Virtual reconnection high-speed train dynamic interval prediction method and system based on collaborative trajectory prediction

Abstract

The invention provides a virtual reconnection high-speed train dynamic interval prediction method and system based on collaborative trajectory prediction, and belongs to the technical field of train operation control. An operation control framework of a virtual reconnection train group is established, and a hierarchical collaborative prediction mechanism of a leading train and a following train is defined; based on a train longitudinal dynamic model, a position uncertainty safety boundary is constructed by considering factors such as positioning errors and communication delay; the LSTM neural network and physical constraints are fused, physical correction is carried out on the trajectory through acceleration amplitude limiting and kinematics recursion, then a prediction result driven by data and a prediction result driven by a physical model are fused, and high-precision prediction of the running trajectory of the front vehicle is achieved; and dynamically calculating the safety interval and the target interval between the trains based on the prediction result. According to the method, the problems of insufficient dynamism, prediction distortion and poor multi-train adaptability in the prior art are solved, the tracking interval is effectively reduced on the premise of guaranteeing safety, and the trafficability and the operation efficiency of a line are improved.

Description

Technical Field

[0001] This invention relates to the field of train operation control technology, specifically to a method and system for predicting the dynamic intervals of virtual double-unit high-speed trains based on cooperative trajectory prediction. Background Technology

[0002] Traditional CTCS-3 level train control systems use fixed block signaling, with train intervals relying on preset block sections, resulting in large interval redundancy and low track resource utilization. Virtual multiple-unit (MMU) technology, on the other hand, enables multi-train collaborative control through wireless communication (such as GSM-R / LTE-R), reducing train intervals in a "virtual formation" mode, and represents a core technological direction for improving transportation efficiency.

[0003] Currently, there has been considerable research both domestically and internationally on train interval control in virtual multiple-unit (MMU) technology. Besides traditional low-cost control methods such as PID control, many experts and scholars have applied modern intelligent control algorithms to train interval control in virtual MMU operation modes, such as model predictive control, sliding mode control, and control methods based on deep learning and reinforcement learning. However, existing virtual MMU interval calculation methods have many shortcomings, such as: insufficient dynamism: many methods treat train traction / braking performance as static constants, ignoring its nonlinear changes with speed, leading to redundancy or safety risks in interval calculations; lack of trajectory prediction: interval calculations rely solely on the current state of the train (position, speed), without considering the future trajectory of preceding trains, making it difficult to adapt to dynamic interval adjustments under complex operating conditions (such as acceleration and braking); lack of physical constraints: some trajectory prediction methods (such as single LSTM) do not incorporate train dynamics, and the prediction results may violate kinematic laws (such as exceeding acceleration limits), leading to distorted interval calculations; poor adaptability to multiple trains: trajectory prediction logic has not been designed for hierarchical control scenarios of multiple trains in virtual MMU, making it difficult to support coordinated interval control of three or more trains. Summary of the Invention

[0004] The purpose of this invention is to provide a method and system for predicting the dynamic interval of virtual double-unit high-speed trains based on cooperative trajectory prediction, so as to solve at least one of the technical problems existing in the background art.

[0005] To achieve the above objectives, the present invention adopts the following technical solution:

[0006] In a first aspect, the present invention provides a method for predicting the dynamic interval of virtual double-unit high-speed trains based on cooperative trajectory prediction, comprising:

[0007] To construct the input dataset for collaborative trajectory prediction, historical trajectory data of each train in a virtual coupled train set is collected. This historical trajectory data includes train position and speed. (The text then abruptly shifts to a different topic:) The historical trajectory of the train and the first The historical trajectory of the train, for the first Predicting the trajectory of trains;

[0008] An LSTM prior prediction network is constructed. The input data is standardized, and the standardized spliced ​​sequence is fed into a multi-layer LSTM network. The LSTM network captures long-term dependencies in the sequence through its internal gating mechanism and cell state, and the network outputs the future sequence. The initial predicted trajectory at each time step;

[0009] Establish a state correction iterative model to correct parts of the LSTM prediction results that may violate physical laws. Based on the velocity sequence predicted by LSTM, the initial acceleration is obtained by differential calculation. The maximum traction acceleration and maximum braking deceleration of the train at the current predicted speed are queried. The acceleration is limited. Using the limited acceleration, the train state transition matrix is ​​used for forward recursion to obtain the predicted trajectory based entirely on the physical model.

[0010] Based on the train dynamics model, considering positioning error, speed measurement error, speed measurement update frequency, maximum communication delay, and train length, a forward safety boundary and a backward safety boundary for train position uncertainty are established, and a position uncertainty safety boundary model is constructed. The dynamic safety interval and the dynamic target interval are calculated. Based on the train position uncertainty, the front end and back end of the train safety envelope are calculated. Combined with the emergency braking distance, the dynamic safety interval is calculated. Based on the corrected predicted trajectory, the target interval at each predicted time is calculated, and finally, the dynamic target interval is obtained.

[0011] As a further limitation of the first aspect of the present invention, the input dataset for constructing the cooperative trajectory prediction specifically includes: based on the virtual coupled train set, the historical trajectory data. Including train location and train speed The historical trajectory of the train is The predicted trajectory is , define the first The historical trajectory of trains , No. The historical trajectory of trains and the The historical trajectory of trains Through the first Column and number The trajectory of the train is related to the first The trajectory of each train is predicted; for the first train, a supplementary reference trajectory is used. The lead train tracks a preset reference trajectory; the second train predicts and tracks based on the lead train's reference trajectory and historical trajectories; the third and subsequent trains need to simultaneously receive historical trajectory information from the preceding and even earlier trains for collaborative prediction and tracking.

[0012] As a further limitation of the first aspect of the present invention, the construction of an LSTM prior prediction network specifically includes: standardizing all trajectory data; designing an LSTM with a dropout mechanism between layers to prevent overfitting; concatenating the standardized data into an input sequence; capturing temporal dependencies through the forget gate, input gate, and output gate of the LSTM; outputting prior prediction results; and then inversely standardizing the prior prediction results to finally obtain the prior prediction trajectory of the train.

[0013] As a further limitation of the first aspect of the present invention, a state correction iterative model is established, specifically as follows: the position and speed of the train are extracted from the obtained prior predicted trajectory, the predicted prior acceleration is calculated, and then the acceleration is limited to a range that conforms to the actual situation according to the traction / braking characteristics of the train, so that the acceleration does not exceed the maximum braking acceleration and maximum traction acceleration of the train. When it exceeds the range, it is replaced by a boundary value; then, based on the longitudinal dynamic equation formula of the train, the physical correction trajectory is calculated; the data-driven prediction results and the model-driven prediction results are fused, the prior error covariance matrix is ​​calculated, the gain is calculated, and finally the optimized predicted trajectory that conforms to the physical laws after correction is output, and the posterior error covariance matrix is ​​updated.

[0014] As a further limitation of the first aspect of the present invention, a position uncertainty safety boundary model is established based on the train dynamics model, specifically including: calculating the front and rear envelopes of the train operation; and considering positioning errors. Speed ​​measurement error Speed ​​test update frequency Maximum communication delay and train length A location uncertainty safety boundary model is established, and the dynamic safety interval is calculated. Based on the corrected prediction trajectory, the target interval at each prediction time is calculated, and finally the dynamic target interval is obtained to ensure safety throughout the entire prediction period.

[0015] As a further limitation of the first aspect of this invention, the data-driven prediction results are fused with the model-driven prediction results to finally output an optimized prediction trajectory. To obtain the trajectory across the entire prediction time domain, iterative prediction is performed, updating the optimized prediction result at the current moment to the historical trajectory queue as input for the next prediction step. Simultaneously, the error covariance matrix... It also iterates and updates at each step, providing a basis for the next fusion filtering.

[0016] Secondly, the present invention provides a virtual double-unit high-speed train dynamic interval prediction system based on cooperative trajectory prediction, comprising:

[0017] The acquisition module is used to construct the input dataset for cooperative trajectory prediction, collecting historical trajectory data of each train in the virtual coupled train set. This historical trajectory data includes train position and speed. The historical trajectory of the train and the first The historical trajectory of the train, for the first Predicting the trajectory of trains;

[0018] The first prediction module is used to construct the LSTM prior prediction network. It standardizes the input data and feeds the standardized spliced ​​sequence into a multi-layer LSTM network. The LSTM network captures long-term dependencies in the sequence through its internal gating mechanism and cell state, and the network outputs the future sequence. The initial predicted trajectory at each time step;

[0019] The iterative module is used to establish a state correction iterative model and correct parts of the LSTM prediction results that may violate physical laws. Based on the velocity sequence predicted by LSTM, the initial acceleration is obtained through differential calculation. The maximum traction acceleration and maximum braking deceleration of the train at the current predicted speed are queried. The acceleration is limited. Using the limited acceleration, the train state transition matrix is ​​used for forward recursion to obtain the predicted trajectory based entirely on the physical model.

[0020] The second prediction module is used to establish forward and backward safety boundaries for train position uncertainty based on the train dynamics model, taking into account positioning error, speed measurement error, speed measurement update frequency, maximum communication delay, and train length. It also establishes a position uncertainty safety boundary model. The module calculates dynamic safety intervals and dynamic target intervals. Based on train position uncertainty, it calculates the front and back ends of the train safety envelope. Combined with emergency braking distance, it calculates dynamic safety intervals. Based on the corrected prediction trajectory, it calculates the target intervals at each prediction time, and finally obtains the dynamic target intervals.

[0021] Thirdly, the present invention provides a non-transitory computer-readable storage medium for storing computer instructions, which, when executed by a processor, implement the method for predicting the dynamic interval of virtual double-unit high-speed trains based on cooperative trajectory prediction as described in the first aspect.

[0022] Fourthly, the present invention provides a computer device including a memory and a processor, wherein the processor and the memory communicate with each other, the memory stores program instructions that can be executed by the processor, and the processor calls the program instructions to execute the method for predicting the dynamic interval of virtual double-unit high-speed trains based on cooperative trajectory prediction as described in the first aspect.

[0023] Fifthly, the present invention provides an electronic device, comprising: a processor, a memory, and a computer program; wherein the processor is connected to the memory, the computer program is stored in the memory, and when the electronic device is running, the processor executes the computer program stored in the memory to cause the electronic device to execute instructions for implementing the method for predicting the dynamic interval of virtual multiple-unit high-speed trains based on cooperative trajectory prediction as described in the first aspect.

[0024] The beneficial effects of this invention are: by integrating dynamic traction / braking characteristics, cooperative trajectory prediction and physical constraints, compared with other high-speed train prediction methods, it considers the running trajectories of multiple trains, and achieves precise intervals while ensuring safety, thereby improving the efficiency of line transportation.

[0025] The advantages of additional aspects of the invention will be set forth more clearly in the following description or will be learned by practice of the invention. Attached Figure Description

[0026] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the following description of the embodiments will be briefly introduced. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0027] Figure 1 This is a schematic diagram illustrating the basic principle of the virtual double-unit high-speed train dynamic interval calculation method based on cooperative trajectory prediction as described in an embodiment of the present invention.

[0028] Figure 2 This is a diagram of the location uncertainty safety boundary model described in an embodiment of the present invention.

[0029] Figure 3 This is a diagram of the LSTM neural network structure described in an embodiment of the present invention.

[0030] Figure 4 This is a framework diagram for iterative prediction of train state trajectory according to an embodiment of the present invention. Detailed Implementation

[0031] 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 denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain the present invention, and should not be construed as limiting the present invention.

[0032] It will be understood by those skilled in the art that, unless otherwise defined, all terms used herein (including technical and scientific terms) have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains.

[0033] It should also be understood that terms such as those defined in general dictionaries should be understood to have meanings consistent with their meanings in the context of the prior art, and should not be interpreted in an idealized or overly formal sense unless defined as here.

[0034] Those skilled in the art will understand that, unless specifically stated otherwise, the singular forms “a,” “an,” “the,” and “the” used herein may also include the plural forms. It should be further understood that the term “comprising” as used in this specification means the presence of the stated 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, and / or groups thereof.

[0035] In the description of this specification, references to terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of the present invention. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples. Without contradiction, those skilled in the art can combine and integrate the different embodiments or examples described in this specification, as well as the features of those different embodiments or examples.

[0036] To facilitate understanding of the present invention, the present invention will be further explained and described below with reference to the accompanying drawings and specific embodiments. However, the specific embodiments do not constitute a limitation on the embodiments of the present invention.

[0037] Those skilled in the art should understand that the accompanying drawings are merely schematic diagrams of embodiments, and the components in the drawings are not necessarily essential for implementing the present invention.

[0038] Example 1

[0039] In this embodiment 1, a dynamic interval prediction system for virtual coupled high-speed trains based on cooperative trajectory prediction is first provided, including: an acquisition module, used to construct an input dataset for cooperative trajectory prediction, and to collect historical operating trajectory data of each train in the virtual coupled train group, wherein the historical trajectory data includes train position and speed, and through the... The historical trajectory of the train and the first The historical trajectory of the train, for the first The system predicts the trajectory of a train. The first prediction module constructs an LSTM prior prediction network, standardizing the input data. The standardized concatenated sequence is then fed into a multi-layer LSTM network. The LSTM network, through its internal gating mechanism and cell states, captures long-term dependencies within the sequence, and outputs the future... The initial predicted trajectory at each time step; the iterative module is used to establish a state correction iterative model, correcting parts of the LSTM prediction results that may violate physical laws. Based on the velocity sequence predicted by LSTM, the initial acceleration is obtained through differential calculation. The maximum traction acceleration and maximum braking deceleration of the train at the current predicted speed are queried, the acceleration is limited, and the limited acceleration is used to perform forward recursion using the train state transition matrix to obtain a predicted trajectory based entirely on the physical model; the second prediction module is used to establish forward and backward safety boundaries for train position uncertainty based on the train dynamics model, considering positioning error, speed measurement error, speed measurement update frequency, maximum communication delay, and train length, and to establish a position uncertainty safety boundary model; the dynamic safety interval and dynamic target interval are calculated. Based on the train position uncertainty, the front and back ends of the train safety envelope are calculated. Combined with the emergency braking distance, the dynamic safety interval is calculated. Based on the corrected predicted trajectory, the target interval at each prediction time is calculated, and finally the dynamic target interval is obtained.

[0040] In this embodiment, based on the above system, a method for calculating the dynamic interval of virtual double-unit high-speed trains based on cooperative trajectory prediction is implemented, including the following steps:

[0041] To construct the input dataset for cooperative trajectory prediction, historical trajectory data of each train in a Virtual Multiple Unit Train System (VCTS) is collected. This historical trajectory data includes train position and speed. The historical trajectory of the train (the historical trajectory of the preceding train) and the first The historical trajectory of the train (the historical trajectory of the preceding train), for the first Predicting the trajectory of trains.

[0042] An LSTM prior prediction network is constructed by standardizing the input data to improve the stability and training efficiency of the LSTM network. The standardized concatenated sequence is then input into a multi-layer LSTM network. The LSTM network, through its internal gating mechanisms (forget gate, input gate, output gate) and cell states, captures long-term dependencies in the sequence, and the network outputs the future sequence. The initial predicted trajectory (including position and velocity) at each time step.

[0043] A state correction iterative model is established to correct potentially physical violations in the LSTM prediction results. Based on the velocity sequence predicted by the LSTM, the initial acceleration is calculated using differential calculations. The maximum traction acceleration and maximum braking deceleration of the train at the current predicted speed are retrieved, and the acceleration is limited. Using the limited acceleration, the train's kinematic equations (state transition matrix) are used for forward recursion to obtain a predicted trajectory entirely based on the physical model. The data-driven prediction results are fused with the model-driven prediction results to output an optimized predicted trajectory that is both accurate and conforms to physical laws. To obtain the trajectory across the entire prediction time domain, iterative prediction is performed, updating the optimized prediction result at the current moment to the historical trajectory queue as input for the next prediction step. Simultaneously, the error covariance matrix... It also iterates and updates at each step, providing a basis for the next fusion filtering.

[0044] Based on the train dynamics model, considering positioning error, speed measurement error, speed measurement update frequency, maximum communication delay, and train length, forward and backward safety boundaries for train position uncertainty are established, and a position uncertainty safety boundary model is constructed. Dynamic safety intervals and dynamic target intervals are calculated. Based on train position uncertainty, the front and back ends of the train safety envelope are calculated. Combined with emergency braking distance, the dynamic safety interval is calculated. Based on the corrected predicted trajectory, the target interval at each predicted time is calculated, ultimately yielding the dynamic target interval.

[0045] Specifically, constructing the input dataset for cooperative trajectory prediction includes defining the Virtual Multiple-Unit Train Set (VCTS) as follows:

[0046] (1)

[0047] in For the lead car, To track the train, its preceding train is The train that follows is The historical trajectory data Including train location and train speed The train's historical trajectory is as follows: The predicted trajectory is The specific formula is as follows:

[0048] (2)

[0049] in, For the length of the historical trajectory, To predict the trajectory length. Define the first... The historical trajectory of trains , No. The historical trajectory of trains and the The historical trajectory of trains Through the first Column and number The trajectory of the train is related to the first The trajectory of each train is predicted. For the first train ( =1), supplementary reference trajectory The lead train tracks a preset reference trajectory; the second train predicts and tracks based on the lead train's reference trajectory and historical trajectories; the third and subsequent trains need to simultaneously receive historical trajectory information from the preceding and even earlier trains for collaborative prediction and tracking.

[0050] (3)

[0051] Constructing an LSTM prior prediction network specifically includes:

[0052] First, all trajectory data are standardized according to equation (4). The mean of the trajectory data. denoted as the standard deviation of the trajectory data.

[0053] (4)

[0054] The design incorporates an LSTM, with dropout mechanisms added between layers to prevent overfitting. The standardized data is concatenated into an input sequence according to equation (5). Temporal dependencies are captured through the LSTM's forget gate, input gate, and output gate, and the prior prediction results are output. Then, the prior prediction results are inversely standardized according to equation (6) to finally obtain the prior prediction trajectory of the train. .

[0055] (5)

[0056] (6)

[0057] The gating unit calculation of the LSTM network satisfies the formula, where It is the sigmoid activation function. , , This is the weight matrix. This is the hidden state from the previous moment. , , For bias, This is the current input.

[0058] (7)

[0059] The LSTM hidden state and cell state updates follow the formula below, where In cellular state, Candidate cell state, This is the weight matrix. For bias.

[0060] (8)

[0061] Specifically, the state correction iterative model is established as follows:

[0062] First, the prior predicted trajectory obtained from claim 3. Extract the train's location With speed Then, the predicted prior acceleration is calculated according to formula (9), and the acceleration is limited to a range that conforms to reality based on the traction / braking characteristics of the train. To ensure that the acceleration does not exceed the train's maximum braking acceleration. and maximum traction acceleration If the value is outside the range, the boundary value is used instead.

[0063] (9)

[0064] Then, based on the train longitudinal dynamics equation (10), the physical correction trajectory is calculated.

[0065] (10)

[0066] Next, the data-driven prediction results are fused with the model-driven prediction results. First, the prior error covariance matrix is ​​calculated according to formula (11), and the gain is calculated according to formula (12), where It is the identity matrix. and The noise covariance matrices of the physical model and the prior prediction are respectively. Finally, the optimized prediction trajectory that conforms to the physical law after correction is output according to formula (13). The posterior error covariance matrix is ​​updated according to formula (14).

[0067] (11)

[0068] (12)

[0069] (13)

[0070] (14)

[0071] Finally, there is iterative prediction. A queue structure is used, where old data is popped and new data is pushed in each iteration. To ensure real-time input, repeat the above steps until the output of the forward-moving train is obtained. Complete predicted sequence , used for the dynamic target interval calculation described in claim 1.

[0072] A safety boundary model for position uncertainty is established based on the train dynamics model, specifically including:

[0073] The front and rear envelopes of the train's movement are calculated using the following formulas:

[0074] (15)

[0075] (16)

[0076] Considering positioning error Speed ​​measurement error Speed ​​test update frequency Maximum communication delay and train length Establish a location uncertainty safety boundary model, wherein the dynamic safety interval The calculation method is as follows:

[0077] (17)

[0078] Based on the corrected predicted trajectory in claim 4 The target interval for each prediction time is calculated according to formula (18), and the dynamic target interval is finally obtained according to formula (19) to ensure safety throughout the entire prediction period. For the train ahead exist Predicting speed at any given moment For the train at speed Common braking distances.

[0079] (18)

[0080] (19)

[0081] Example 2

[0082] Virtual multiple-unit (VMU) is a cutting-edge technology in the rail transit field. Through wireless communication and cooperative control, it allows multiple trains to operate in close formation, as if connected by a "virtual coupler," while maintaining a safe distance. This shortens train intervals and increases line capacity. However, as an emerging technology, VMU is still somewhat immature. Existing VMU train interval calculation methods suffer from problems such as insufficient dynamism, distorted trajectory prediction, and poor adaptability among multiple trains. Furthermore, traditional prediction methods that rely solely on data-driven models may produce prediction results that violate physical laws, affecting the reliability and safety of the control system. This embodiment 2 provides a dynamic interval calculation method based on cooperative trajectory prediction.

[0083] like Figures 1 to 4 As shown, this method mainly includes: First, defining the Virtual Multiple-Unit Train Set (VCTS) and train operation state variables, establishing a multi-train control strategy, and on this basis, constructing a cooperative trajectory predictor of "LSTM prior prediction + physical correction iteration," and fusing the data-driven prediction results with the model-driven prediction results using Kalman filtering. A longitudinal dynamic model considering the basic and additional resistances of train operation is constructed, the safety envelope model of train operation is analyzed, and a safety model for train position uncertainty is established considering speed measurement and positioning errors and communication delays. Then, combined with the obtained high-precision forward train prediction trajectory, the safety interval and target interval are dynamically calculated. The specific implementation steps of this method are as follows:

[0084] Step 1: Construct the input dataset for collaborative trajectory prediction by collecting historical trajectory data of each train in the Virtual Multiple Unit Train System (VCTS).

[0085] In the scenario of virtual multiple-unit operation on high-speed railways, multiple trains (no fewer than 3) run in formation along the track. The trains are numbered according to their direction of travel, with the first train defined as the lead car and the following trains as the trailing trains. The train numbering is as follows:

[0086]

[0087] Among them, Representing the lead car, To track the train, its preceding train is The train that follows is .

[0088] Subsequently, real-time operational data of each train in the Virtual Multiple Unit Train System (VCTS) was collected, including the position sampled once per sampling period. and speed Its representative is the train. exist Position and velocity at any given moment:

[0089] Define the historical trajectory of train operation as Before recording The state of the step defines the predicted trajectory as After prediction The state of the step.

[0090]

[0091] Different trajectory tracking and predictive control rules are formulated for the lead train, the second train, and subsequent trains. These control rules include:

[0092] when =1, i.e., the lead car According to the preset trajectory Tracking is performed, but there is no predictor.

[0093]

[0094] when When =2, that is based on and the historical trajectory of the lead car To make predictions and thus obtain .

[0095]

[0096] when When ≥3, based on Historical trajectory and Historical trajectory To make predictions and thus obtain .

[0097]

[0098] Step 2: Construct an LSTM network for prior prediction.

[0099] Firstly, because the absolute values ​​of the position coordinates are very large, matrix operations in neural networks can easily lead to numerical overflow or loss of precision. Activation functions such as sigmoid and tanh will enter the saturation region when the input value is very large, and the gradient will approach 0, causing the gradient to vanish. In order to make the prediction algorithm more efficient and accurate and speed up the convergence speed of the algorithm, the collected data is standardized according to the following formula.

[0100] ;

[0101] The standardized spliced ​​sequences are input into a multilayer LSTM network. The LSTM network captures long-term dependencies in the sequences through its internal gating mechanisms (forget gate, input gate, output gate) and cell states.

[0102] The gating unit calculation of the LSTM network satisfies the following formula:

[0103]

[0104] The LSTM hidden state and cell state update follow the following formula:

[0105]

[0106] The network outputs the initial predicted trajectory for the next time step according to the following formula:

[0107]

[0108] Then, denormalize the output according to the following formula to restore it to the original data dimensions.

[0109]

[0110] Step 3: Establish a state correction iterative model.

[0111] To correct for potentially physics-insensitive parts of the LSTM prediction results and make the predictions more closely reflect reality, we first use the velocity sequence extracted from the prior predicted trajectory to calculate its acceleration using the following formula.

[0112]

[0113] Query the train's maximum traction acceleration and maximum braking deceleration at the current predicted speed, and limit the acceleration according to the following formula to ensure it does not exceed the train's physical capabilities. .

[0114]

[0115] Then correct the speed according to the following formula:

[0116] ;

[0117] Then, physical corrections are performed based on the following vehicle longitudinal dynamics equations.

[0118] .

[0119] LSTM predictions may be inaccurate due to model errors or data noise, and physical models may also have errors due to model simplification or inaccurate parameters. If the uncertainty of LSTM predictions is small (i.e., more reliable), the fusion result will place more trust in the LSTM predictions; conversely, it will place more trust in the physical model. Therefore, to fuse data-driven and model-driven predictions, the prior error covariance matrix is ​​first calculated: ; Calculate the gain: ,in It is the identity matrix. and These are the noise covariance matrices of the physical model and the prior prediction, respectively. The final output is the optimized predicted trajectory that conforms to physical laws after correction. The posterior error covariance matrix is ​​then updated. .

[0120] Finally, iterative prediction is performed to obtain the complete prediction sequence. A queue structure is used, where old data is popped and new data is pushed in each iteration. To ensure real-time input, repeat step 2 and the steps above until the output of the forward-moving train is obtained. Complete predicted sequence .

[0121] Step 4: Establish a position uncertainty safety boundary model based on the train dynamics model and calculate the dynamic target interval.

[0122] In virtual reconnection scenarios, establishing a safety envelope aims to transform all unavoidable uncertainties and risks in reality into a concrete and calculable "safety space," thereby absolutely guaranteeing that trains will not collide in the event of unknowns and accidents. Therefore, positioning errors are considered. Speed ​​measurement error Speed ​​test update frequency Maximum communication delay and train length The front and back envelopes of train operation are established according to the following formula.

[0123]

[0124] Based on the above analysis, a location uncertainty safety boundary model is established, wherein the dynamic safety interval... The calculation method is as follows:

[0125]

[0126] Based on the corrected predicted trajectory in step 3 The target interval for each prediction time is calculated using the following formula:

[0127]

[0128] The dynamic target interval is finally obtained according to the following formula, ensuring safety throughout the entire prediction period. Wherein, For the train ahead exist Predicting speed at any given moment For the train at speed Common braking distances.

[0129]

[0130] In summary, the present invention provides a method for calculating the dynamic interval of virtual double-unit high-speed trains based on cooperative trajectory prediction. Its core lies in achieving dynamic and accurate calculation of the interval through a progressive process of "cooperative prediction - physical correction - safety calculation". The specific implementation proceeds in the following steps: First, construct the input dataset for cooperative trajectory prediction. The system collects historical operating trajectory data (including position and speed) of each train in the virtual double-unit train group. For the train in the formation... The train followed the train ahead ( ) and the preceding train ( Historical trajectory data is stitched together and standardized to improve the training efficiency and numerical stability of subsequent models. Then, an LSTM prior prediction network is constructed. The standardized sequence is input into a multi-layer LSTM network, utilizing its internal gating mechanisms (forget gate, input gate, output gate) and the long-term dependencies of the cell state to output an initial predicted trajectory for a certain time period in the future. To correct the problem that pure data-driven prediction may violate physical laws, a state correction iterative model is established. This model first calculates the initial acceleration based on the velocity sequence predicted by the LSTM, and then limits the acceleration according to the train's dynamic traction / braking performance curve to ensure its physical feasibility. Next, using the limited acceleration, forward recursion is performed through the train's kinematic equations to obtain the predicted trajectory based on the physical model. Finally, the LSTM prior prediction result and the physical model recursion result are optimally fused. Trust weights are dynamically assigned based on the uncertainty of both predictions, outputting an optimized predicted trajectory that is both accurate and conforms to physical laws. This process is iterative prediction; the optimization result of the current step is fed back to the historical queue for the next prediction, thus achieving rolling optimization. At the control level, the basic and additional resistances experienced by the train are first analyzed to establish a longitudinal train dynamics model per unit mass. Simultaneously, to quantify uncertainties in the real world, factors such as positioning error, speed measurement error, communication delay, and train length are comprehensively considered to establish forward and backward safety boundaries for train position uncertainty, forming a safety envelope for train operation. Based on this safety envelope, the minimum dynamic interval ensuring absolute safety is calculated. Furthermore, using the optimized predicted trajectory obtained in the second step, the target interval for each future time moment is calculated proactively within the prediction time domain, and its minimum value is taken as the optimal dynamic target interval for the current moment. This maximizes the reduction of train intervals and improves line capacity while ensuring safety.

[0131] Example 3

[0132] This embodiment 3 provides a non-transitory computer-readable storage medium for storing computer instructions. When executed by a processor, the computer instructions implement the virtual double-unit high-speed train dynamic interval prediction method based on cooperative trajectory prediction as described above. The method includes:

[0133] To construct the input dataset for collaborative trajectory prediction, historical trajectory data of each train in a virtual coupled train set is collected. This historical trajectory data includes train position and speed. (The text then abruptly shifts to a different topic:) The historical trajectory of the train and the first The historical trajectory of the train, for the first Predicting the trajectory of trains;

[0134] An LSTM prior prediction network is constructed. The input data is standardized, and the standardized spliced ​​sequence is fed into a multi-layer LSTM network. The LSTM network captures long-term dependencies in the sequence through its internal gating mechanism and cell state, and the network outputs the future sequence. The initial predicted trajectory at each time step;

[0135] Establish a state correction iterative model to correct parts of the LSTM prediction results that may violate physical laws. Based on the velocity sequence predicted by LSTM, the initial acceleration is obtained by differential calculation. The maximum traction acceleration and maximum braking deceleration of the train at the current predicted speed are queried. The acceleration is limited. Using the limited acceleration, the train state transition matrix is ​​used for forward recursion to obtain the predicted trajectory based entirely on the physical model.

[0136] Based on the train dynamics model, considering positioning error, speed measurement error, speed measurement update frequency, maximum communication delay, and train length, a forward safety boundary and a backward safety boundary for train position uncertainty are established, and a position uncertainty safety boundary model is constructed. The dynamic safety interval and the dynamic target interval are calculated. Based on the train position uncertainty, the front end and back end of the train safety envelope are calculated. Combined with the emergency braking distance, the dynamic safety interval is calculated. Based on the corrected predicted trajectory, the target interval at each predicted time is calculated, and finally, the dynamic target interval is obtained.

[0137] Example 4

[0138] This embodiment 4 provides a computer device, including a memory and a processor, wherein the processor and the memory communicate with each other, and the memory stores program instructions that can be executed by the processor. The processor calls the program instructions to execute the virtual double-unit high-speed train dynamic interval prediction method based on cooperative trajectory prediction as described above. The method includes:

[0139] To construct the input dataset for collaborative trajectory prediction, historical trajectory data of each train in a virtual coupled train set is collected. This historical trajectory data includes train position and speed. (The text then abruptly shifts to a different topic:) The historical trajectory of the train and the first The historical trajectory of the train, for the first Predicting the trajectory of trains;

[0140] An LSTM prior prediction network is constructed. The input data is standardized, and the standardized spliced ​​sequence is fed into a multi-layer LSTM network. The LSTM network captures long-term dependencies in the sequence through its internal gating mechanism and cell state, and the network outputs the future sequence. The initial predicted trajectory at each time step;

[0141] Establish a state correction iterative model to correct parts of the LSTM prediction results that may violate physical laws. Based on the velocity sequence predicted by LSTM, the initial acceleration is obtained by differential calculation. The maximum traction acceleration and maximum braking deceleration of the train at the current predicted speed are queried. The acceleration is limited. Using the limited acceleration, the train state transition matrix is ​​used for forward recursion to obtain the predicted trajectory based entirely on the physical model.

[0142] Based on the train dynamics model, considering positioning error, speed measurement error, speed measurement update frequency, maximum communication delay, and train length, a forward safety boundary and a backward safety boundary for train position uncertainty are established, and a position uncertainty safety boundary model is constructed. The dynamic safety interval and the dynamic target interval are calculated. Based on the train position uncertainty, the front end and back end of the train safety envelope are calculated. Combined with the emergency braking distance, the dynamic safety interval is calculated. Based on the corrected predicted trajectory, the target interval at each predicted time is calculated, and finally, the dynamic target interval is obtained.

[0143] Example 5

[0144] This embodiment 5 provides an electronic device, including: a processor, a memory, and a computer program; wherein, the processor is connected to the memory, and the computer program is stored in the memory. When the electronic device is running, the processor executes the computer program stored in the memory, so that the electronic device executes instructions to implement the virtual multiple-unit high-speed train dynamic interval prediction method based on cooperative trajectory prediction as described above. The method includes:

[0145] To construct the input dataset for collaborative trajectory prediction, historical trajectory data of each train in a virtual coupled train set is collected. This historical trajectory data includes train position and speed. (The text then abruptly shifts to a different topic:) The historical trajectory of the train and the first The historical trajectory of the train, for the first Predicting the trajectory of trains;

[0146] An LSTM prior prediction network is constructed. The input data is standardized, and the standardized spliced ​​sequence is fed into a multi-layer LSTM network. The LSTM network captures long-term dependencies in the sequence through its internal gating mechanism and cell state, and the network outputs the future sequence. The initial predicted trajectory at each time step;

[0147] Establish a state correction iterative model to correct parts of the LSTM prediction results that may violate physical laws. Based on the velocity sequence predicted by LSTM, the initial acceleration is obtained by differential calculation. The maximum traction acceleration and maximum braking deceleration of the train at the current predicted speed are queried. The acceleration is limited. Using the limited acceleration, the train state transition matrix is ​​used for forward recursion to obtain the predicted trajectory based entirely on the physical model.

[0148] Based on the train dynamics model, considering positioning error, speed measurement error, speed measurement update frequency, maximum communication delay, and train length, a forward safety boundary and a backward safety boundary for train position uncertainty are established, and a position uncertainty safety boundary model is constructed. The dynamic safety interval and the dynamic target interval are calculated. Based on the train position uncertainty, the front end and back end of the train safety envelope are calculated. Combined with the emergency braking distance, the dynamic safety interval is calculated. Based on the corrected predicted trajectory, the target interval at each predicted time is calculated, and finally, the dynamic target interval is obtained.

[0149] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0150] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0151] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0152] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment, whereby a series of operational steps are performed to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0153] While the specific embodiments of the present invention have been described above in conjunction with the accompanying drawings, this is not intended to limit the scope of protection of the present invention. Those skilled in the art should understand that, based on the technical solutions disclosed in the present invention, various modifications or variations that can be made by those skilled in the art without creative effort should be included within the scope of protection of the present invention.

Claims

1. A method for predicting the dynamic interval of virtual double-unit high-speed trains based on cooperative trajectory prediction, characterized in that, include: To construct the input dataset for collaborative trajectory prediction, historical trajectory data of each train in a virtual coupled train set is collected. This historical trajectory data includes train position and speed. (The text then abruptly shifts to a different topic:) The historical trajectory of the train and the first The historical trajectory of the train, for the first Predicting the trajectory of trains; An LSTM prior prediction network is constructed. The input data is standardized, and the standardized spliced ​​sequence is fed into a multi-layer LSTM network. The LSTM network captures long-term dependencies in the sequence through its internal gating mechanism and cell state, and the network outputs the future sequence. The initial predicted trajectory at each time step; Establish a state correction iterative model to correct parts of the LSTM prediction results that may violate physical laws. Based on the velocity sequence predicted by LSTM, the initial acceleration is obtained by differential calculation. The maximum traction acceleration and maximum braking deceleration of the train at the current predicted speed are queried. The acceleration is limited. Using the limited acceleration, the train state transition matrix is ​​used for forward recursion to obtain the predicted trajectory based entirely on the physical model. Based on the train dynamics model, considering positioning error, speed measurement error, speed measurement update frequency, maximum communication delay, and train length, a forward safety boundary and a backward safety boundary for train position uncertainty are established, and a position uncertainty safety boundary model is constructed. The dynamic safety interval and the dynamic target interval are calculated. Based on the train position uncertainty, the front end and back end of the train safety envelope are calculated. Combined with the emergency braking distance, the dynamic safety interval is calculated. Based on the corrected predicted trajectory, the target interval at each predicted time is calculated, and finally, the dynamic target interval is obtained.

2. The method for predicting the dynamic interval of virtual double-unit high-speed trains based on cooperative trajectory prediction according to claim 1, characterized in that, The input dataset for constructing collaborative trajectory prediction specifically includes: historical trajectory data based on the virtual coupled train sets. Including train location and train speed The historical trajectory of the train is The predicted trajectory is , define the first The historical trajectory of trains , No. The historical trajectory of trains and the The historical trajectory of trains Through the first Column and number The trajectory of the train is related to the first The trajectory of each train is predicted; for the first train, a supplementary reference trajectory is used. The lead train tracks a preset reference trajectory; the second train predicts and tracks based on the lead train's reference trajectory and historical trajectories; the third and subsequent trains need to simultaneously receive historical trajectory information from the preceding and even earlier trains for collaborative prediction and tracking.

3. The method for predicting the dynamic interval of virtual double-unit high-speed trains based on cooperative trajectory prediction according to claim 1, characterized in that, The construction of an LSTM prior prediction network includes: standardizing all trajectory data; designing an LSTM-based network with dropout mechanism between layers to prevent overfitting; concatenating the standardized data into an input sequence; capturing temporal dependencies through the LSTM's forget gate, input gate, and output gate; outputting prior prediction results; and then inversely standardizing the prior prediction results to finally obtain the train's prior prediction trajectory.

4. The method for predicting the dynamic interval of virtual double-unit high-speed trains based on cooperative trajectory prediction according to claim 1, characterized in that, A state correction iterative model is established, specifically as follows: the position and velocity of the train are extracted from the obtained prior predicted trajectory, the predicted prior acceleration is calculated, and then the acceleration is limited to a range that conforms to reality based on the train's traction / braking characteristics, ensuring that the acceleration does not exceed the train's maximum braking acceleration and maximum traction acceleration; if it exceeds the range, boundary values ​​are used instead. Then, based on the train's longitudinal dynamics equation, the physical correction trajectory is calculated. The data-driven prediction results are fused with the model-driven prediction results, the prior error covariance matrix is ​​calculated, the gain is calculated, and finally, the optimized predicted trajectory that conforms to physical laws after correction is output, and the posterior error covariance matrix is ​​updated.

5. The method for predicting the dynamic interval of virtual double-unit high-speed trains based on cooperative trajectory prediction according to claim 1, characterized in that, A safety boundary model for position uncertainty is established based on the train dynamics model, specifically including: calculating the front and rear envelopes of train operation; and considering positioning errors. Speed ​​measurement error Speed ​​test update frequency Maximum communication delay and train length A location uncertainty safety boundary model is established, and the dynamic safety interval is calculated. Based on the corrected prediction trajectory, the target interval at each prediction time is calculated, and finally the dynamic target interval is obtained to ensure safety throughout the entire prediction period.

6. The method for predicting the dynamic interval of virtual double-unit high-speed trains based on cooperative trajectory prediction according to claim 1, characterized in that, The data-driven and model-driven prediction results are fused to output an optimized predicted trajectory. To obtain the trajectory across the entire prediction time domain, iterative predictions are performed, updating the historical trajectory queue with the current optimized prediction result as input for the next prediction step. Simultaneously, the error covariance matrix... It also iterates and updates at each step, providing a basis for the next fusion filtering.

7. A dynamic interval prediction system for virtual multiple-unit high-speed trains based on cooperative trajectory prediction, characterized in that, include: The acquisition module is used to construct the input dataset for cooperative trajectory prediction, collecting historical trajectory data of each train in the virtual coupled train set. This historical trajectory data includes train position and speed. The historical trajectory of the train and the first The historical trajectory of the train, for the first Predicting the trajectory of trains; The first prediction module is used to construct the LSTM prior prediction network. It standardizes the input data and feeds the standardized spliced ​​sequence into a multi-layer LSTM network. The LSTM network captures long-term dependencies in the sequence through its internal gating mechanism and cell state, and the network outputs the future sequence. The initial predicted trajectory at each time step; The iterative module is used to establish a state correction iterative model and correct parts of the LSTM prediction results that may violate physical laws. Based on the velocity sequence predicted by LSTM, the initial acceleration is obtained through differential calculation. The maximum traction acceleration and maximum braking deceleration of the train at the current predicted speed are queried. The acceleration is limited. Using the limited acceleration, the train state transition matrix is ​​used for forward recursion to obtain the predicted trajectory based entirely on the physical model. The second prediction module is used to establish forward and backward safety boundaries for train position uncertainty based on the train dynamics model, taking into account positioning error, speed measurement error, speed measurement update frequency, maximum communication delay, and train length. It also establishes a position uncertainty safety boundary model. The module calculates dynamic safety intervals and dynamic target intervals. Based on train position uncertainty, it calculates the front and back ends of the train safety envelope. Combined with emergency braking distance, it calculates dynamic safety intervals. Based on the corrected prediction trajectory, it calculates the target intervals at each prediction time, and finally obtains the dynamic target intervals.

8. A non-transitory computer-readable storage medium, characterized in that, The non-transitory computer-readable storage medium is used to store computer instructions, which, when executed by a processor, implement the method for predicting the dynamic interval of virtual double-unit high-speed trains based on cooperative trajectory prediction as described in any one of claims 1-6.

9. A computer device, characterized in that, The system includes a memory and a processor, which communicate with each other. The memory stores program instructions that can be executed by the processor, and the processor calls the program instructions to execute the method for predicting the dynamic interval of virtual double-unit high-speed trains based on cooperative trajectory prediction as described in any one of claims 1-6.

10. An electronic device, characterized in that, include: The electronic device includes a processor, a memory, and a computer program; wherein the processor is connected to the memory, the computer program is stored in the memory, and when the electronic device is running, the processor executes the computer program stored in the memory to cause the electronic device to execute instructions for implementing the method for predicting the dynamic interval of virtual multiple-unit high-speed trains based on cooperative trajectory prediction as described in any one of claims 1-6.

Images