo-range calculated by satellite observation pseudo-range and axle speed measurement.

11. The train satellite positioning deception warning system as claimed in claim 9, characterized in that: A warning information interface module is also provided, which is used to trigger the generation of deception warning information according to the degree of deception influence of the train autonomous positioning unit determined by the deception situation prediction module, and output the deception warning information to the positioning computer contained in the train autonomous positioning unit.

12. A computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the computer program, it implements the train satellite positioning deception warning method as described in any one of claims 1 to 8.

13. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the steps of the train satellite positioning deception warning method as described in any one of claims 1 to 8 are implemented.

Images