A train positioning method based on ultra-wideband enhanced Beidou and virtual balise

Abstract

The present invention provides a train positioning method based on ultra-wideband enhanced Beidou and virtual transponders. The method includes: establishing a virtual transponder basic database and an ultra-wideband device basic database based on the collected data of the target track line; establishing a positioning observation quality basic database based on the Beidou positioning observation quality of the target track line, and compiling an ultra-wideband / Beidou joint positioning strategy set; dynamically selecting a positioning strategy based on the real-time observation conditions of the Beidou signal during the train operation, calculating the current three-dimensional spatial position of the train, obtaining the ultra-wideband positioning solution data source of the train, calculating the train's mileage information at the next moment, determining the train's mileage and pre-capture trigger time for the next positioning solution cycle, and sending position correction information at the capture time of the nearest target virtual transponder to be captured. The method of the present invention combines ultra-wideband technology with Beidou positioning, effectively improving the capture accuracy and stability of virtual transponders in complex environments.

Description

Technical Field

[0001] The present invention relates to the technical field of rail transit train operation control, and in particular to a train positioning method based on ultra-wideband enhanced Beidou and a virtual transponder. Background Art

[0002] The virtual balise uses devices such as Beidou positioning terminals to obtain information such as the train's real-time position and speed. By comparing the train's position calculated by Beidou positioning with a preset virtual reference position, the train determines the exact time it passes through the reference position. When the train approaches the real balise's position, the onboard device triggers internal logic, using the same interface and information structure as the real balise to report the train's exact position and execute balise message transmission. In this process, ensuring continuous and accurate acquisition of the train's real-time position and, based on this, seamlessly and accurately determining whether the train has passed the preset reference position is a key condition for timely triggering the virtual balise function and ensuring normal train operation. To this end, seamless capture of virtual balises requires the system to provide stable, accurate, and reliable virtual balise capture services under various environmental conditions, avoiding missed captures and maintaining high capture accuracy. Seamless capture can be divided into two levels: at the positioning level, the system ensures reliable and usable positioning information under all circumstances, avoiding positioning failures and interruptions caused by changes in Beidou signal observation quality; and at the capture level, the system avoids possible missed captures while maintaining the highest possible capture accuracy.

[0003] Ultra-wideband (UWB) technology is valued for its high-precision positioning capabilities in complex environments. As an effective regional non-Beidou positioning method, it can locate a variety of elements in its deployment area, including personnel, in addition to trains. UWB-based positioning has been used to locate trains and personnel on rail transit lines such as the Beijing Subway. Many smart stations and construction sites have introduced UWB to provide regional positioning for equipment and personnel, demonstrating the advantages and value of UWB positioning technology in railway systems. Considering the core characteristics and key advantages of UWB positioning technology, as well as its significantly lower cost compared to Beidou alternative / augmentation solutions such as Beidou pseudo-satellites, it has significant potential for development and application as an effective Beidou positioning enhancement method for specific regions. For traditional Beidou-based virtual balise solutions, introducing the regional enhancement mechanism provided by UWB positioning under specific conditions, such as limited Beidou service or degraded Beidou positioning performance, will effectively ensure a continuous and reliable source of basic information for virtual balise implementation, thereby supporting seamless optimization of virtual balise functionality.

[0004] Existing technologies focus on the typical scenarios in railway environments and study the precise single-point positioning method for trains under conditions of limited satellite signals. For railway environments, on-board sensors such as INS and ODO are selected to build a satellite augmentation system, and accurate GNSS / INS and INS / ODO mathematical models and error equations are established. A GNSS / INS / ODO combined positioning method is further constructed, and digital track maps are used to assist positioning to meet the continuous positioning requirements of trains in different environments.

[0005] One existing virtual balise capture scheme operates based on a fixed capture radius (Capture Interval, CI). The virtual balise detects the train's distance from its true position within a radius R, centered around a pre-set virtual reference point. The balise transmits a corresponding balise message the first time it appears within the capture circle.

[0006] Another virtual transponder capture scheme in the prior art includes using a GPS receiver to provide positioning information, using a fitting function to extrapolate the speed of several points when the GPS cannot provide positioning information, or reinitializing after receiving the GPS positioning information again, and using the current speed model to dynamically predict the speed at the current moment to determine whether the train is within the capture range of the virtual transponder at the current moment, thereby achieving the capture of the virtual transponder.

[0007] Another virtual balise capture solution in the prior art uses an improved Sage-Husa adaptive filtering algorithm to reduce satellite positioning errors, providing accurate train positioning data for virtual balise capture. It adopts a composite map matching method, uses different map matching algorithms according to the different characteristics of the railway lines, and uses a data interpolation method based on the Kriging algorithm to increase sample points to reduce the risk of missed virtual balise capture.

[0008] The disadvantages of the virtual transponder capture solution in the above prior art include:

[0009] (1) In existing BeiDou replacement / augmentation solutions, inertial sensors, odometers, and other technologies all have the problem of error accumulation. In the context of reducing the use of physical transponders and BeiDou being limited, there is a lack of information capabilities to accurately calibrate the accumulated errors.

[0010] (2) The existing virtual transponder capture scheme has insufficient environmental adaptability and does not fully consider the switching operation of trains in different environments and the treatment of complex factors that may affect Beidou positioning performance, such as Beidou signal interference. It also lacks a dynamic compensation mechanism when Beidou signal performance degrades in complex operating environments.

[0011] (3) The capture accuracy of existing virtual transponders is limited, and the use of a fixed capture radius makes it difficult to balance the accuracy and the risk of missed capture, especially in complex environments where the risk of missed capture is exacerbated. Summary of the Invention

[0012] The present invention provides a train positioning method based on ultra-wideband enhanced Beidou and virtual transponder to achieve accurate capture of the target virtual transponder by the train, providing higher protection for continuous positioning and safety monitoring of on-board equipment.

[0013] In order to achieve the above-mentioned purpose, the present invention adopts the following technical solutions.

[0014] A train positioning method based on ultra-wideband enhanced Beidou and virtual balise, comprising:

[0015] Establish a virtual balise basic database and an ultra-wideband device basic database based on the collected target track line data;

[0016] Establish a basic positioning observation quality database based on the Beidou positioning observation quality of the target track line, and compile an ultra-wideband / Beidou joint positioning strategy set based on prior terrain environment information, Beidou positioning observation quality exploration information, and historical train operation data along the line;

[0017] During the train operation, a positioning strategy is dynamically selected from the UWB / Beidou joint positioning strategy set according to the real-time observation conditions of the Beidou signal, the three-dimensional spatial position of the train is calculated, and multiple UWB base station observation data sets are extracted based on the current mileage of the train and the UWB device spatial data in the UWB device basic database to obtain the UWB positioning solution data source of the train;

[0018] Constructing a train state characteristic parameter sequence, establishing a train mileage prediction network model, inputting the train state characteristic parameter sequence during the real-time operation of the train into the train mileage prediction network model, and obtaining the train mileage information at the next moment;

[0019] Extracting information related to the current train and the mileage information of the train at the next moment from the virtual balise basic database, determining the train mileage and pre-capture trigger time of the next positioning solution cycle, calculating the predicted position of the train closest to the nearest target virtual balise from the pre-capture trigger time, estimating the capture time difference of the nearest target virtual balise, and determining the capture time of the nearest target virtual balise to be captured;

[0020] The position correction information is sent at the capture moment of the nearest target virtual balise to be captured, and the train corrects the current position according to the position correction information and updates the real-time status sequence of the virtual balise at the same time.

[0021] Preferably, the step of establishing a virtual transponder basic database and an ultra-wideband device basic database based on the collected target track line data includes:

[0022] Set the target operation route from the starting station A to the terminal station B, and its mileage range is from the starting station center mileage S A Mileage to terminal station center S B , build a basic database of virtual transponders and ultra-wideband equipment covering the entire mileage range of the line;

[0023] The virtual transponder basic database includes track line spatial data, virtual transponder identity sequence and virtual transponder spatial data. The track line spatial data describes the spatial data of the track line between the starting station A and the terminal station B. Combined with the actual positioning measurement, data processing and format conversion of several key points on each track line in the interval and station, a complete track line spatial database is formed according to a specific data item format. The database stores the spatial information and attribute information of all key points on the line. index,λ The spatial data items stored in the key points of the route include: key point mileage S (P index,λ ), three-dimensional coordinate position of key points on the line (L index,λ ,B index,λ ,H index,λ ), the stored attribute data items include: running direction D (P index,λ ), terrain environment attributes U(P index,λ ), the subscript index indicates the key point number of the line, the subscript λ indicates the line number, and the key point mileage S(P index ) is directly related to the up and down directions of the line. The mileage in the up direction decreases, and the mileage in the down direction increases. For any two adjacent key points of the line, the numbers are P α 、P α+1 , the following conditions are met:

[0024] dis{P α (L α ,B α ,H α ),P α+1 (L α+1 ,B α+1 ,H α+1 )}-|S(P α )-S(P α+1 )|≤h s

[0025] Where dis{*} represents the straight-line distance solution function between two points based on three-dimensional space coordinates, h S is the tolerable orbital space data error limit;

[0026] The virtual transponder identity sequence describes the interval from the origin station A to the terminal station B and all N AB The identity data of the virtual transponder includes: target line index number λ, direction flag bit D N and virtual transponder number b i ;

[0027] The virtual transponder space data items include: virtual transponder number b i , virtual transponder mileage and the three-dimensional coordinate position of the virtual transponder

[0028] The UWB basic database contains UWB equipment spatial data, specifically including UWB base station number A j , ultra-wideband base station mileage and the three-dimensional coordinate position of the ultra-wideband base station

[0029] Preferably, the method of establishing a positioning observation quality basic database based on the Beidou positioning observation quality of the target track line, and compiling an ultra-wideband / Beidou joint positioning strategy set based on prior terrain environment information, Beidou positioning observation quality exploration information, and historical train operation data along the line includes:

[0030] Use the historical driving data along the target track line to explore the quality of Beidou positioning observations, and establish a basic positioning observation quality database to store Beidou positioning quality observation data, including: Beidou satellite signal-to-noise ratio (SNR) mean avg (P index,λ ), the number of visible BeiDou satellites n(P index,λ ), Position Dilution of Precision PDOP(P index,λ ), average pseudorange residual e P,avg (P index,λ ), Beidou satellite signal-to-noise ratio standard deviation σ SNR (P index,λ ) and the average signal-to-noise ratio decrease of Beidou satellites SNR drop (P index,λ );

[0031] Based on BeiDou positioning observation quality detection information and historical train operation data of multiple lines, the target track line is analyzed for various possible BeiDou observation quality conditions, including normal BeiDou signal quality, degraded BeiDou signal quality, and BeiDou signal failure.

[0032] The positioning strategy content Method contained in the ultra-wideband / Beidou joint positioning strategy set is determined according to the common Beidou observation quality of the line. The positioning strategy content Method includes three types: ultra-wideband independent positioning (UWB), Beidou independent positioning (GNSS), and ultra-wideband / Beidou joint positioning (UWB / GNSS).

[0033] Preferably, the method of dynamically selecting a positioning strategy in the ultra-wideband / Beidou joint positioning strategy set according to the real-time observation conditions of the Beidou signal during the train operation to calculate the three-dimensional spatial position of the train currently running includes:

[0034] During the real-time operation of the train, at time t, the train control onboard equipment receives the signal from the BeiDou positioning receiver and processes it to obtain the current navigation BeiDou ephemeris and the pseudo-ranges of n visible BeiDou satellites {ρ1(t),ρ2(t),…,ρ n (t)} real-time observation data, using the extracted spatial data of m ultra-wideband devices to obtain the relative distance information between the device and the current train {d1(t), d2(t), …, d m (t)};

[0035] During each positioning solution cycle, the system determines whether the Beidou signal meets the availability conditions:

[0036] SNR avg ≥SNR min and n≥n min

[0037] Determine the signal quality status of the Beidou signal that meets the availability conditions, and dynamically select a positioning strategy in the UWB / Beidou joint positioning strategy set based on the signal quality status of the Beidou signal

[0038] C) Signal quality is normal

[0039] PDOP≤PDOP max and e p,avg ≤e p,max and

[0040] If the conditions are met, the Beidou independent positioning strategy is selected;

[0041] D) Signal quality degradation due to occlusion

[0042] PDOP>PDOP max or e p,avg >e p,max

[0043] If the conditions are met, the UWB / Beidou joint positioning strategy is selected;

[0044] C) Signal quality degradation due to interference

[0045] or SNR drop >SNR drop,max

[0046] If the conditions are met, the UWB / Beidou joint positioning strategy is selected in real time;

[0047] If the system determines that the BeiDou signal does not meet the availability conditions, it selects the ultra-wideband independent positioning strategy;

[0048] Construct the measurement vector z according to the selected positioning strategy and real-time observation information t The train's three-dimensional position {X(t), Y(t), Z(t)} and the clock error Δt constitute the state vector to be estimated, and the state vector estimate is calculated using a nonlinear state filter. Get the three-dimensional position of the train

[0049] Preferably, the extracting of multiple ultra-wideband base station observation data sets based on the current running mileage of the train and the ultra-wideband device spatial data in the ultra-wideband device basic database to obtain the ultra-wideband positioning solution data source of the train includes:

[0050] When the Beidou signal quality is normal, the Beidou independent positioning strategy is used to construct the measurement vector:

[0051] z t =[ρ1(t),ρ2(t),…,ρ n (t)] T

[0052] When the BeiDou signal is still available but the quality is degraded, the UWB / BeiDou joint positioning strategy is used to construct the measurement vector:

[0053] z t =[ρ1(t),ρ2(t),…,ρ n (t),d1(t),d2(t),…,d m (t)] T

[0054] When the BeiDou signal fails, the ultra-wideband independent positioning strategy is used to construct the measurement vector:

[0055] z t =[d1(t),d2(t),…,d m (t)] T

[0056] The state vector to be estimated is composed of the three-dimensional position of the train {X(t), Y(t), Z(t)} and the clock error Δt:

[0057] xt =[X(t),Y(t),Z(t),Δt] T

[0058] Combined with the three-dimensional coordinate position of each visible Beidou satellite {X i (t),Y i (t),Z i (t)} Write the pseudo-range observation model based on the simplified three-dimensional coordinate position of the connected ultra-wideband device Write the ranging model and use the nonlinear Kalman filter to calculate the state vector estimate State vector estimate The specific process includes the following steps:

[0059]

[0060]

[0061] P t =(IK t H t )P t,t-1

[0062] in, P t,t-1 are the state vector prediction value, the one-step prediction variance matrix, and F t,t-1 is the state transfer matrix, K t is the equivalent filter gain matrix, Q t 、R t are the variance matrices of system error and measurement error, respectively, H t is the linearized equivalent measurement matrix obtained by transforming the observation model, P t is the state estimation variance matrix;

[0063] Set up dynamic iterative optimization judgment logic, and use the current train operation information and observation information at time t to iteratively optimize and obtain the optimal solution for train position estimation. Estimate the optimal train position Convert to coordinate position Get the track key point P closest to the current position of the train index,λ , determine the train's running direction and line number;

[0064] The state vector estimate obtained by Kalman filtering is used as the initial state of the dynamic iterative optimization judgment logic Set the optimization objective function X * To minimize the sum of squared residuals, three optimization objective functions are specifically corresponding to the three different strategies in the selected ultra-wideband / Beidou joint positioning:

[0065]

[0066] Where: e UWB,j =d j -||xp UWB,j || is the ranging residual of the jth ultra-wideband base station, d j is the actual measured ultra-wideband ranging distance, x is the train position to be optimized, and p UWB,j is the location coordinate of the known ultra-wideband base station; is the pseudorange residual of the i-th BeiDou satellite, is the Doppler velocity residual; and is the corresponding covariance matrix; x GNSS and x UWB is the train position status to be optimized calculated based on BeiDou and UWB, e Fus is the Euclidean distance difference between BeiDou and UWB positioning results, e Fus =||x GNSS -x UWB ||, ∑Fus is the covariance of the fusion residual;

[0067] Use nonlinear least squares method to perform state iterative optimization to determine whether the optimization accuracy is met:

[0068] ||e(x (k) )-e(x (k-1) )||<ε r

[0069] Where, ε r is the maximum acceptable variation of the residual, and if the above formula is satisfied, the optimal solution of the optimized train position estimation is output. If the above formula is not satisfied and the maximum number of iterations has been reached, the iteration is stopped and the current optimal estimation result is output;

[0070] Depend on Iteratively solve the corresponding geographic coordinates of the train Based on the current coordinate position of the train Determine the track key point P closest to the current position of the train index,λ , thereby determining the train's running direction D(t) and the line number λ;

[0071] Based on the train running status, the track line spatial data in the virtual balise basic database is extracted, and the optimal train position is estimated based on the calculated Use the nearest neighbor principle to find the segment between key points of the track closest to the optimal position estimate in the track line spatial data Calculate the current train mileage S(t):

[0072] S(t)=S(PF )±γ||P F -P E ||

[0073] Where S(P F ) is the key point P of the line F Mileage, γ is the projection coefficient, ||P F -P E || is a line segment length;

[0074] At the time t when the nearest virtual transponder to be captured is captured VB , use the correction information provided by the nearest virtual balise to be captured to update the current position of the train, and replace the current train's optimal position estimate with the three-dimensional coordinate position of the nearest virtual balise to be captured Update mileage at the same time Implement correction information integration in the position calculation process;

[0075] According to the current train mileage S(t) and the ultra-wideband device spatial data in the ultra-wideband device basic database, multiple ultra-wideband base station observation data sets are extracted according to the distance principle to obtain the ultra-wideband positioning solution data source of the train.

[0076] Preferably, the process of constructing a train state characteristic parameter sequence, establishing a train mileage prediction network model, inputting the train state characteristic parameter sequence during real-time train operation into the train mileage prediction network model, and obtaining the train mileage information at the next moment includes:

[0077] Using length W len The sliding window extracts the mathematical and statistical characteristics of train operation data from the historical driving data along the target track line, and constructs the train state characteristic parameter sequence V t , establish a sample set containing time series information, where each sample contains sequence data of train state characteristic parameters, and the mathematical statistical characteristics of the train operation data include the current train position x t , train speed v t , train acceleration a t , train mileage ΔS t , the cumulative mileage of the train is S t , average train speed Mean train acceleration Train speed maximum sequence v t,max , train acceleration maximum sequence a t,max And the standard deviation v of the velocity maximum sequence and acceleration maximum sequence t,std 、a t,std , extreme value v t,range 、at,range and the ratio v of the maximum to minimum value t,rate 、a t,rate ;

[0078] Then the train state characteristic parameter sequence is:

[0079]

[0080] The train state characteristic parameter sequence is normalized by the min-max normalization method to obtain the continuous W len The characteristic parameter sequence of the positioning solution cycle is used as the input of the train mileage prediction network model, and the predicted train mileage in the next positioning solution cycle is used as the output. The characteristic sequence is input into the train mileage prediction network model f(·). The model extracts and calculates the features according to the input sequence, and calculates the changing trend of the train driving status layer by layer. Through model training, the relationship between the characteristic parameter sequence and the mileage is learned, thereby generating the mileage information ΔS of the train in the next positioning solution cycle ΔT t ' +1 The specific process includes the following steps:

[0081] (1) Let the input sequence V = [V1, V2, V3, ..., V t ]

[0082] (2) The input sequence V undergoes a nonlinear transformation to generate a hidden state sequence H (1)

[0083]

[0084] (3) Using the first hidden state H (1) As input, the new hidden state sequence H is obtained after the second layer transformation (2)

[0085]

[0086] (4) According to Q = XW Q , K=XW K ,P=XW P Generate Query, Key and Value matrices, where W Q 、W K 、W P is a trainable weight matrix; then calculate the dynamic weighted sum weight A,

[0087]

[0088] The softmax function is used to standardize the weight value to the [0,1] interval;

[0089] (5) Perform weighted summation on the Value matrix P according to the dynamic weight and weight A to obtain the weighted aggregation output H agg :

[0090] H agg =AP

[0091] The eigenvector H agg Input into the fully connected layer to generate the predicted mileage ΔS′ of the train t+1

[0092] ΔS′ t+1 =W o H agg +b o

[0093] Where W o and b o are the weights and biases of the fully connected layer respectively;

[0094] (6) Use the mean square error (MSE) as the loss function to calculate the error between the predicted mileage and the actual mileage:

[0095]

[0096] Where N is the number of samples, ΔS t+1 is the actual running mileage of the train;

[0097] Select mean square error As the target loss function, where N is the number of samples, ΔS t+1 Based on the actual mileage of the train, the model is trained iteratively to improve its ability to predict the train's operating status until the model converges.

[0098] For the train mileage prediction network model obtained through training after convergence, we selected historical running data of multiple trains as the test set, adjusted the model parameters according to the model prediction accuracy, and achieved model optimization. The specific structure of the train mileage prediction network model is as follows:

[0099] V=[V1,V2,V3,...,V t ]

[0100]

[0101] Q=XW Q , K=XW K ,P=XW P

[0102]

[0103] H agg =AP

[0104] ΔS′ t+1 =W o H agg +b o

[0105] Among them, f (1) is the activation function of the first layer, θ (1) is the weight and bias set of the first layer network, f (2) is the activation function of the second layer, θ (2) is the weight and bias set of the second layer network, W Q 、W K 、W P is the trainable weight matrix, A is the dynamic weighted sum weight, W o and b o are the weights and biases of the fully connected layer, respectively.

[0106] Preferably, the extracting of information related to the current train and the mileage information of the train at the next moment from the virtual transponder basic database, determining the train mileage and pre-capture trigger time of the next positioning solution cycle, calculating the predicted position of the train closest to the nearest target virtual transponder from the pre-capture trigger time, estimating the capture time difference of the nearest target virtual transponder, and determining the capture time of the nearest target virtual transponder to be captured, includes:

[0107] According to the current train running direction D(t), extract the virtual balise identity sequence in the virtual balise basic database to construct the virtual balise real-time state sequence, and determine the virtual balise number b closest to the target to be captured in this direction i The virtual transponder real-time status sequence describes the interval from the starting station A to the terminal station B and all N AB The state information data of the virtual responder is obtained by the virtual responder capture state Acq(i,D N ), where i represents the virtual transponder number, i = 1, 2, ..., N AB , D N Used to mark the actual running direction of the train during the current balise capture process, Acq(i,D N ) takes a value of 0, indicating that the virtual responder is not captured; Acq(i,D N ) takes a value of 1, indicating that the virtual transponder has been captured;

[0108] According to the number b of the nearest target virtual transponder to be captured i Extract the corresponding virtual transponder spatial data and determine the target virtual transponder mileage information When the predicted running mileage of the train in the next positioning solution cycle calculated at time t meets the following conditions, the current time t is determined to be the pre-capture trigger time of the target virtual balise to be captured

[0109] ΔS′ t+1 ≥ΔS NVB

[0110] Where, ΔS NVB is the distance between the train and the target virtual balise at the current moment;

[0111] In the unit positioning solution cycle ΔT, multi-step forward prediction is performed at fixed time intervals δt, and the train mileage S′(t+kδt), k=1,2,...,N-1, Compute the predicted position of the train at each time step.

[0112] Perform multi-step forward prediction of train mileage S′(t+kδt) at time interval δt. Construct an input feature sequence based on the current train state and historical state. This feature sequence serves as the initial input to the model for the first forward prediction. The model outputs the predicted mileage S′(t+δt) for the first time step, which represents the mileage of the train within the time interval δt. Based on the prediction of the first time step, the prediction result is recursively used as the input for the next time step to continue multi-step forward prediction. In each recursion, the model outputs the mileage for the next time step, resulting in a sequence of predicted mileages for multiple future time steps {S′(t+δt), S′(t+2δt), ..., S′(t+(N-1)δt)}.

[0113] Calculate the train mileage S′(t+kδt) and the target virtual balise mileage at each time step The distance ΔS k (ΔS k =|S′(t+kδt)-S(t)|), find the predicted train position closest to the target virtual balise position among all predicted moments, i.e. ΔS k The minimum time t+kδt, ​​the target virtual transponder capture time difference kδt is obtained;

[0114] The judgment time t+kδt is numbered b i The capture time T of the nearest virtual transponder to be captured cap , that is, T cap =t+kδt.

[0115] Preferably, the sending of position correction information at the capture moment of the nearest target virtual balise to be captured, the train correcting the current position according to the position correction information, and updating the real-time state sequence of the virtual balise at the same time, comprises:

[0116] Extract the three-dimensional coordinate position of the captured virtual transponder closest to the target to be captured and mileage Get the train's current position correction information;

[0117] At the capture moment of the nearest target virtual transponder to be captured, the real-time state sequence of the virtual transponder is updated, and the capture state Acq(b i ,D N ) is captured;

[0118] Along the actual running direction of the train, the next virtual balise closest to the front is selected as the nearest target virtual balise to be captured in the subsequent running process of the system, and the number of the nearest target virtual balise to be captured is updated to b. i+1 .

[0119] It can be seen from the technical solutions provided by the above-mentioned embodiments of the present invention that the method of the present invention combines ultra-wideband technology with Beidou positioning, avoiding the problem of difficult balance between capture accuracy and risk of missed capture in conventional capture solutions from the source, ensuring continuous and stable high-precision positioning regardless of whether Beidou signals are available in complex environments, eliminating the availability limitations of virtual transponder capture, and effectively improving the capture accuracy and stability of virtual transponders in complex environments.

[0120] Additional aspects and advantages of the present invention will be set forth in part in the following description, will become apparent from the following description, or may be learned by practice of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0121] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0122] Figure 1 1 is a flow chart of a method for capturing a virtual transponder based on ultra-wideband enhanced BeiDou according to an embodiment of the present invention;

[0123] Figure 2 2 is a schematic diagram of the principle of extracting observation data sets from multiple ultra-wideband base stations according to an embodiment of the present invention;

[0124] Figure 3Schematic diagram of the regional enhancement mechanism provided by the ultra-wideband unit under Beidou-restricted conditions in an embodiment of the present invention;

[0125] Figure 4 2 is a schematic diagram of the principle of determining the pre-capture triggering time of the nearest target virtual transponder to be captured according to an embodiment of the present invention;

[0126] Figure 5 Schematic diagram of the principle of predictive capture and identification of the nearest virtual transponder to be captured according to an embodiment of the present invention;

[0127] Figure 6 Schematic diagram of the capture process and capture error of a virtual transponder in an embodiment of the present invention;

[0128] Figure 7 3 is a schematic diagram comparing the capture solution of an embodiment of the present invention and a traditional capture solution. DETAILED DESCRIPTION

[0129] The embodiments of the present invention are described in detail below, examples of which are shown in the accompanying drawings, wherein the same or similar reference numerals throughout represent the same or similar elements or elements having the same or similar functions. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain the present invention, and are not to be construed as limiting the present invention.

[0130] It will be understood by those skilled in the art that, unless expressly stated otherwise, the singular forms "a", "an", "said" and "the" used herein may also include the plural forms. It should be further understood that the term "comprising" used in the description of the present invention refers to the presence of the features, integers, steps, operations, elements and / or components, but does not exclude the presence or addition of one or more other features, integers, steps, operations, elements, components and / or groups thereof. It should be understood that when we refer to an element as being "connected" or "coupled" to another element, it may be directly connected or coupled to the other element, or there may be intermediate elements. In addition, "connected" or "coupled" as used herein may include wireless connections or couplings. The term "and / or" used herein includes any unit and all combinations of one or more associated listed items.

[0131] It will be understood by those skilled in the art that, unless otherwise defined, all terms (including technical and scientific terms) used herein have the same meaning as commonly understood by those skilled in the art in the art to which the present invention pertains. It should also be understood that terms such as those defined in common dictionaries should be understood to have meanings consistent with their meanings in the context of the prior art and, unless defined as such herein, will not be interpreted in an idealized or overly formal sense.

[0132] To facilitate understanding of the embodiments of the present invention, several specific embodiments will be further explained below with reference to the accompanying drawings, and each embodiment does not constitute a limitation on the embodiments of the present invention.

[0133] In order to solve the problem of seamless capture of virtual transponders in the complex environment of a wide-area railway network, the present invention provides a virtual transponder capture method based on ultra-wideband enhanced Beidou. This method analyzes the possible quality of various Beidou observations of the target track line, sets a positioning solution logic adjustment and reconstruction mechanism, introduces ultra-wideband positioning technology to enhance Beidou positioning, and uses dynamic iterative optimization decision logic to suppress positioning errors, ensuring the continuous provision of real-time train positioning information. At the same time, a train mileage prediction network model is used to predict the train mileage in the next positioning cycle as the basis for capture judgment, avoiding the problem of difficult balance between capture accuracy and missed capture risk in conventional capture solutions from the source, eliminating the availability limitations of virtual transponder capture, and effectively supporting the design, development, and equipment manufacturing of new train control systems.

[0134] This embodiment provides a processing flow of a virtual transponder acquisition method based on ultra-wideband enhanced Beidou. Figure 1 As shown, the processing steps include the following:

[0135] Step S1, collecting data of the target track line, and establishing a virtual balise basic database and an ultra-wideband device basic database based on the collected data;

[0136] Step S2: Exploring the Beidou positioning observation quality of the target track line to establish a basic database of positioning observation quality, and compiling an ultra-wideband / Beidou joint positioning strategy set based on prior terrain environment information, Beidou positioning observation quality exploration information, and historical train operation data along the line;

[0137] Step S3: During the operation of the train, a positioning strategy is dynamically selected from the ultra-wideband / Beidou joint positioning strategy set according to the real-time observation conditions of the Beidou signal, the positioning solution logic is adjusted and reconstructed, the three-dimensional spatial position of the current train is calculated, the positioning error is suppressed by using dynamic iterative optimization decision logic, the current mileage of the train and the ultra-wideband device spatial data in the ultra-wideband device basic database are calculated, multiple ultra-wideband base station observation data sets are extracted, and the ultra-wideband positioning solution data source is obtained.

[0138] Step S4: constructing a train state characteristic parameter sequence and establishing a train mileage prediction network model. Inputting the train state characteristic parameter sequence during the real-time operation of the train into the train mileage prediction network model to obtain the train mileage information at the next moment.

[0139] Step S5, extracting information related to the current train from the virtual transponder basic database, determining the train mileage and pre-capture trigger time of the next positioning solution cycle, calculating the predicted position of the train closest to the nearest target virtual transponder starting from the pre-capture trigger time, estimating the capture time difference of the nearest target virtual transponder, and determining the capture time of the nearest target virtual transponder to be captured;

[0140] Step S6: sending position correction information at the moment of capturing the nearest target virtual balise to be captured. The train corrects its current position according to the position correction information and updates the real-time status sequence of the virtual balise at the same time.

[0141] The step S1 collects data of the target track line to establish a virtual transponder basic database and an ultra-wideband device basic database, specifically including:

[0142] Combined with the actual situation of the target operation line, a mileage range covering the entire line (from the starting station A to the terminal station B, the mileage range is from the starting station center mileage S A Mileage to terminal station center S B )'s virtual transponder basic database and ultra-wideband device basic database.

[0143] 1) Virtual balise basic database, including track line spatial data, virtual balise identity sequence, and virtual balise spatial data:

[0144] The track line spatial data mainly describes the spatial data of the track line between the starting station A and the terminal station B. It is combined with the actual positioning measurement, data processing and format conversion of several key points on each track line in the interval and station, and is organized into a complete track line spatial database according to a specific data item format. The database mainly stores the spatial information and attribute information of all key points on the line. index,λ The key points of the route, the main stored spatial data items include: route key point mileage S (P index,λ ), three-dimensional coordinate position of key points on the line (L index,λ ,B index,λ ,H index,λ ), the main stored attribute data items include: running direction D (P index,λ ), terrain environment attributes U(P index,λ ). Among them, the subscript index represents the key point number of the line, the subscript λ represents the line number, and the key point mileage S(P index) is directly related to the up and down directions of the line. The mileage in the up direction decreases, while the mileage in the down direction increases. The spatial data accuracy of the key points of the track line directly determines the capture performance of the virtual balise. The accuracy level of the track line spatial database must meet the line segment approximation error requirements, that is, for any two adjacent key points of the line, the numbers are P and α 、P α+1 , the following conditions are met:

[0145] dis{P α (L α ,B α ,H α ),P α+1 (L α+1 ,B α+1 ,H α+1 )}-|S(P α )-S(P α+1 )|≤h S

[0146] Where dis{*} represents the straight-line distance solution function between two points based on three-dimensional space coordinates, h S is the tolerable error limit of orbital space data.

[0147] The virtual transponder identity sequence mainly describes the interval from the starting station A to the terminal station B and all N AB The identity information data of a virtual balise is as follows: for double-track lines and areas with multiple tracks, it is necessary to use line index number, default running direction and other information to mark the virtual balise identity. The virtual balise identity sequence mainly includes the identity data of the virtual balise, where: the virtual balise identity data includes: target line index number λ, direction flag D N , virtual transponder number b i ,Based on this data, the identity of any virtual transponder can be uniquely determined;

[0148] The virtual transponder space data is collected and processed in advance to ensure the uniqueness of the virtual transponder position. The virtual transponder space data items mainly include: virtual transponder number b i , virtual transponder mileage Virtual transponder three-dimensional coordinate position

[0149] 2) Ultra-wideband basic database, including ultra-wideband equipment spatial data, specifically including ultra-wideband base station number A j , ultra-wideband base station mileage Ultra-wideband base station three-dimensional coordinate position

[0150] The UWB device spatial data is generally collected and processed simultaneously during the measurement and processing phase of the track line spatial data and the virtual transponder spatial data to ensure the uniqueness of the UWB device location. Similar to the virtual transponder spatial data, the UWB device spatial data includes the UWB base station number A j , ultra-wideband base station mileage Ultra-wideband base station three-dimensional coordinate position

[0151] The step S2 specifically includes the following steps:

[0152] Step S21, using the historical driving data along the target track line to explore the Beidou positioning observation quality, establish a positioning observation quality basic database, which is used to store Beidou positioning quality observation data, mainly Beidou positioning quality related information of key points of the line, usually carried out simultaneously during the line data survey and measurement process, specifically including: Beidou satellite signal-to-noise ratio mean SNR avg (P index,λ ), the number of visible BeiDou satellites n(P index,λ ), Position Dilution of Precision PDOP(P index,λ ), average pseudorange residual e P,avg (P index,λ ), Beidou satellite signal-to-noise ratio standard deviation σ SNR (P index,λ ), Beidou satellite signal-to-noise ratio reduction average SNR drop (P index,λ );

[0153] Step S22: Based on the Beidou positioning observation quality detection information and the historical train operation data of multiple lines, the target track line is analyzed for various possible Beidou observation quality conditions, including normal Beidou signal quality, degraded Beidou signal quality, and Beidou signal failure. The reasons for Beidou signal degradation and failure can be divided into two factors: terrain shielding and interference. The situations that may occur on actual lines are only a subset of the above three Beidou signal quality conditions. The specific process of Beidou observation quality category judgment is as follows:

[0154] (1) Beidou signal availability judgment: SNR avg ≥SNR min and n≥n min , where SNR avg is the mean signal-to-noise ratio of Beidou satellites, SNR i is the signal-to-noise ratio of the i-th visible Beidou satellite.

[0155] (2) If the signal meets the availability conditions, the signal quality status is judged and divided into normal Beidou signal quality, Beidou signal degradation due to terrain shielding, and Beidou signal degradation due to interference, corresponding to:

[0156] PDOP≤PDOP max and e p,avg ≤e p,max and

[0157] PDOP>PDOP max or e p,avg >e p,max

[0158] or SNR drop >SNR drop,max

[0159] Among them, PDOP is the position precision factor, which reflects the impact of the geometric distribution of Beidou satellites on positioning error. is the coordinate of the i-th visible BeiDou satellite.

[0160] (3) If the signal does not meet the availability conditions, the BeiDou signal is directly judged to be invalid, and the cause of the failure is determined. It can be divided into terrain shielding failure and interference failure, corresponding to the following conditions:

[0161] And SNR drop ≤SNR drop,max

[0162] or SNR drop >SNR drop,max

[0163] in, is the standard deviation of Beidou satellite signal-to-noise ratio; SNR drop is the average decrease in Beidou satellite signal-to-noise ratio, calculated using a sliding window: Where: N t is the number of observations in the sliding window, SNR h (i) is the maximum signal-to-noise ratio in the i-th sliding window, SNR l (i) is the minimum value of the signal-to-noise ratio in the i-th sliding window; SNR min is the minimum threshold of the Beidou satellite signal-to-noise ratio mean, n is the number of visible Beidou satellites, and n min The minimum number of visible Beidou satellites, PDOP max is the maximum acceptable value of the position precision dilution, e p,max is the maximum acceptable value of pseudorange residual, σ SNR，max is the maximum acceptable value of the Beidou satellite signal-to-noise ratio fluctuation, SNRdrop,max It is the maximum acceptable value of the Beidou satellite signal-to-noise ratio degradation.

[0164] According to the common Beidou observation quality of the route, the specific positioning strategy content contained in the ultra-wideband / Beidou joint positioning strategy set is determined accordingly, which mainly includes three types: ultra-wideband independent positioning (UWB), Beidou independent positioning (GNSS), and ultra-wideband / Beidou joint positioning (UWB / GNSS):

[0165]

[0166] In this step, the specific positioning strategy content contained in the ultra-wideband / Beidou joint positioning strategy set is determined, specifically including three strategies: ultra-wideband independent positioning (UWB), Beidou independent positioning (GNSS), and ultra-wideband / Beidou joint positioning (UWB / GNSS):

[0167] (1) Signal quality is normal (Method = GNSS)

[0168] Beidou signals have a high signal-to-noise ratio and stability, which can be seen when there are sufficient Beidou satellites. Selecting the Beidou independent positioning strategy means that only Beidou positioning technology can provide continuous, stable and accurate position information. In this case, ultra-wideband positioning serves as a backup and does not actively participate in positioning calculations.

[0169] (2) Signal quality degradation (Method = UWB / GNSS)

[0170] Beidou signals experience signal-to-noise ratio fluctuations, a decrease in the number of visible Beidou satellites, and an increase in PDOP, resulting in signal quality degradation. However, the signal is still usable. In this case, the UWB / Beidou combined positioning strategy is selected to improve positioning stability and accuracy.

[0171] (3) Beidou signal failure (Method = UWB)

[0172] When BeiDou signals cannot be received normally, the signal-to-noise ratio is extremely low or fails, and the number of BeiDou satellites is far below the minimum requirement, the system switches to the ultra-wideband independent positioning strategy and relies solely on ultra-wideband positioning technology to provide continuous train position information.

[0173] The step S3 specifically includes the following steps:

[0174] Step S31, during the operation of the train, the real-time observation information obtained by the Beidou positioning receiver and the ultra-wideband unit contained in the train control on-board equipment at the current time t is extracted in real time, and the current Beidou signal availability is determined according to the extracted real-time observation data and the Beidou observation quality category judgment process, and the positioning strategy is dynamically selected in the ultra-wideband / Beidou joint positioning strategy set based on this; in this step, the real-time observation information obtained is used to collect Beidou signals in real time by the Beidou positioning receiver contained in the train control on-board equipment. At the same time, in order to ensure the continuity of positioning, the relative distance information between the ultra-wideband unit and the train is synchronously collected in real time. In addition, the ultra-wideband unit is used to detect and diagnose the integrity of the Beidou positioning signal, and to assist in confirming whether the Beidou signal is interfered with, so that when the Beidou signal quality is degraded or the Beidou signal fails, the ultra-wideband unit and the Beidou positioning receiver are selected to realize the combined positioning solution or the ultra-wideband independent positioning solution is used to replace the Beidou independent positioning for the capture judgment of the virtual transponder.

[0175] The BeiDou signal quality judgment condition is used to dynamically select the specific positioning strategy of the UWB / BeiDou joint positioning strategy set to adapt to the changes in BeiDou signal quality, thereby ensuring the accuracy of positioning solution; the system uses the BeiDou satellite signal-to-noise ratio (SNR) avg , the number of visible Beidou satellites n, position precision dilution PDOP, average pseudorange residual e P,avg , Beidou satellite signal-to-noise ratio standard deviation σ SNR , Beidou satellite signal-to-noise ratio reduction average SNR drop To determine the quality of Beidou signals, the specific determination process is as follows:

[0176] (1) Real-time signal acquisition

[0177] At time t, the train control vehicle-mounted equipment receives the signal from the BeiDou positioning receiver and processes it to obtain the current navigation BeiDou ephemeris and the pseudo-ranges of n visible BeiDou satellites {ρ1(t),ρ2(t),…,ρ n (t)} and other real-time observation information, and use the extracted spatial data of m ultra-wideband devices to obtain the relative distance information {d1(t), d2(t), …, d m (t)};

[0178] (2) Real-time judgment of signal availability

[0179] During real-time train operation, the system determines whether the Beidou signal meets the availability conditions in each positioning solution cycle:

[0180] SNR avg ≥SNR min and n≥n min

[0181] If the conditions are met, the signal is determined to be available and the process enters into real-time judgment of the signal quality status. If the conditions are not met, the signal is directly determined to be invalid and the process enters into judgment of the cause of invalidation (step (4)).

[0182] (3) Real-time judgment of signal quality status

[0183] For signals that meet availability conditions, the system further determines the signal quality status and determines the positioning strategy in real time:

[0184] 1: Signal quality is normal

[0185] PDOP≤PDOP max and e p,avg ≤e p,max and

[0186] If the conditions are met, the Beidou independent positioning strategy is selected in real time, and only Beidou signals are used for positioning;

[0187] 2: Signal quality degraded due to terrain obstruction

[0188] PDOP>PDOP max or e p,avg >e p,max

[0189] If the conditions are met, the UWB / Beidou joint positioning strategy is selected in real time, using UWB signals and Beidou signals for joint positioning to improve positioning stability and accuracy;

[0190] 3: Signal quality degraded due to interference

[0191] or SNR drop >SNR drop,max

[0192] If the conditions are met, the UWB / Beidou joint positioning strategy is selected in real time to address the degradation risk caused by signal interference;

[0193] 4: Real-time judgment of failure causes

[0194] If the signal does not meet the availability conditions, the system will directly determine that the Beidou signal is invalid and determine the specific cause of the failure in real time:

[0195] A) Signal failure due to terrain obstruction

[0196] And SNR drop ≤SNR drop,max

[0197] B) The signal fails due to interference

[0198] or SNR drop >SNR drop,max

[0199] If the conditions are met, the UWB independent positioning strategy is selected in real time, relying entirely on UWB positioning to ensure the continuity of positioning information. After determining the Beidou signal quality, the system dynamically selects a positioning strategy from the "UWB / Beidou joint positioning strategy set" according to the following rules to adapt to the current signal quality status:

[0200] (1) Normal signal quality – Beidou independent positioning strategy

[0201] The system relies solely on Beidou signals for positioning. In this case, the Beidou signal quality is stable and no ultra-wideband assistance is required, saving computing and resources.

[0202] (2) Signal quality degradation – UWB / Beidou joint positioning strategy

[0203] In the event of signal degradation, the system selects an ultra-wideband / Beidou joint positioning strategy, introducing ultra-wideband data and Beidou signals for joint positioning and calculation to ensure positioning accuracy and stability. Ultra-wideband plays a reinforcing role in this situation to compensate for the degradation of Beidou signal quality.

[0204] (3) Signal failure – UWB independent positioning strategy

[0205] When the Beidou signal fails, the system will completely switch to the ultra-wideband independent positioning strategy, relying on the ultra-wideband signal to provide positioning information, replacing the failed Beidou positioning, to ensure the continuity of positioning information in the event of Beidou signal failure or severe interference;

[0206] Figure 3 Schematic diagram of the regional enhancement mechanism provided by the ultra-wideband unit under Beidou-restricted conditions described in an embodiment of the present invention.

[0207] Step S32: Start the position calculation process and calculate the current three-dimensional position of the train based on the selected positioning strategy and real-time observation information. Construct the measurement vector z t , which is divided into three situations:

[0208] (1) When the BeiDou signal quality is normal, the measurement vector is

[0209] z t =[ρ1(t),ρ2(t),…,ρ n (t)] T

[0210] (2) When the BeiDou signal is available but the BeiDou signal quality is degraded, the measurement vector is:

[0211] zt =[ρ1(t),ρ2(t),…,ρ n (t),d1(t),d2(t),…,d m (t)] T

[0212] (3) When the BeiDou signal fails, the measurement vector is:

[0213] z t =[d1(t),d2(t),…,d m (t)] T

[0214] The state vector to be estimated is formed by the three-dimensional position of the train {X(t), Y(t), Z(t)} and the clock error Δt: t =[X(t),Y(t),Z(t),Δt] T , in the above three cases, the corresponding measurement vector z is used respectively t , using a nonlinear Kalman filter to calculate the state vector estimate The specific process of filtering estimation includes the following steps:

[0215]

[0216] P t =(IK t H t )P t,t-1

[0217] in, P t,t-1 are the state vector prediction value, the one-step prediction variance matrix, and F t,t-1 is the state transfer matrix, K t is the equivalent filter gain matrix, Q t 、R t are the variance matrices of system error and measurement error, respectively, H t is the linearized equivalent measurement matrix obtained by transforming the observation model, P t is the state estimation variance matrix, thus obtaining the train's three-dimensional coordinate position estimation result

[0218] In this step, a measurement vector is constructed based on the current BeiDou signal quality observations and the selected positioning strategy:

[0219] When the Beidou signal quality is normal, use the Beidou independent positioning strategy to construct the measurement vector:

[0220] z t =[ρ1(t),ρ2(t),…,ρ n (t)]T

[0221] When the BeiDou signal is still available but the quality is degraded, the UWB / BeiDou joint positioning strategy is used to construct the measurement vector:

[0222] z t =[ρ1(t),ρ2(t),…,ρ n (t),d1(t),d2(t),…,d m (t)] T

[0223] When the BeiDou signal fails, the ultra-wideband unit is used to replace the BeiDou signal, and the ultra-wideband independent positioning strategy is used to construct the measurement vector:

[0224] z t =[d1(t),d2(t),…,d m (t)] T

[0225] At the same time, the state vector to be estimated is formed by the three-dimensional position of the train {X(t), Y(t), Z(t)} and the clock error Δt:

[0226] x t =[X(t),Y(t),Z(t),Δt] T

[0227] Combined with the three-dimensional coordinate position of each visible Beidou satellite {X i (t),Y i (t),Z i (t)} Write the pseudo-range observation model based on the simplified three-dimensional coordinate position of the connected ultra-wideband device Write the ranging model and use the nonlinear Kalman filter to calculate the state vector estimate The specific process of filtering estimation includes the following steps:

[0228]

[0229] P t =(IK t H t )P t,t-1

[0230] in, P t,t-1 are the state vector prediction value, the one-step prediction variance matrix, and F t,t-1 is the state transfer matrix, K t is the equivalent filter gain matrix, Q t 、R t are the variance matrices of system error and measurement error, respectively, H tis the linearized equivalent measurement matrix obtained by transforming the observation model, P t is the state estimation variance matrix, thus obtaining the train's three-dimensional coordinate position estimation result

[0231] Step S33: Using the train operation information and observation information at the current time t, enter the position calculation process optimization stage, set the dynamic iterative optimization judgment logic, and iteratively optimize to obtain the optimal solution for train position estimation. Estimate the optimal train position Convert to coordinate position Get the track key point P closest to the current position of the train index,λ , determine its running direction and line number.

[0232] The state vector estimate obtained by Kalman filtering is used as the initial state of the dynamic iterative optimization judgment logic Set the optimization objective function X * To minimize the sum of squared residuals, three optimization objective functions are specifically corresponding to the three different strategies in the selected ultra-wideband / Beidou joint positioning:

[0233]

[0234] Where: e UWB,j =d j -||xp UWB,j || is the ranging residual of the jth ultra-wideband base station, d j is the actual measured ultra-wideband ranging distance, x is the train position to be optimized, and p UWB,j is the location coordinate of the known ultra-wideband base station; is the pseudorange residual of the i-th BeiDou satellite, is the Doppler velocity residual; and is the corresponding covariance matrix; x GNSS and x UWB is the train position status to be optimized calculated based on BeiDou and UWB, e Fus is the Euclidean distance difference between BeiDou and UWB positioning results (e Fus =||x GNSS -x UWB ||), ∑Fus is the covariance of the fusion residual;

[0235] On this basis, the nonlinear least squares method is used to perform state iterative optimization to determine whether the optimization accuracy is met:

[0236] ||e(x (k) )-e(x (k-1) )||<ε r

[0237] Where, ε r is the maximum acceptable variation of the residual, and if the above formula is satisfied, the optimal solution of the optimized train position estimation is output. If the above formula is not satisfied and the maximum number of iterations has been reached, the iteration is stopped and the current optimal estimation result is output;

[0238] In this step, positioning error is suppressed by dynamic iterative optimization logic. At the current time t, the initial state estimate obtained by inputting Kalman filter is and various observation quantities z k , including BeiDou pseudorange ρ, BeiDou clock error Δt, ultra-wideband ranging information d, etc., used to build the relationship between state variables and observations; set up dynamic iterative optimization logic, and use the state estimation value obtained by Kalman filtering as the initial state of the dynamic iterative optimization logic Set the optimization objective function X * To minimize the residual sum of squares, according to the three different strategies in the selected ultra-wideband / Beidou joint positioning, three specific optimization objective functions are respectively corresponding. The specific optimization objective functions are:

[0239]

[0240] Use nonlinear least squares method to perform state iterative optimization to determine whether sufficient optimization accuracy has been achieved:

[0241] ||e(x (k) )-e(x (k-1) )||<ε r

[0242] Where, ε r is the maximum acceptable variation of the residual. If the above formula is satisfied, the optimized state estimate is output. If the above equation is still not satisfied when the maximum number of iterations is reached, the iteration is stopped and the current optimal estimation result is output.

[0243] Depend on Iteratively solve its corresponding geographic coordinates The reference ellipsoid is selected as WGS-84 ellipsoid, and the major axis a is defined as 6378137.0m, the flattening f is defined as 1 / 298.257223563, and the coordinates of the three-dimensional space point are (X * ,Y * ,Z * ), the specific conversion process is as follows:

[0244]

[0245]

[0246] (Meet the accuracy requirements | B i+1 -B i Converges when |<ε, where ε=10 -10 )

[0247] B≈B i+1 ,

[0248] The running direction D(t) and the line number λ are determined, and the track line spatial data in the virtual balise basic database are extracted. Based on the current coordinate position of the train Determine the track key point P closest to the current position of the train index,λ , thereby determining its running direction and line number.

[0249] Step S34: based on the train running status, extract the track line space data in the virtual balise basic database, and estimate the optimal train position according to the calculated train running status. Use the nearest neighbor principle to find the segment between key points of the track closest to the optimal position estimate in the track line spatial data Calculate the current train mileage S(t):

[0250] S(t)=S(P F )±γ||P F -P E ||

[0251] Where S(P F ) is the key point P of the line F Mileage (upward mileage decreases, downward mileage increases), γ is the projection coefficient, ||P F -P E || is a line segment length;

[0252] At the time t when the nearest virtual transponder to be captured is captured VB , use the correction information provided to update the current position of the train, and replace the current train's optimal position estimate coordinates with the three-dimensional coordinate position of the nearest virtual balise to be captured Update mileage at the same time Achieve correction information integration in the position calculation process;

[0253] According to the current train mileage S(t) and the ultra-wideband device spatial data in the ultra-wideband device basic database, multiple ultra-wideband base station observation data sets are extracted according to the distance principle to obtain the ultra-wideband positioning solution data source.

[0254] In this step, at the current time t, the track line spatial data in the virtual balise database is extracted and the optimal train position is estimated based on the calculated In the track line spatial data, the nearest neighbor principle of the key point segment is searched to determine the line key point segment associated with the current position. The segment consists of two adjacent line key points P. F 、P E The straight line segment formed is uniquely determined. On this basis, by calculating the current position Clip between route keypoints The projection position within the line is calculated using the key point P F Mileage S(P F ) and the offset of the projected position in the segment, calculate the current running mileage of the train S(t)=S(P F )±γ||P F -P E ||, where γ is the projection coefficient, ||P F -P E || is a line segment length;

[0255] The multi-UWB base station observation data set extraction screens all devices in the front range of the train that have established signal transmission relationships with the train according to the distance priority principle, discards devices that are far away, and only retains devices that meet the distance requirements to avoid signal quality degradation. Figure 2 Schematic diagram of the principle of extracting observation data sets from multiple ultra-wideband base stations in this embodiment.

[0256] The step S4 specifically includes the following steps:

[0257] Step S41, using a sliding window (length W len ) Extract the mathematical and statistical characteristics of train operation data from the historical driving data along the target track line and construct the train state characteristic parameter sequence V t , establish a sample set containing time series information, where each sample contains sequence data of train state characteristic parameters, and the mathematical statistical characteristics of the train operation data include the current train position x t , train speed v t , train acceleration a t , train mileage ΔS t , the cumulative mileage of the train is S t , average train speed Mean train acceleration Train speed maximum sequence v t,max , train acceleration maximum sequence a t,max And the standard deviation v of the velocity maximum sequence and acceleration maximum sequence t,std 、a t,std , extreme value v t,range 、a t,range and the ratio v of the maximum to minimum valuet,rate 、a t,rate ;

[0258] In this step, set the sliding window W len , based on the historical driving data along the target track line, the mathematical statistical characteristics of train operation data are calculated, and the train state characteristic parameter sequence V is constructed t , establish a sample set containing time series information, the mathematical statistical characteristics of the train operation data include the current train position x t , train speed v t , train acceleration a t , train mileage ΔS t And the train's cumulative mileage S t , take the average speed at each moment in the sliding window and the mean acceleration a t Composition sequence;

[0259] Take the maximum value of the speed and acceleration sequence at each moment in the sliding window to form the train operation data speed maximum sequence v t,max and acceleration maximum sequence a t,max , calculate the standard deviation v of the two sequences respectively t,std 、a t,std , extreme value v t,range 、a t,range and the ratio v of the maximum to minimum value t,rate 、a t,rate ;

[0260] Then the train state characteristic parameter sequence is:

[0261]

[0262] Step S42: normalize the train state characteristic parameter sequence by using the min-max normalization method to obtain the continuous W len The characteristic parameter sequence of the positioning solution cycle is used as the input of the train mileage prediction network model, and the train predicted mileage in the next positioning solution cycle is used as the output. The train mileage prediction network model f(·) is constructed to predict the train mileage information at the next moment. The mean square error is selected. As the target loss function, where N is the number of samples, ΔS t+1 Based on the actual mileage of the train, the model is trained iteratively multiple times to improve its ability to predict the train's operating status until the model converges.

[0263] For the train mileage prediction network model obtained through training after convergence, we selected historical running data of multiple trains as the test set, adjusted the model parameters according to the model prediction accuracy, and achieved model optimization. The specific structure of the train mileage prediction network model is as follows:

[0264] V=[V1,V2,V3,...,V t ]

[0265]

[0266] Q=XW Q , K=XW K ,P=XW P

[0267]

[0268] H agg =AP

[0269] ΔS′ t+1 =W o H agg +b o

[0270] Among them, f (1) is the activation function of the first layer, θ (1) is the weight and bias set of the first layer network, f (2) is the activation function of the second layer, θ (2) is the weight and bias set of the second layer network, W Q 、W K 、W P is the trainable weight matrix, A is the dynamic weighted sum weight, W o and b o are the weights and biases of the fully connected layer respectively;

[0271] In this step, the train mileage prediction network model is constructed to continuously W len The characteristic parameter sequence of a positioning solution cycle is used as the input of the train mileage prediction network, and the train mileage of the next positioning solution cycle is used as the output. The expression of the prediction model is: Where: ΔS′ t+1 is the train mileage within the next positioning solution cycle ΔT output by the model at time t+1, f(·) is the prediction model of the train mileage prediction network, V t is the train state characteristic parameter sequence input to the model at time t;

[0272] The feature sequence is input into the prediction model f(·). The model extracts and calculates features based on the input sequence, and calculates the changing trend of the train's driving status layer by layer. Through model training, the relationship between the feature parameter sequence and the mileage is learned, thereby generating the train mileage ΔS′ for the next positioning solution cycle ΔT. t+1 The specific process includes the following steps:

[0273] (1) Let the input sequence V = [V1, V2, V3, ..., V t ]

[0274] (2) The input sequence V undergoes a nonlinear transformation to generate a hidden state sequence H (1)

[0275]

[0276] (3) Using the first hidden state H (1) As input, the new hidden state sequence H is obtained after the second layer transformation (2)

[0277]

[0278] (4) According to Q = XW Q , K=XW K ,P=XW P Generate Query, Key and Value matrices, where W Q 、W K 、W P is a trainable weight matrix; then calculate the dynamic weighted sum weight A,

[0279]

[0280] The softmax function is used to standardize the weight value to the [0,1] interval;

[0281] (5) Perform weighted summation on the Value matrix P according to the dynamic weight and weight A to obtain the weighted aggregation output H agg :

[0282] H agg =AP

[0283] The eigenvector H agg Input into the fully connected layer to generate the predicted mileage ΔS′ of the train t+1

[0284] ΔS′ t+1 =W o H agg +b o

[0285] Where W o and b o are the weights and biases of the fully connected layer respectively;

[0286] (6) Use the mean square error (MSE) as the loss function to calculate the error between the predicted mileage and the actual mileage:

[0287]

[0288] Where N is the number of samples, ΔS t+1 is the actual running mileage of the train;

[0289] For the train mileage prediction network model trained after convergence, multiple train operation data are selected as the test set, and the model parameters are adjusted according to the model prediction accuracy to achieve model optimization;

[0290] Step S43: The train status data is acquired in real time during the train operation. len Construct train state characteristic parameter sequence V t Whenever new data arrives, add the new data to the feature sequence and remove the earliest data, keeping the window length W len , with the updated feature sequence As the latest input of the train mileage prediction network model, the train predicted mileage ΔS′ in the next positioning solution cycle t+1 As output, a forward prediction of the train operation status is performed.

[0291] In this step, the train status data includes the current train position x t , train speed v t , train acceleration a t , train mileage ΔS t And the train's cumulative mileage S t , the train state characteristic parameter sequence V t Including the current train position x t , train speed v t , train acceleration a t And the train's cumulative mileage S t , take the average speed at each moment in the sliding window and the mean acceleration Composition sequence;

[0292] Take the maximum value of the speed and acceleration sequence at each moment in the sliding window to form the train operation data speed maximum sequence v t,max and acceleration maximum sequence a t,max , calculate the standard deviation v of the two sequences respectively t,std 、a t,std , extreme value v t,range 、a t,range And the ratio of maximum to minimum value v t,rate 、a t,rate ;

[0293] Whenever new data arrives, add the new data to the feature sequence and remove the earliest data, keeping the window length W len, with the updated feature sequence As the latest input of the train mileage prediction network model, the real-time output of the mileage prediction value ΔS′ of the next positioning solution cycle is t+1 , to achieve forward prediction of the train’s operating status.

[0294] The step S5 specifically includes the following steps:

[0295] Step S51: According to the current train running direction D(t), extract the virtual transponder identity sequence in the virtual transponder basic database to construct the virtual transponder real-time state sequence, and determine the virtual transponder number b that is closest to the target to be captured in this direction. i ;

[0296] In this step, the virtual transponder real-time status sequence mainly describes the interval from the starting station A to the terminal station B and all N AB The state information data of the virtual responder is obtained by the virtual responder capture state Acq(i,D N ), where i represents the virtual transponder number, i = 1, 2, ..., N AB , D N Used to mark the actual running direction of the train during the current balise capture process, Acq(i,D N ) takes a value of 0, indicating that the virtual responder is not captured; Acq(i,D N ) takes a value of 1, indicating that the virtual balise has been captured. During the initialization phase before the train runs on the target line, the capture status of all virtual balises must be reset to zero.

[0297] Step S52: according to the number b of the nearest target virtual transponder to be captured, i Extract the corresponding virtual transponder spatial data and determine the target virtual transponder mileage information When the predicted running mileage of the train in the next positioning solution cycle calculated at time t meets the following conditions, the current time t is determined to be the pre-capture trigger time of the target virtual balise to be captured

[0298] ΔS′ t+1 ≥ΔS NVB

[0299] Where, ΔS NVB is the distance between the train and the target virtual balise at the current moment;

[0300] Figure 4This is a schematic diagram of the principle of predicting the pre-capture trigger time of the target virtual transponder to be captured in this embodiment. This figure intuitively shows the comparison process of the predicted mileage and the target virtual transponder position at different time steps to determine the optimal pre-capture trigger time.

[0301] Step S53: within the unit positioning solution cycle ΔT, perform multi-step forward prediction at fixed time intervals δt, and the train mileage S′(t+kδt), k=1,2,...,N-1, Compute the predicted position of the train at each time step.

[0302] In this step, a multi-step forward prediction of train mileage S′(t+kδt) is performed at time interval δt. An input feature sequence is constructed based on the current train state and historical state. This feature sequence serves as the initial input to the model for the first forward prediction. The model outputs the predicted mileage S′(t+δt) for the first time step, representing the mileage traveled by the train within the time interval δt. Based on the prediction of the first time step, the prediction result is recursively used as the input for the next time step to continue the multi-step forward prediction. The specific process is as follows:

[0303] 1) Add the predicted mileage S′(t+δt) of the first time step to the feature sequence and use Estimate the velocity eigenvector for the next time step, Estimate the acceleration eigenvector of the next time step, and based on it calculate the mathematical statistical eigenvalue of the train state feature sequence of the next time step, and update the input feature sequence to include the latest prediction information;

[0304] 2) Forward prediction again: input the updated input feature sequence into the model to obtain the predicted mileage S′(t+2δt) for the next time step;

[0305] 3) Repeat the above process: continue to recursively update the input feature sequence and input it into the model for prediction until the prediction of all steps is completed;

[0306] In each recursion, the model outputs the mileage of the next time step, and finally obtains the predicted mileage sequence {S′(t+δt),S′(t+2δt),...,S′(t+(N-1)δt)} for multiple time steps in the future

[0307] Step S54, calculate the train running mileage S'(t+kδt) and the target virtual balise mileage at each time step The distance ΔS k (ΔS k=|S′(t+kδt)-S(t)|), find the predicted train position closest to the target virtual balise position among all predicted moments, i.e. ΔS k The minimum time t+kδt, ​​the target virtual transponder capture time difference kδt is obtained;

[0308] In this step, for each prediction time kδt, the distance between the train's predicted mileage and the target virtual balise is calculated:

[0309] ΔS k =|S′(t+kδt)-S(t)|

[0310] At all N predicted moments, traverse the calculated distances, find the minimum value, and determine the moment corresponding to the minimum distance:

[0311] k min =argmin(S′(t+δt),S′(t+2δt),...,S′(t+(N-1)δt))

[0312] Step S55: Determine the time t+kδt as the time number b i The capture time of the nearest target virtual transponder to be captured, that is, T cap =t+kδt.

[0313] Figure 5 Schematic diagram of the principle of predicting the capture decision of the nearest target virtual transponder to be captured in this embodiment. Figure 4 In

[15] , the train's mileage is accumulated through multi-step forward prediction, and the distance between the train and the nearest virtual balise to be captured is calculated at each time step to determine the capture time.

[0314] Figure 6 Schematic diagram of the capture process and capture error of the virtual transponder in this embodiment. Figure 5 As shown in the figure, 50 virtual balises are arranged at equal intervals of 1.5 km. The train runs at a speed of 80 km / h along the straight track. The eastward position of the virtual balise extends from 500 meters to 2000 meters, and the northward position gradually decreases from 2000 meters to 1700 meters. The capture system predicts the train position near each virtual balise. The black line represents the train track, the triangle represents the virtual balise position, and the diamond is the capture position. Due to measurement errors, there is a certain deviation between the capture position and the virtual balise position. The bar chart shows the capture error of the target virtual balise to be captured.

[0315] The step S6 specifically includes the following steps:

[0316] Step S61: extract the three-dimensional coordinate position of the captured virtual transponder closest to the target to be captured mileage Provides the current position correction information to the position calculation process;

[0317] Step S62: At the capture time of the nearest target virtual transponder to be captured, update the real-time state sequence of the virtual transponder and modify the capture state Acq(b i ,D N ) is captured;

[0318] Step S63: along the actual running direction of the train, select the next virtual transponder in front of the train as the nearest target virtual transponder to be captured in the subsequent running process of the system, and update the number of the nearest target virtual transponder to be captured to b. i+1 .

[0319] Figure 7 Schematic diagram of a capture solution according to an embodiment of the present invention, wherein the gray portion shows the differences between the capture solution designed according to the embodiment of the present invention and the traditional capture solution.

[0320] In summary, the method of the present invention combines ultra-wideband technology with Beidou positioning, avoiding the problem of difficult balance between capture accuracy and missed capture risk in conventional capture schemes from the source, ensuring continuous and stable high-precision positioning regardless of whether Beidou signals are available in complex environments, eliminating the availability limitations of virtual transponder capture, and effectively improving the capture accuracy and stability of virtual transponders in complex environments.

[0321] The present invention fully considers the variability of Beidou observation conditions during train operation, and enhances Beidou by introducing ultra-wideband technology. It not only replaces Beidou positioning in environments where Beidou signals are unstable or blocked, such as tunnels, but also reinforces Beidou signals in interfered areas, providing continuous positioning results. By constructing a mileage prediction model, it changes the limitation of relying on a fixed capture radius in conventional capture schemes, so that capture judgment no longer relies solely on real-time positioning solution results, reduces the impact of Beidou signal fluctuations, ensures the train's accurate capture of the target virtual transponder, and provides higher protection for continuous positioning and safety monitoring of on-board equipment.

[0322] The ultra-wideband technology in the embodiments of the present invention can effectively ensure that the implementation of virtual transponder technology has a continuous and reliable source of basic information. By constructing a train mileage prediction network model, it avoids the problem of difficult balance between capture accuracy and missed capture risk in conventional capture schemes, thereby achieving seamless capture of virtual transponders.

[0323] The virtual transponder capture method based on the ultra-wideband enhanced global navigation Beidou system provided in this embodiment improves the limitations of the traditional method that solely relies on Beidou signal positioning in the virtual transponder capture judgment process. The method uses the train mileage prediction as the judgment basis in the capture process, constructs a train running mileage prediction model, and significantly improves the timeliness and accuracy of the capture judgment through the effective use of the train historical status data, overcoming the problem of the difficulty in achieving both positioning accuracy and the risk of missed capture in the traditional method that relies on a fixed capture space radius. While appropriately controlling the amount of calculation, the method is suitable for various Beidou signal observation conditions and has strong adaptability.

[0324] The method provided in this embodiment makes full use of the information support of the virtual transponder basic database, and introduces the regional enhancement mechanism provided by ultra-wideband positioning under special conditions such as specific Beidou service limitations and Beidou positioning performance degradation. It will be able to effectively ensure that the implementation of virtual transponder technology has a continuous and reliable source of basic information, and thus provide support for the seamless optimization of virtual transponder functions. It has broad application prospects in railway applications such as train dynamic status monitoring and train tracking and early warning.

[0325] Those skilled in the art will appreciate that the accompanying drawings are merely schematic diagrams of an embodiment, and the modules or processes in the accompanying drawings are not necessarily required to implement the present invention.

[0326] From the above description of the embodiments, it can be seen that those skilled in the art can clearly understand that the present invention can be implemented by means of software plus the necessary general-purpose hardware platform. Based on this understanding, the technical solution of the present invention, or the portion that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a storage medium such as ROM / RAM, a magnetic disk, or an optical disk, and includes a number of instructions for enabling a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in various embodiments of the present invention or certain parts of the embodiments.

[0327] Each embodiment in this specification is described in a progressive manner. The same or similar parts between the embodiments can be referred to each other. Each embodiment focuses on the differences from other embodiments. In particular, for the device or system embodiments, since they are basically similar to the method embodiments, the description is relatively simple. For the relevant parts, refer to the partial description of the method embodiments. 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 can be selected according to actual needs to achieve the purpose of the scheme of this embodiment. A person of ordinary skill in the art can understand and implement it without making any creative efforts.

[0328] The above description is merely a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any changes or substitutions that can be easily conceived by a person skilled in the art within the technical scope disclosed in the present invention should be included in the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be based on the scope of protection of the claims.

Claims

1. A train positioning method based on ultra-wideband enhanced Beidou and virtual transponder, characterized in that: include: Establish a virtual balise basic database and an ultra-wideband device basic database based on the collected target track line data; Establish a basic positioning observation quality database based on the Beidou positioning observation quality of the target track line, and compile an ultra-wideband / Beidou joint positioning strategy set based on prior terrain environment information, Beidou positioning observation quality exploration information, and historical train operation data along the line; During the train operation, a positioning strategy is dynamically selected from the UWB / Beidou joint positioning strategy set according to the real-time observation conditions of the Beidou signal, the three-dimensional spatial position of the train is calculated, and multiple UWB base station observation data sets are extracted based on the current mileage of the train and the UWB device spatial data in the UWB device basic database to obtain the UWB positioning solution data source of the train; Constructing a train state characteristic parameter sequence, establishing a train mileage prediction network model, inputting the train state characteristic parameter sequence during the real-time operation of the train into the train mileage prediction network model, and obtaining the train mileage information at the next moment; Extracting information related to the current train and the mileage information of the train at the next moment from the virtual balise basic database, determining the train mileage and pre-capture trigger time of the next positioning solution cycle, calculating the predicted position of the train closest to the nearest target virtual balise from the pre-capture trigger time, estimating the capture time difference of the nearest target virtual balise, and determining the capture time of the nearest target virtual balise to be captured; The position correction information is sent at the capture moment of the nearest target virtual balise to be captured, and the train corrects the current position according to the position correction information and updates the real-time status sequence of the virtual balise at the same time.

2. The train positioning method based on ultra-wideband enhanced Beidou and virtual balise according to claim 1 is characterized in that: The method of establishing a virtual balise basic database and an ultra-wideband device basic database based on the collected target track line data includes: Set the target operation route from the starting station A to the terminal station B, and its mileage range is from the starting station center mileage S A Mileage to terminal station center S B , build a basic database of virtual transponders and ultra-wideband equipment covering the entire mileage range of the line; The virtual transponder basic database includes track line spatial data, virtual transponder identity sequence and virtual transponder spatial data. The track line spatial data describes the spatial data of the track line between the starting station A and the terminal station B. Combined with the actual positioning measurement, data processing and format conversion of several key points on each track line in the interval and station, a complete track line spatial database is formed according to a specific data item format. The database stores the spatial information and attribute information of all key points on the line. index,λ The spatial data items stored in the key points of the route include: key point mileage S (P index,λ ), three-dimensional coordinate position of key points on the line (L index,λ ,B index,λ ,H index,λ ), the stored attribute data items include: running direction D (P index,λ ), terrain environment attributes U(P index,λ ), the subscript index indicates the key point number of the line, the subscript λ indicates the line number, and the key point mileage S(P index ) is directly related to the up and down directions of the line. The mileage in the up direction decreases, and the mileage in the down direction increases. For any two adjacent key points of the line, the numbers are P α 、P α+1 , the following conditions are met: dis{P α (L α ,B α ,H α ),P α+1 (L α+1 ,B α+1 ,H α+1 )}-|S(P α )-S(P α+1 )|≤h s Where dis{*} represents the straight-line distance solution function between two points based on three-dimensional space coordinates, h S is the tolerable orbital space data error limit; The virtual transponder identity sequence describes the interval from the origin station A to the terminal station B and all N AB The identity data of the virtual transponder includes: target line index number λ, direction flag bit D N and virtual transponder number b i ; The virtual transponder space data items include: virtual transponder number b i , virtual transponder mileage and the three-dimensional coordinate position of the virtual transponder The UWB basic database contains UWB equipment spatial data, specifically including UWB base station number A j , ultra-wideband base station mileage and the three-dimensional coordinate position of the ultra-wideband base station 3. The train positioning method based on ultra-wideband enhanced Beidou and virtual transponder according to claim 1 is characterized in that: The method of establishing a basic positioning observation quality database based on the Beidou positioning observation quality of the target track line and compiling an ultra-wideband / Beidou joint positioning strategy set based on prior terrain environment information, Beidou positioning observation quality exploration information, and historical train operation data along the line includes: Use the historical driving data along the target track line to explore the quality of Beidou positioning observations, and establish a basic positioning observation quality database to store Beidou positioning quality observation data, including: Beidou satellite signal-to-noise ratio (SNR) mean avg (P index,λ ), the number of visible BeiDou satellites n(P index,λ ), position precision dilution PDOP (P index,λ ), average pseudorange residual e P,avg (P index,λ ), Beidou satellite signal-to-noise ratio standard deviation σ SNR (P index,λ ) and the average signal-to-noise ratio decrease of Beidou satellites SNR drop (P index,λ ); Based on BeiDou positioning observation quality detection information and historical train operation data of multiple lines, the target track line is analyzed for various possible BeiDou observation quality conditions, including normal BeiDou signal quality, degraded BeiDou signal quality, and BeiDou signal failure. The positioning strategy content Method contained in the ultra-wideband / Beidou joint positioning strategy set is determined according to the common Beidou observation quality of the line. The positioning strategy content Method includes three types: ultra-wideband independent positioning (UWB), Beidou independent positioning (GNSS), and ultra-wideband / Beidou joint positioning (UWB / GNSS).

4. The train positioning method based on ultra-wideband enhanced Beidou and virtual balise according to claim 3 is characterized in that: The method of dynamically selecting a positioning strategy in the ultra-wideband / Beidou joint positioning strategy set according to the real-time observation conditions of the Beidou signal during the train operation to calculate the three-dimensional spatial position of the train currently running includes: During the real-time operation of the train, at time t, the train control onboard equipment receives the signal from the BeiDou positioning receiver and processes it to obtain the current navigation BeiDou ephemeris and the pseudo-ranges of n visible BeiDou satellites {ρ1(t),ρ2(t),…,ρ n (t)} real-time observation data, using the extracted spatial data of m ultra-wideband devices to obtain the relative distance information between the device and the current train {d1(t), d2(t), …, d m (t)}; During each positioning solution cycle, the system determines whether the Beidou signal meets the availability conditions: SNR avg ≥SNR min and n≥n min Determine the signal quality status of the Beidou signal that meets the availability conditions, and dynamically select a positioning strategy in the UWB / Beidou joint positioning strategy set based on the signal quality status of the Beidou signal A) Signal quality is normal PDOP≤PDOP max and e p,avg ≤e p,max and If the conditions are met, the Beidou independent positioning strategy is selected; B) Signal quality degradation due to occlusion PDOP > PDOP max or e p,avg > e p,max If the conditions are met, the UWB / Beidou joint positioning strategy is selected; C) Signal quality degradation due to interference or SNR drop > SNR drop,max If the conditions are met, the UWB / Beidou joint positioning strategy is selected in real time; If the system determines that the BeiDou signal does not meet the availability conditions, it selects the ultra-wideband independent positioning strategy; Construct the measurement vector z according to the selected positioning strategy and real-time observation information t The train's three-dimensional position {X(t), Y(t), Z(t)} and the clock error Δt constitute the state vector to be estimated, and the state vector estimate is calculated using a nonlinear state filter. Get the three-dimensional position of the train 5. The train positioning method based on ultra-wideband enhanced Beidou and virtual transponder according to claim 4 is characterized in that: The method of extracting multiple ultra-wideband base station observation data sets based on the current running mileage of the train and the ultra-wideband device spatial data in the ultra-wideband device basic database to obtain the ultra-wideband positioning solution data source of the train includes: When the Beidou signal quality is normal, the Beidou independent positioning strategy is used to construct the measurement vector: z t =[ρ1(t),ρ2(t),…,ρ n (t)] T When the BeiDou signal is still available but the quality is degraded, the UWB / BeiDou joint positioning strategy is used to construct the measurement vector: z t =[ρ1(t),ρ2(t),…,ρ n (t),d1(t),d2(t),…,d m (t)] T When the BeiDou signal fails, the ultra-wideband independent positioning strategy is used to construct the measurement vector: z t =[d1(t),d2(t),…,d m (t)] T The state vector to be estimated is composed of the three-dimensional position of the train {X(t), Y(t), Z(t)} and the clock error Δt: x t =[X(t),Y(t),Z(t),Δt] T Combined with the three-dimensional coordinate position of each visible Beidou satellite {X i (t),Y i (t),Z i (t)} Write the pseudo-range observation model based on the simplified three-dimensional coordinate position of the connected ultra-wideband device Write the ranging model and use the nonlinear Kalman filter to calculate the state vector estimate State vector estimate The specific process includes the following steps: P t =(I-K t H t )P t,t-1 in, P t,t-1 are the state vector prediction value, the one-step prediction variance matrix, and F t,t-1 is the state transfer matrix, K t is the equivalent filter gain matrix, Q t 、R t are the variance matrices of system error and measurement error, respectively, H t is the linearized equivalent measurement matrix obtained by transforming the observation model, P t is the state estimation variance matrix; Set up dynamic iterative optimization judgment logic, and use the current train operation information and observation information at time t to iteratively optimize and obtain the optimal solution for train position estimation. Estimate the optimal train position Convert to coordinate position Get the track key point P closest to the current position of the train index,λ , determine the train's running direction and line number; The state vector estimate obtained by Kalman filtering is used as the initial state of the dynamic iterative optimization judgment logic Set the optimization objective function X * To minimize the sum of squared residuals, three optimization objective functions are specifically corresponding to the three different strategies in the selected ultra-wideband / Beidou joint positioning: Where: e UWB,j =d j -||xp UWB,j || is the ranging residual of the jth ultra-wideband base station, d j is the actual measured ultra-wideband ranging distance, x is the train position to be optimized, and p UWB,j is the location coordinate of the known ultra-wideband base station; is the pseudorange residual of the i-th BeiDou satellite, is the Doppler velocity residual; and is the corresponding covariance matrix; x GNSS and x UWB is the train position status to be optimized calculated based on BeiDou and UWB, e Fus is the Euclidean distance difference between BeiDou and UWB positioning results, e Fus =||x GNSS -x UWB ||, ∑Fus is the covariance of the fusion residual; Use nonlinear least squares method to perform state iterative optimization to determine whether the optimization accuracy is met: ||e(x (k) )-former (k-1) )||<ε r Where, ε r is the maximum acceptable variation of the residual, and if the above formula is satisfied, the optimal solution of the optimized train position estimation is output. If the above formula is not satisfied and the maximum number of iterations has been reached, the iteration is stopped and the current optimal estimation result is output; Depend on Iteratively solve the corresponding geographic coordinates of the train Based on the current coordinate position of the train Determine the track key point P closest to the current position of the train index,λ , thereby determining the train's running direction D(t) and the line number λ; Based on the train running status, the track line spatial data in the virtual balise basic database is extracted, and the optimal train position is estimated based on the calculated Use the nearest neighbor principle to find the segment between key points of the track closest to the optimal position estimate in the track line spatial data Calculate the current train mileage S(t): S(t)=S(P F )±γ||P F -P E || Where S(P F ) is the key point P of the line F Mileage, γ is the projection coefficient, ||P F -P E || is a line segment length; At the time t when the nearest virtual transponder to be captured is captured VB , use the correction information provided by the nearest virtual balise to be captured to update the current position of the train, and replace the current train's optimal position estimate with the three-dimensional coordinate position of the nearest virtual balise to be captured Update mileage at the same time Achieve correction information integration in the position calculation process; According to the current train mileage S(t) and the ultra-wideband device spatial data in the ultra-wideband device basic database, multiple ultra-wideband base station observation data sets are extracted according to the distance principle to obtain the ultra-wideband positioning solution data source of the train.

6. The train positioning method based on ultra-wideband enhanced Beidou and virtual transponder according to claim 5 is characterized in that: The method of constructing a train state characteristic parameter sequence, establishing a train mileage prediction network model, inputting the train state characteristic parameter sequence during real-time train operation into the train mileage prediction network model, and obtaining the train mileage information at the next moment includes: Using length W len The sliding window extracts the mathematical and statistical characteristics of train operation data from the historical driving data along the target track line, and constructs the train state characteristic parameter sequence V t , establish a sample set containing time series information, where each sample contains sequence data of train state characteristic parameters, and the mathematical statistical characteristics of the train operation data include the current train position x t , train speed v t , train acceleration a t , train mileage ΔS t , the cumulative mileage of the train is S t , average train speed Mean train acceleration Train speed maximum sequence v t,max , train acceleration maximum sequence a t,max And the standard deviation v of the velocity maximum sequence and acceleration maximum sequence t,std 、a t,std , extreme value v t,range 、a t,range and the ratio v of the maximum to minimum value t,rate 、a t,rate ; Then the train state characteristic parameter sequence is: The train state characteristic parameter sequence is normalized by the min-max normalization method to obtain the continuous W len The characteristic parameter sequence of the positioning solution cycle is used as the input of the train mileage prediction network model, and the predicted train mileage in the next positioning solution cycle is used as the output. The characteristic sequence is input into the train mileage prediction network model f(·). The model extracts and calculates the features according to the input sequence, and calculates the changing trend of the train driving status layer by layer. Through model training, the relationship between the characteristic parameter sequence and the mileage is learned, thereby generating the mileage information ΔS′ of the train in the next positioning solution cycle ΔT. t+1 The specific process includes the following steps: (1) Let the input sequence V = [V1, V2, V3, ..., V t ] (2) The input sequence V undergoes a nonlinear transformation to generate a hidden state sequence H (1) (3) Using the first hidden state H (1) As input, the new hidden state sequence H is obtained after the second layer transformation (2) (4) According to Q = XW Q , K=XW K ,P=XW P Generate Query, Key and Value matrices, where W Q 、W K 、W P is a trainable weight matrix; then calculate the dynamic weighted sum weight A, The softmax function is used to standardize the weight value to the [0,1] interval; (5) Perform weighted summation on the Value matrix P according to the dynamic weight and weight A to obtain the weighted aggregation output H agg : H=AP agg The eigenvector H agg Input into the fully connected layer to generate the train's predicted mileage ΔS t ' +1 ΔS t ′ +1 =W o H agg +b o Where W o and b o are the weights and biases of the fully connected layer respectively; (6) Use the mean square error (MSE) as the loss function to calculate the error between the predicted mileage and the actual mileage: Where N is the number of samples, ΔS t+1 is the actual running mileage of the train; Select mean square error As the target loss function, where N is the number of samples, ΔS t+1 Based on the actual mileage of the train, the model is trained iteratively to improve its ability to predict the train's operating status until the model converges. For the train mileage prediction network model obtained through training after convergence, we selected historical running data of multiple trains as the test set, adjusted the model parameters according to the model prediction accuracy, and achieved model optimization. The specific structure of the train mileage prediction network model is as follows: V=[V1,V2,V3,...,V t ] Q=XW Q ,K=XW K ,P=XW P H=AP agg ΔS t ′ +1 =W o H agg +b o Among them, f (1) is the activation function of the first layer, θ (1) is the weight and bias set of the first layer network, f (2) is the activation function of the second layer, θ (2) is the weight and bias set of the second layer network, W Q 、W K 、W P is the trainable weight matrix, A is the dynamic weighted sum weight, W o and b o are the weights and biases of the fully connected layer, respectively.

7. The train positioning method based on ultra-wideband enhanced Beidou and virtual transponder according to claim 6 is characterized in that: The method of extracting information related to the current train and the mileage information of the train at the next moment from the virtual transponder basic database, determining the train mileage and pre-capture trigger time of the next positioning solution cycle, calculating the predicted position of the train closest to the nearest target virtual transponder from the pre-capture trigger time, estimating the capture time difference of the nearest target virtual transponder, and determining the capture time of the nearest target virtual transponder to be captured, includes: According to the current train running direction D(t), extract the virtual balise identity sequence in the virtual balise basic database to construct the virtual balise real-time state sequence, and determine the virtual balise number b closest to the target to be captured in this direction i The virtual transponder real-time status sequence describes the interval from the starting station A to the terminal station B and all N AB The state information data of the virtual responder is obtained by the virtual responder capture state Acq(i,D N ), where i represents the virtual transponder number, i = 1, 2, ..., N AB , D N Used to mark the actual running direction of the train during the current balise capture process, Acq(i,D N ) takes a value of 0, indicating that the virtual responder is not captured; Acq(i,D N ) takes a value of 1, indicating that the virtual transponder has been captured; According to the number b of the nearest target virtual transponder to be captured i Extract the corresponding virtual transponder spatial data and determine the target virtual transponder mileage information When the predicted running mileage of the train in the next positioning solution cycle calculated at time t meets the following conditions, the current time t is determined to be the pre-capture trigger time of the target virtual balise to be captured ΔS t ' +1 ≥ΔS NVB Where, ΔS NVB is the distance between the train and the target virtual balise at the current moment; In the unit positioning solution cycle ΔT, multi-step forward prediction is performed at fixed time intervals δt, and the train mileage S′(t+kδt), k=1,2,...,N-1, Compute the predicted position of the train at each time step. Perform multi-step forward prediction of train mileage S′(t+kδt) at time interval δt. Construct an input feature sequence based on the current train state and historical state. This feature sequence serves as the initial input to the model for the first forward prediction. The model outputs the predicted mileage S′(t+δt) for the first time step, which represents the mileage of the train within the time interval δt. Based on the prediction of the first time step, the prediction result is recursively used as the input for the next time step to continue multi-step forward prediction. In each recursion, the model outputs the mileage for the next time step, resulting in a sequence of predicted mileages for multiple future time steps {S′(t+δt), S′(t+2δt), ..., S′(t+(N-1)δt)}. Calculate the train mileage S′(t+kδt) and the target virtual balise mileage at each time step The distance ΔS k (ΔS k =|S′(t+kδt)-S(t)|), find the predicted train position closest to the target virtual balise position among all predicted moments, i.e. ΔS k The minimum time t+kδt, ​​the target virtual transponder capture time difference kδt is obtained; The judgment time t+kδt is numbered b i The capture time T of the nearest virtual transponder to be captured cap , that is, T cap =t+kδt.

8. The train positioning method based on ultra-wideband enhanced Beidou and virtual transponder according to claim 7 is characterized in that: The sending of position correction information at the capture moment of the nearest target virtual balise to be captured, the train correcting the current position according to the position correction information, and updating the real-time state sequence of the virtual balise at the same time, comprises: Extract the three-dimensional coordinate position of the captured virtual transponder closest to the target to be captured and mileage Get the train's current position correction information; At the capture moment of the nearest target virtual transponder to be captured, the real-time state sequence of the virtual transponder is updated, and the capture state Acq(b i ,D N ) is captured; Along the actual running direction of the train, the next virtual balise closest to the front is selected as the nearest target virtual balise to be captured in the subsequent running process of the system, and the number of the nearest target virtual balise to be captured is updated to b. i+1 .

Images