Rail transit multi-system information interaction control method based on train-ground network cooperation

Abstract

The application provides a track traffic multi-system information interaction control method based on train-ground network cooperation, relates to the track traffic control technical field, and comprises the following steps: obtaining train-mounted equipment and ground infrastructure operation data, extracting a semantic feature vector and realizing heterogeneous data standardization, executing train-ground data space-time alignment to establish a synchronous state set, constructing a directed graph structure to identify a key state subset, further constructing a multi-objective optimization function to solve a Pareto optimal solution set, and deriving an optimal cooperative control instruction sequence. The application realizes information interaction and cooperative control among multiple systems, and improves track traffic operation safety and energy utilization efficiency.

Description

Technical Field

[0001] This invention relates to the field of rail transit control technology, and in particular to a multi-system information interaction control method for rail transit based on vehicle-ground network collaboration. Background Technology

[0002] As a crucial component of modern urban transportation systems, urban rail transit plays a vital role in ensuring smooth urban traffic flow through its safe, efficient, and reliable operation. With the continuous expansion of urban rail transit scale and technological advancements, vehicle-to-ground (V2G) network cooperative control technology has become a key means to improve the efficiency and safety of rail transit operations. The V2G cooperative system achieves coordinated control during train operation through information exchange between onboard equipment and ground infrastructure, thereby enhancing the safety and efficiency of train operation.

[0003] Traditional rail transit control systems primarily rely on independently operating onboard equipment and ground facilities for information processing and decision-making, resulting in limited data interaction between the vehicle and the ground, making comprehensive collaborative control difficult. With the development of the Internet of Things (IoT), big data, and artificial intelligence (AI) technologies, vehicle-ground network collaborative control technology is gradually becoming a development trend in rail transit control systems. Through efficient information interaction and intelligent decision-making, it enables precise control and optimized scheduling of train operations. Summary of the Invention

[0004] This invention provides a multi-system information interaction control method for rail transit based on vehicle-ground network collaboration, which can solve the problems in the prior art.

[0005] A first aspect of this invention provides a multi-system information interaction control method for rail transit based on vehicle-to-ground network cooperation, comprising:

[0006] Acquire operational data of train onboard equipment and ground infrastructure;

[0007] Extract semantic feature vectors from the vehicle equipment operation data and infrastructure operation data, learn the inter-domain mapping transformation matrix by minimizing the semantic space distribution distance between the source domain and the target domain, and project heterogeneous data fields onto a unified semantic space to obtain standardized operation data;

[0008] Based on the timestamps and spatial location information in the standardized operational data, calculate the spatiotemporal offset of vehicle-to-ground data and perform spatiotemporal alignment to establish a set of vehicle-to-ground synchronization states.

[0009] The vehicle-to-ground synchronization state set is constructed as a directed graph structure. Directed edges between nodes are established based on temporal mutual information, and topological sorting is performed to obtain a causal sequence. The rate of change of state parameters with respect to control output is calculated, and a subset of key states is selected.

[0010] Based on the aforementioned key state subset and safety constraint rules, a multi-objective optimization function is constructed that minimizes operating adjustment time, energy consumption, and safety margin. The multi-objective optimization function is solved using the particle swarm optimization algorithm to obtain the Pareto optimal solution set. The optimal cooperative control command sequence is then calculated by combining the expert knowledge base rules.

[0011] The optimal cooperative control command sequence is issued to the vehicle-to-ground execution unit, and feedback status is collected.

[0012] Semantic feature vectors of the vehicle-mounted equipment operation data and infrastructure operation data are extracted. An inter-domain mapping transformation matrix is ​​learned by minimizing the semantic space distribution distance between the source and target domains. Heterogeneous data fields are then projected onto a unified semantic space to obtain standardized operation data, including:

[0013] For each data field in the vehicle-mounted equipment operation data, extract the field identifier, data type descriptor, and business context descriptor, input them into the pre-trained semantic encoder, and obtain the source domain semantic feature vector. Perform the same extraction and encoding operations on each data field in the infrastructure operation data to obtain the target domain semantic feature vector.

[0014] The Frobenius norm distance between the covariance matrices of the source domain semantic feature vector and the target domain semantic feature vector is calculated. The inter-domain mapping transformation matrix is ​​initialized and multiplied with the source domain semantic feature vector to obtain the transformed source domain feature vector.

[0015] Calculate the maximum mean difference between the transformed source domain feature vector and the target domain semantic feature vector, construct a loss function in combination with the Frobenius norm distance, and iteratively update the inter-domain mapping transformation matrix using gradient descent until the loss function converges to a preset convergence condition.

[0016] Using the converged inter-domain mapping transformation matrix, a linear transformation is performed on the source domain semantic feature vector to obtain the feature representation in the unified semantic space. Based on the feature representation in the unified semantic space, the standardized format and standardized values ​​of the data fields are reconstructed to obtain standardized running data.

[0017] Based on the timestamps and spatial location information in the standardized operational data, the spatiotemporal offset of the vehicle-to-ground data is calculated and spatiotemporal alignment is performed to establish a vehicle-to-ground synchronization state set, including:

[0018] From the standardized operational data, timestamp information of onboard equipment and infrastructure is extracted respectively. The timestamps of onboard equipment are converted into the local time reference of the train, and the timestamps of infrastructure are converted into the time reference of the ground system. The time offset vector is then calculated.

[0019] Based on the standardized operation data, the track coordinates of the current mileage position of the train and the installation position of the infrastructure are extracted, the spatial distance between the train and the infrastructure at the same moment is calculated, and the spatial offset vector is obtained. The time offset vector and the spatial offset vector are combined to construct the spatiotemporal offset matrix.

[0020] Based on the spatiotemporal offset matrix, the timestamps of the on-board equipment data are corrected to achieve time alignment, and the spatial reference system of the infrastructure data is transformed to achieve spatial registration. The time-aligned train status data and the spatially registered equipment status data are extracted. The spatiotemporal offset matrix is ​​used to establish vehicle-ground data association rules, and the spatially related ground equipment status is found based on the train position. The status parameters that are the same in time and spatially related are combined into a vehicle-ground synchronization status set.

[0021] The vehicle-to-ground synchronization state set is constructed as a directed graph structure. Directed edges between nodes are established based on temporal mutual information, and topological sorting is performed to obtain a causal sequence. The rate of change of state parameters with respect to control output is calculated, and a subset of key states is selected, including:

[0022] For each state parameter in the vehicle-to-ground synchronization state set, a temporal arrangement is performed, the parameter value sequence is extracted within the sliding time window, and the probability density function of the state parameter is calculated by kernel density estimation. The temporal mutual information is calculated based on the integral of the logarithmic ratio of the joint probability density and the marginal probability density.

[0023] Iterate through all state parameter pairs to calculate the temporal mutual information under different time delays. The maximum value of the temporal mutual information is taken as the causal association strength. When it exceeds the temporal threshold, a directed edge is established between the state parameters to form a directed graph structure. Count the number of incoming connections for each state parameter. Add the state parameters with zero incoming connections to the causal sequence in turn, update the number of incoming connections of their associated parameters, and repeat this process until the causal sorting of all state parameters is completed to obtain the causal sequence.

[0024] Based on the causal sequence, a set of all causal paths from each state parameter to the control output is constructed, and the mapping weight matrix from state to control is calculated. The rate of change of the control output is calculated by applying a perturbation to the state parameters, and the state parameters with a rate of change greater than the change threshold are taken as the key state subset.

[0025] Based on the causal sequence, a set of all causal paths from each state parameter to the control output is constructed, and the mapping weight matrix from state to control is calculated. The rate of change of the control output is calculated by applying a perturbation to the state parameters. State parameters with a rate of change greater than a threshold are selected as a subset of key states, including:

[0026] Based on the sorting position of the state parameters in the causal sequence, a breadth-first traversal is performed from the source node of the directed graph. The temporal mutual information values ​​of all node sequences and their connecting edges from each state parameter to the control output are recorded. The node sequences and their edge information are stored as causal paths. All nodes that can reach the control output are traversed to form a complete set of causal paths.

[0027] Calculate the product of the temporal mutual information values ​​on each causal path to obtain the path propagation coefficient. Sum the propagation coefficients of all paths with the state parameter as the starting node to obtain the direct influence factor. Calculate the sum of the product of the path propagation coefficient with the state parameter as the intermediate node and the mutual information of the incoming edge as the indirect influence factor. Combine the direct influence factor to calculate the mapping weight matrix from state to control.

[0028] The standard deviation of the historical sequence of each state parameter is selected as the disturbance reference. Positive and negative disturbances are superimposed on the current value. The change in control output caused by the disturbance is calculated by mapping weights. The difference between the positive and negative disturbance influence values ​​is extracted to obtain the bidirectional change rate of the state parameters.

[0029] The interquartile range is calculated based on the bidirectional rate of change of all state parameters. The sum of the third quartile and the multiples of the interquartile range is set as the screening threshold. State parameters whose rate of change exceeds the screening threshold are extracted and rearranged according to the causal sequence to obtain the key state subset.

[0030] Based on the aforementioned key state subset and safety constraint rules, a multi-objective optimization function is constructed that minimizes operating adjustment time, energy consumption, and maximizes safety margin. The particle swarm optimization algorithm is used to solve the multi-objective optimization function, obtaining a Pareto optimal solution set. The optimal cooperative control command sequence is then calculated using expert knowledge base rules, including:

[0031] The deviation between the current value and the target value of each state parameter in the key state subset is calculated. The running adjustment time is calculated in combination with the control command adjustment rate. The energy consumption coefficient per unit adjustment is obtained. The total energy consumption value is calculated. The minimum interval between the target value of the state parameter and the safety boundary is used as the safety margin. The running adjustment time, total energy consumption value and safety margin are constructed into a multi-objective optimization function.

[0032] Within the constraints of the control command parameters, an initial particle swarm is generated, the particle position and velocity are iteratively updated, the multi-objective optimization function value is non-dominated and the crowding distance is calculated, and the final non-dominated solution set is extracted as the Pareto optimal solution.

[0033] The Pareto optimal solution is substituted into the mapping weight matrix to calculate the state response. The response sensitivity is calculated through perturbation analysis. The current scenario strategy is extracted based on the expert rule base. The node attribute distribution of the key state subset is analyzed. The weights of each objective in the multi-objective optimization function are adjusted. The Pareto optimal solution with the lowest response sensitivity and the largest multi-objective optimization function value is selected as the optimal cooperative control command sequence.

[0034] Within the constraints of the control command parameters, an initial particle swarm is generated, and the particle positions and velocities are iteratively updated. Non-dominated sorting is performed on the multi-objective optimization function values, and crowding distances are calculated. The final non-dominated solution set is extracted as the Pareto optimal solution, including:

[0035] Within the constraints of the control command parameters, the particle position and velocity vectors are randomly initialized to form an initial particle swarm. The running adjustment time, total energy consumption and safety margin objective function value of each particle are calculated. Non-dominated sorting is performed to divide the particles into multiple non-dominated levels.

[0036] For particles in the first non-dominated level, calculate the distance between each particle and its neighboring particles in the three target dimensions, and normalize and sum them to obtain the comprehensive crowding distance. Particles whose comprehensive crowding distance exceeds the density distinction threshold are marked as sparse regions, and particles whose comprehensive crowding distance is below the density distinction threshold are marked as dense regions.

[0037] Based on the non-dominated hierarchy and region labeling of particles, the inertial weight of particles in dense regions is dynamically adjusted to enhance the global search, while the inertial weight of particles in sparse regions is reduced to enhance the local search. The velocity update amount is calculated by combining the historical best position and the global best position of each particle.

[0038] The velocity update is fused with the current velocity and superimposed with the position vector to obtain the new position of the particle. After performing boundary mapping correction, the non-dominated sorting iteration is repeated until the convergence limit is reached. The particles of the final first non-dominated level are extracted as the Pareto optimal solution set.

[0039] A second aspect of the present invention provides a multi-system information interaction control system for rail transit based on vehicle-to-ground network cooperation, comprising:

[0040] The first unit is used to acquire operational data of train onboard equipment and ground infrastructure.

[0041] The second unit is used to extract semantic feature vectors of the vehicle equipment operation data and infrastructure operation data, learn the inter-domain mapping transformation matrix by minimizing the semantic space distribution distance between the source domain and the target domain, and project heterogeneous data fields onto a unified semantic space to obtain standardized operation data.

[0042] The third unit is used to calculate the spatiotemporal offset of vehicle-to-ground data and perform spatiotemporal alignment based on the timestamp and spatial location information in the standardized operation data, and to establish a set of vehicle-to-ground synchronization states.

[0043] The fourth unit is used to construct the vehicle-to-ground synchronization state set into a directed graph structure, establish directed edges between nodes based on temporal mutual information and perform topological sorting to obtain a causal sequence, calculate the rate of change of state parameters with respect to control output, and filter key state subsets.

[0044] The fifth unit is used to construct a multi-objective optimization function based on the key state subset and safety constraint rules, which minimizes the operation adjustment time, energy consumption, and safety margin. The multi-objective optimization function is solved using the particle swarm optimization algorithm to obtain the Pareto optimal solution set. The optimal cooperative control command sequence is then calculated by combining the expert knowledge base rules.

[0045] The sixth unit is used to issue the optimal cooperative control command sequence to the vehicle-to-ground execution unit and collect feedback status.

[0046] A third aspect of the embodiments of the present invention,

[0047] An electronic device is provided, comprising:

[0048] processor;

[0049] Memory used to store processor-executable instructions;

[0050] The processor is configured to invoke instructions stored in the memory to execute the aforementioned method.

[0051] Fourth aspect of the present invention,

[0052] A computer-readable storage medium is provided, having stored thereon computer program instructions that, when executed by a processor, implement the aforementioned method.

[0053] The beneficial effects of this application are as follows:

[0054] By extracting semantic feature vectors from the operational data of onboard equipment and infrastructure, and learning the inter-domain mapping transformation matrix by minimizing the semantic space distribution distance between the source and target domains, a standardized expression of heterogeneous data fields in a unified semantic space is achieved, solving the technical problem of inconsistent data formats and large semantic differences among different subsystems in the rail transit system.

[0055] By constructing the vehicle-to-ground synchronization state set as a directed graph structure, establishing directed edges between nodes based on temporal mutual information and performing topological sorting, the causal relationships between system parameters are scientifically identified. Furthermore, by calculating the rate of change to filter key state subsets, the computational complexity is significantly reduced and the system response speed is improved.

[0056] A multi-objective optimization function was constructed, which includes minimizing operating adjustment time, energy consumption, and maximizing safety margin. The Pareto optimal solution set was obtained by using the particle swarm optimization algorithm. The optimal cooperative control command sequence was calculated by combining the expert knowledge base rules, realizing intelligent cooperative control among multiple systems in rail transit. This optimized the system operating efficiency and energy consumption while ensuring safety. Attached Figure Description

[0057] Figure 1 This is a flowchart illustrating the multi-system information interaction control method for rail transit based on vehicle-to-ground network collaboration, according to an embodiment of the present invention.

[0058] Figure 2 This is a schematic diagram of the calculation method for standardized operating data in an embodiment of the present invention. Detailed Implementation

[0059] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0060] The technical solution of the present invention will be described in detail below with reference to specific embodiments. These specific embodiments can be combined with each other, and the same or similar concepts or processes may not be described again in some embodiments.

[0061] Figure 1 This is a flowchart illustrating the multi-system information interaction control method for rail transit based on vehicle-to-ground network collaboration, as described in an embodiment of the present invention. Figure 1 As shown, the method includes:

[0062] Acquire operational data of train onboard equipment and ground infrastructure;

[0063] Extract semantic feature vectors from the vehicle equipment operation data and infrastructure operation data, learn the inter-domain mapping transformation matrix by minimizing the semantic space distribution distance between the source domain and the target domain, and project heterogeneous data fields onto a unified semantic space to obtain standardized operation data;

[0064] Based on the timestamps and spatial location information in the standardized operational data, calculate the spatiotemporal offset of vehicle-to-ground data and perform spatiotemporal alignment to establish a set of vehicle-to-ground synchronization states.

[0065] The vehicle-to-ground synchronization state set is constructed as a directed graph structure. Directed edges between nodes are established based on temporal mutual information, and topological sorting is performed to obtain a causal sequence. The rate of change of state parameters with respect to control output is calculated, and a subset of key states is selected.

[0066] Based on the aforementioned key state subset and safety constraint rules, a multi-objective optimization function is constructed that minimizes operating adjustment time, energy consumption, and safety margin. The multi-objective optimization function is solved using the particle swarm optimization algorithm to obtain the Pareto optimal solution set. The optimal cooperative control command sequence is then calculated by combining the expert knowledge base rules.

[0067] The optimal cooperative control command sequence is issued to the vehicle-to-ground execution unit, and feedback status is collected.

[0068] Figure 2 This is a schematic diagram of the calculation method for standardized operational data according to an embodiment of the present invention. In an optional implementation, semantic feature vectors of the vehicle-mounted equipment operational data and infrastructure operational data are extracted. An inter-domain mapping transformation matrix is ​​learned by minimizing the semantic space distribution distance between the source and target domains. The heterogeneous data fields are then projected onto a unified semantic space to obtain standardized operational data, including:

[0069] For each data field in the vehicle-mounted equipment operation data, extract the field identifier, data type descriptor, and business context descriptor, input them into the pre-trained semantic encoder, and obtain the source domain semantic feature vector. Perform the same extraction and encoding operations on each data field in the infrastructure operation data to obtain the target domain semantic feature vector.

[0070] The Frobenius norm distance between the covariance matrices of the source domain semantic feature vector and the target domain semantic feature vector is calculated. The inter-domain mapping transformation matrix is ​​initialized and multiplied with the source domain semantic feature vector to obtain the transformed source domain feature vector.

[0071] Calculate the maximum mean difference between the transformed source domain feature vector and the target domain semantic feature vector, construct a loss function in combination with the Frobenius norm distance, and iteratively update the inter-domain mapping transformation matrix using gradient descent until the loss function converges to a preset convergence condition.

[0072] Using the converged inter-domain mapping transformation matrix, a linear transformation is performed on the source domain semantic feature vector to obtain the feature representation in the unified semantic space. Based on the feature representation in the unified semantic space, the standardized format and standardized values ​​of the data fields are reconstructed to obtain standardized running data.

[0073] In this specific embodiment, semantic information is extracted from each data field in the vehicle-mounted device's operational data. This operational data includes vehicle speed, location information, door status, etc., and each of these data fields has a specific identifier, type description, and business context. For example, for the speed data field, its identifier "SPEED", data type "float", and business context "real-time vehicle speed" are extracted. After extraction, these three descriptors are concatenated into the text sequence "SPEED float vehicle real-time speed", which is then input into a pre-trained BERT semantic encoder. This encoder uses a 12-layer Transformer structure. After the input text sequence is processed by a multi-head attention mechanism, a 768-dimensional vector representation is extracted from the [CLS] position of the last layer as the source domain semantic feature vector.

[0074] The same operation is performed on infrastructure operation data, which includes signal system status, power supply parameters, etc. Taking traffic light status as an example, the identifier "SIGNAL_STATUS", data type "enum" and business context "track signal indication status" are extracted and concatenated into the text "SIGNAL_STATUS enum track signal indication status". This text is then input into the BERT encoder to obtain the semantic feature vector of the target domain.

[0075] Calculate the inter-domain distribution distance by forming a matrix X from all source domain semantic feature vectors and a matrix Y from all target domain semantic feature vectors, and then calculate their covariance matrix C. x and C y The covariance matrix reflects the correlation structure between the semantic features of each field. This can be achieved by calculating the Frobenius norm distance D = ||C| between two covariance matrices. x -C y ||F is used to quantify the difference in distribution between the two domains. The inter-domain mapping transformation matrix W is initialized as the identity matrix and multiplied with the source domain eigenvector X to obtain the transformed source domain eigenvector X'=WX.

[0076] The maximum mean difference (MMD) between the transformed source domain eigenvector X' and the target domain eigenvector Y is calculated. MMD measures the difference in distribution by calculating the mean distance between the two sets of vectors in the reproducing kernel Hilbert space, using the Gaussian kernel function K(x,y)=exp(-||xy||). 2 / 2σ 2 ), where σ is the kernel width parameter, set as the median distance between the feature vectors of the source domain and the target domain, and MMD can be expressed as the average distance between the features of the two domains in the kernel space.

[0077] Construct the loss function L=MMD(X',Y)+λ||C x '-C y||F, where λ is the balancing parameter, set to 0.5, uses the Adam optimizer to perform gradient descent, with a learning rate of 0.001, iteratively updates the mapping matrix W, and considers the preset convergence condition to be met when the loss change is less than the threshold of 0.0001 for five consecutive iterations, and stops iterating.

[0078] Using the converged mapping matrix W, a linear transformation is performed on the semantic feature vectors of the source domain to obtain the feature representation Z=WX in the unified semantic space. Based on these unified representations, the data fields are standardized and reconstructed. Specifically, for each transformed feature vector z, its cosine similarity with all semantic feature vectors of the target domain is calculated, and the target domain field format with the highest similarity is selected as the standardized format.

[0079] For numerical standardization, a value mapping function is constructed based on the field correspondence. For example, for speed values ​​in vehicle-mounted devices, if the source domain unit is "km / h" while the standard format is "m / s", then a unit conversion is performed, dividing the value by 3.6. For enumeration types, such as door status, a mapping table is established to map the "0 / 1" of the source domain to the "CLOSED / OPEN" of the standard format.

[0080] Through the above processing, the operational data of on-board equipment and infrastructure are represented in a unified semantic space, and standardized operational data is generated. This standardized data maintains the information integrity of the original data while achieving a unified format, facilitating subsequent comprehensive analysis and application.

[0081] In practical applications, for a subway line's train operation monitoring system, the "vehicle speed" field (unit: km / h, floating-point) recorded by the onboard ATP equipment and the "trainSpeed" field (unit: m / s, floating-point) recorded by the ground ATS system are mapped to a unified semantic space using this method. They are identified as fields expressing the same semantics and standardized as the "vehicleSpeed" field (unit: m / s, floating-point). This approach resolves the inconsistency in data representation between different systems, providing a unified data foundation for rail transit operation analysis.

[0082] In one optional implementation, based on the timestamps and spatial location information in the standardized operational data, the vehicle-to-ground data spatiotemporal offset is calculated and spatiotemporal alignment is performed to establish a vehicle-to-ground synchronization state set, including:

[0083] From the standardized operational data, timestamp information of onboard equipment and infrastructure is extracted respectively. The timestamps of onboard equipment are converted into the local time reference of the train, and the timestamps of infrastructure are converted into the time reference of the ground system. The time offset vector is then calculated.

[0084] Based on the standardized operation data, the track coordinates of the current mileage position of the train and the installation position of the infrastructure are extracted, the spatial distance between the train and the infrastructure at the same moment is calculated, and the spatial offset vector is obtained. The time offset vector and the spatial offset vector are combined to construct the spatiotemporal offset matrix.

[0085] Based on the spatiotemporal offset matrix, the timestamps of the on-board equipment data are corrected to achieve time alignment, and the spatial reference system of the infrastructure data is transformed to achieve spatial registration. The time-aligned train status data and the spatially registered equipment status data are extracted. The spatiotemporal offset matrix is ​​used to establish vehicle-ground data association rules, and the spatially related ground equipment status is found based on the train position. The status parameters that are the same in time and spatially related are combined into a vehicle-ground synchronization status set.

[0086] In practical applications, it is necessary to extract the timestamp information of onboard equipment and infrastructure from the standardized operation data. The timestamp of onboard equipment is usually based on the onboard clock system, while the timestamp of infrastructure is based on the ground system clock. Due to differences in hardware and synchronization mechanisms, these two clock systems have time reference deviations. Therefore, it is necessary to convert the timestamp of onboard equipment to the train's local time reference and the timestamp of infrastructure to the ground system time reference.

[0087] During the conversion process, a time base mapping function is used. For onboard equipment, the deviation parameter between the device's local clock and the standard time source is used for correction. For example, if the onboard system clock is five seconds behind the standard time source, this five-second deviation value needs to be added during conversion. Similarly, a similar conversion is performed on infrastructure timestamps to ensure the time bases of the two systems are comparable. The two converted time series are compared to calculate a time offset vector, which reflects the distribution of time differences between the onboard system and the ground system.

[0088] Based on standardized operational data, the track coordinates of the train's current mileage position and the infrastructure installation positions are extracted. The train's position is typically obtained through an onboard positioning system, such as a track odometer, GNSS positioning, or a multi-sensor fusion positioning system; while the location of the infrastructure is a fixed installation coordinate point. By comparing the train's position at the same moment with the locations of the infrastructure along the line, the spatial distance between them is calculated, thus obtaining a spatial offset vector.

[0089] In a real-world application scenario, suppose a train is passing through a certain section. The onboard data shows that the current train mileage position is K256+500. At the same time, the recorded signal status data comes from the signal installed at K256+680. At this time, the spatial offset is 180 meters, which represents the actual distance between the train and the signal.

[0090] The time offset vector and spatial offset vector obtained above are combined to construct a spatiotemporal offset matrix. This matrix contains the correspondence between the time dimension and the spatial dimension. Each element represents the spatiotemporal relationship between the train and the infrastructure at a specific location at a specific moment. The matrix is ​​constructed using a pairing mapping method, which matches the time synchronization point with the spatial reference point to form a complete spatiotemporal mapping relationship table.

[0091] Based on the constructed spatiotemporal offset matrix, the timestamps of the vehicle-mounted equipment data are corrected to achieve time alignment. The correction process employs a time offset compensation algorithm, adjusting the original timestamps according to the time offset values ​​in the offset matrix to ensure consistency between the vehicle-mounted equipment data and the ground system in the time dimension. Simultaneously, the spatial reference frame of the infrastructure data is transformed to achieve spatial registration. This process uses a coordinate transformation function to unify spatial data from different reference frames into the same coordinate system, ensuring the comparability of spatial locations.

[0092] After alignment, the time-aligned train status data and the spatially registered equipment status data are extracted. The train status data includes parameters such as speed, acceleration, and operating mode, while the equipment status data includes information such as turnout position, signal display status, and track circuit occupancy status. After time-space alignment processing, these data can accurately reflect the overall system status under the same time and space conditions.

[0093] A vehicle-to-ground data association rule is established using a spatiotemporal offset matrix. This rule defines how to find the status of spatially related ground equipment based on the train's location. The association rule adopts a spatial nearest neighbor search algorithm. Taking the current position of the train as the center, a reasonable spatial window range is set (such as 2000 meters in front and 500 meters behind), and all infrastructures located within this range are filtered out and their real-time status data is obtained.

[0094] For example, when a train runs to a certain section, according to the spatiotemporal association rules, it can automatically identify the status of all switches, signal displays and track occupancy within one kilometer ahead, providing the necessary environmental perception information for train operation control. It combines time-related and space-related state parameters into a vehicle-ground synchronized state set. Each record in this set contains a timestamp, train state parameters (position, speed, etc.) and corresponding space-related ground equipment state parameters, forming a complete vehicle-ground cooperative state snapshot.

[0095] By using the above-mentioned spatiotemporal alignment method for vehicle-to-ground data, the problem of spatiotemporal benchmark differences between different data sources is solved, providing a reliable data foundation for vehicle-to-ground collaborative analysis and decision-making, and effectively supporting the application of intelligent railway operation control and safety monitoring.

[0096] In one optional implementation, the vehicle-to-ground synchronization state set is constructed as a directed graph structure. Directed edges between nodes are established based on temporal mutual information, and topological sorting is performed to obtain a causal sequence. The rate of change of state parameters with respect to control output is calculated, and a subset of key states is selected, including:

[0097] For each state parameter in the vehicle-to-ground synchronization state set, a temporal arrangement is performed, the parameter value sequence is extracted within the sliding time window, and the probability density function of the state parameter is calculated by kernel density estimation. The temporal mutual information is calculated based on the integral of the logarithmic ratio of the joint probability density and the marginal probability density.

[0098] Iterate through all state parameter pairs to calculate the temporal mutual information under different time delays. The maximum value of the temporal mutual information is taken as the causal association strength. When it exceeds the temporal threshold, a directed edge is established between the state parameters to form a directed graph structure. Count the number of incoming connections for each state parameter. Add the state parameters with zero incoming connections to the causal sequence in turn, update the number of incoming connections of their associated parameters, and repeat this process until the causal sorting of all state parameters is completed to obtain the causal sequence.

[0099] Based on the causal sequence, a set of all causal paths from each state parameter to the control output is constructed, and the mapping weight matrix from state to control is calculated. The rate of change of the control output is calculated by applying a perturbation to the state parameters, and the state parameters with a rate of change greater than the change threshold are taken as the key state subset.

[0100] In this specific embodiment, a set of vehicle-to-ground synchronization states from the vehicle control system is collected. This set includes vehicle system state parameters and ground control system state parameters, including but not limited to: vehicle speed, position, acceleration, steering angle, battery level, ground control commands, and communication delay. After collection, these state parameters are time-series arranged. By setting an appropriate sampling frequency (e.g., ten times per second), the changes in state parameters over a period of time (e.g., thirty minutes) are recorded. Based on the time-series data, a sliding time window (e.g., ten seconds) is set, and the numerical sequence of each state parameter within the window is extracted from the time-series data.

[0101] For the numerical sequence of state parameters within each sliding window, the kernel density estimation method is used to calculate the probability density function. Specifically, a Gaussian kernel function is used to smooth each data point, and the bandwidth parameter is automatically determined by the Silverman rule, thus obtaining the marginal probability density function of each state parameter. For any two state parameters, their joint probability density function is calculated. Based on the obtained probability density function, the temporal mutual information between the state parameters is calculated. The temporal mutual information is calculated by integrating the logarithm ratio of the joint probability density to the marginal probability density. For example, for state parameters X and Y, the temporal mutual information under time delay τ can be expressed as the integral of the joint probability P(X(t), Y(t+τ)) with the logarithm ratio of the marginal probabilities P(X(t)) and P(Y(t+τ)).

[0102] Iterate through all possible pairs of state parameters and calculate the temporal mutual information under different time delays (e.g., from 0 to 5 seconds, with a step size of 0.1 seconds). For each pair of parameters, record the maximum value of the temporal mutual information as the causal correlation strength and record the corresponding optimal time delay. When the causal correlation strength exceeds the preset temporal threshold (set according to system characteristics, such as 0.3), it is considered that there is a causal relationship between the two state parameters, and a directed edge is established between the corresponding state parameters, thereby forming a directed graph structure.

[0103] After constructing the directed graph, a topological sort is performed to obtain the causal sequence. The number of incoming connections for each state parameter is counted, and state parameters with zero incoming connections (i.e., source nodes that are not affected by other parameters) are identified and added to the causal sequence. The parameter is then removed from the directed graph, and the number of incoming connections for its associated parameters is updated. The above process is repeated, and parameters with zero incoming connections are added to the causal sequence in turn until all state parameters are sorted.

[0104] Based on the obtained causal sequence, a set of all possible causal paths from each state parameter to the control output is constructed. For example, if the causal sequence shows that parameter A affects parameter B, and parameter B affects the control output C, then there exists a causal path from A to C: A→B→C. By identifying all such paths, a comprehensive understanding of how state parameters affect the control output can be obtained.

[0105] To quantify the impact of state parameters on the control output, a small perturbation (such as increasing or decreasing its value by 5%) is applied to each state parameter, and the change in the control output is observed. Specifically, for the state parameter x... i Apply perturbation Δx i Measure the change Δy in the control output y, and calculate the rate of change r. i =Δy / Δx iThis rate of change reflects the sensitivity of the state parameters to the control output. State parameters with a rate of change greater than a preset threshold (such as 0.1) are identified as key state parameters and form a key state subset. These parameters have a significant impact on the control output and should be given priority in the design of the control algorithm.

[0106] In practical applications, by analyzing the relationship between vehicle lateral position, lateral velocity, lateral acceleration, steering wheel angle, lane departure, and other state parameters and steering control output, it can be identified that steering wheel angle and lateral acceleration are the two most critical state parameters affecting steering control. This result can be used to optimize autonomous driving control algorithms and improve the accuracy and response speed of lateral control. This method is also applicable to the cooperative control scenario between intelligent connected vehicles and roadside equipment. By analyzing the causal relationship between vehicle state, roadside perception data, and control commands, communication delay and roadside predicted trajectory can be identified as key state parameters affecting the cooperative control effect, thereby enabling targeted optimization of vehicle-road cooperative control strategies.

[0107] In one optional implementation, based on the causal sequence, a set of all causal paths from each state parameter to the control output is constructed, and a mapping weight matrix from state to control is calculated. The rate of change of the control output is calculated by applying a perturbation to the state parameters, and state parameters with a rate of change greater than a change threshold are selected as a subset of key states, including:

[0108] Based on the sorting position of the state parameters in the causal sequence, a breadth-first traversal is performed from the source node of the directed graph. The temporal mutual information values ​​of all node sequences and their connecting edges from each state parameter to the control output are recorded. The node sequences and their edge information are stored as causal paths. All nodes that can reach the control output are traversed to form a complete set of causal paths.

[0109] Calculate the product of the temporal mutual information values ​​on each causal path to obtain the path propagation coefficient. Sum the propagation coefficients of all paths with the state parameter as the starting node to obtain the direct influence factor. Calculate the sum of the product of the path propagation coefficient with the state parameter as the intermediate node and the mutual information of the incoming edge as the indirect influence factor. Combine the direct influence factor to calculate the mapping weight matrix from state to control.

[0110] The standard deviation of the historical sequence of each state parameter is selected as the disturbance reference. Positive and negative disturbances are superimposed on the current value. The change in control output caused by the disturbance is calculated by mapping weights. The difference between the positive and negative disturbance influence values ​​is extracted to obtain the bidirectional change rate of the state parameters.

[0111] The interquartile range is calculated based on the bidirectional rate of change of all state parameters. The sum of the third quartile and the multiples of the interquartile range is set as the screening threshold. State parameters whose rate of change exceeds the screening threshold are extracted and rearranged according to the causal sequence to obtain the key state subset.

[0112] In this specific embodiment, a causal directed graph model of the vehicle-to-ground network system is obtained, where nodes represent state parameters and control outputs, and edges represent the influence relationships between nodes. State parameters include train speed, acceleration, position information, electromechanical equipment operating status, track status, etc., and control outputs include train traction force, braking force, signal system control commands, etc.

[0113] Based on the sorting position of the state parameters in the causal sequence, a breadth-first traversal is performed starting from the source node of the directed graph. During the traversal, the temporal mutual information value of all node sequences from each state parameter to the control output and the connecting edges is recorded. The temporal mutual information value is calculated by analyzing historical data and represents the degree of influence of changes in state parameters on subsequent nodes.

[0114] Taking train speed state parameters as an example, a breadth-first traversal yields the following paths: train speed → braking distance → braking force control output, and train speed → energy consumption estimation → traction control output. For each path, the temporal mutual information values ​​of each edge are recorded; for example, the mutual information value between speed and braking distance is 0.82, and the mutual information value between braking distance and braking force is 0.91. The node sequences and edge information obtained from the traversal are stored as a causal path set. For each state parameter in the rail transit system, all possible paths from it to each control output are counted, forming a complete causal path set. For example, there are multiple paths from track state parameters to speed control outputs, including directly influential paths and indirect paths through other state parameters.

[0115] Calculate the path propagation coefficient for each causal path, which is the product of the temporal mutual information values ​​of all edges on the path. For example, the propagation coefficient for the path train speed → braking distance → braking force control output is 0.82 × 0.91 = 0.746. For a certain state parameter, sum the path propagation coefficients of all paths starting from it to obtain the direct influence factor. For example, the direct influence factors of train speed on braking force and traction force are 0.746 and 0.623, respectively. Calculate the indirect influence of each state parameter as an intermediate node. For example, when the current speed affects braking control as an intermediate node, sum the product of the path propagation coefficients of all paths passing through speed nodes to the braking control output and the mutual information of the incoming edges, as the indirect influence factor. Combine the direct and indirect influence factors to obtain the mapping weight from state to control.

[0116] For each state parameter in the vehicle-to-ground cooperative system, the standard deviation of its historical sequence data is selected as the parameter disturbance reference. For example, if the standard deviation of the historical data of the train speed parameter is 2.5 km / h, then the disturbance reference is set to 2.5 km / h. Positive disturbance (current value plus standard deviation) and negative disturbance (current value minus standard deviation) are superimposed on the current speed value. Through the mapping weights calculated above, the change in control output caused by the disturbance is obtained.

[0117] Assuming the train's current speed is 60 km / h, the positive disturbance is 62.5 km / h, and the negative disturbance is 57.5 km / h, calculations using mapping weights show that the positive disturbance increases braking force by 8.2%, while the negative disturbance decreases it by 7.9%. Extracting the difference in the impact values ​​of the positive and negative disturbances, the bidirectional change rate of the speed parameter on braking control is found to be 16.1%. The bidirectional change rates of all state parameters in the vehicle-to-ground network are sorted, and the interquartile range is calculated. For example, the first quartile of all parameter change rates is 3.2%, the third quartile is 12.8%, and the interquartile range is 9.6%. A screening threshold is set to the third quartile plus 1.5 times the interquartile range, i.e., 12.8% + 9.6% × 1.5 = 27.2%. State parameters with change rates exceeding the screening threshold are extracted and rearranged according to a causal sequence to obtain a subset of critical states.

[0118] In rail transit applications, the critical state subset includes parameters such as train position, speed, track curvature, gradient, and turnout status. These parameters have a significant impact on control output and require priority monitoring and handling. By filtering the critical state subset, the load on vehicle-to-ground communication can be reduced, and the response speed of the control system can be improved. Based on the filtered critical state subset, precise control of multiple rail transit systems can be achieved. For example, when a change in the critical state parameter of track curvature is detected, the system can quickly adjust the train speed control strategy to ensure safe operation. At the same time, non-critical parameters can be acquired and transmitted using low-frequency strategies, reducing the consumption of vehicle-to-ground communication resources.

[0119] The above method can effectively identify key state parameters affecting rail transit control systems, improve system response speed and control accuracy, and provide technical support for multi-system information interaction control in rail transit with vehicle-ground network collaboration. This method is applicable to various rail transit scenarios, including intercity railways, high-speed railways, and urban rail transit, and has broad application prospects.

[0120] In one optional implementation, based on the critical state subset and safety constraint rules, a multi-objective optimization function is constructed that minimizes operating adjustment time, energy consumption, and maximizes safety margin. The multi-objective optimization function is solved using a particle swarm optimization algorithm to obtain a Pareto optimal solution set. Finally, the optimal cooperative control command sequence is calculated using expert knowledge base rules, including:

[0121] The deviation between the current value and the target value of each state parameter in the key state subset is calculated. The running adjustment time is calculated in combination with the control command adjustment rate. The energy consumption coefficient per unit adjustment is obtained. The total energy consumption value is calculated. The minimum interval between the target value of the state parameter and the safety boundary is used as the safety margin. The running adjustment time, total energy consumption value and safety margin are constructed into a multi-objective optimization function.

[0122] Within the constraints of the control command parameters, an initial particle swarm is generated, the particle position and velocity are iteratively updated, the multi-objective optimization function value is non-dominated and the crowding distance is calculated, and the final non-dominated solution set is extracted as the Pareto optimal solution.

[0123] The Pareto optimal solution is substituted into the mapping weight matrix to calculate the state response. The response sensitivity is calculated through perturbation analysis. The current scenario strategy is extracted based on the expert rule base. The node attribute distribution of the key state subset is analyzed. The weights of each objective in the multi-objective optimization function are adjusted. The Pareto optimal solution with the lowest response sensitivity and the largest multi-objective optimization function value is selected as the optimal cooperative control command sequence.

[0124] In this specific embodiment, it is necessary to calculate the deviation between the current value and the target value of each parameter in the critical state subset, that is, for the critical state subset S={s1, s2, ..., s... n Each state parameter s in} i Let n be the total number of state parameters in the key state subset, and calculate its current value s. i_current With target value s i_target Deviation δ i =|s i_current -s i_target | Then, the rate r is adjusted in conjunction with control commands. i Calculate the running adjustment time T i =δ i / r i The maximum value among all parameter adjustment times is taken as the overall system adjustment time T = max(T1, T2, ..., T). n ).

[0125] Obtain the energy consumption coefficient e of unit adjustment i This represents the energy consumption required for each unit of state parameter adjustment. The total energy consumption value E is calculated as the product of the state parameter adjustment and the energy consumption coefficient: E = Σ(δ). i ×e i Meanwhile, the minimum interval between the target value of the statistical state parameter and the safety boundary is used as the safety margin S=min(|s i_target -s i_boundary |), where s i_boundary Represents the state parameter s i The safety boundary value.

[0126] Based on the above calculation results, a multi-objective optimization function F=[f1(T), f2(E), f3(S)] is constructed, where f1(T) represents the adjustment time objective function with a value of T; f2(E) represents the energy consumption objective function with a value of E; and f3(S) represents the safety margin objective function with a value of -S (the negative sign indicates that the larger the safety margin, the better). The goal of the multi-objective optimization is to minimize f1(T) and f2(E) and maximize f3(S).

[0127] In the multi-objective optimization process, an initial particle swarm is generated within the constraints of the control command parameters. For an N-dimensional control command parameter space, M initial particles are randomly generated, and each particle is represented as an N-dimensional vector X. i =(x i1 , x i2 ,..., x iN ), where i = 1, 2, ..., M, and each parameter satisfies the constraint condition x. ij ∈[x min_j , x max_j At the same time, an initial velocity vector V is assigned to each particle. i =(v i1 , v i2 , ..., v iN The speed range is limited to [-v]. max , v max ]Inside.

[0128] The iterative optimization process of the particle swarm optimization algorithm includes: updating the position and velocity of each particle, evaluating the multi-objective function value, performing non-dominated sorting, and calculating the crowding distance. Specifically, in the (t+1)th iteration, the velocity and position update formula for particle i is: V i (t+1)=w×V i (t)+c1×r1×[p besti -X i (t)]+c2×r2×[g best -X i (t)],X i (t+1)=X i (t)+V i (t+1), where w is the inertia weight, c1 and c2 are learning factors, r1 and r2 are random numbers in the interval [0,1], and p besti For the historical best position of particle i, g best This is the globally optimal position.

[0129] The updated particle swarm is evaluated using multi-objective function values. The particles are stratified using a non-dominated sorting method to determine the Pareto front. For particles in the same non-dominated layer, their crowding distance is calculated to maintain the diversity of solutions. The crowding distance is calculated based on the normalized range of each objective function value, representing the density of particle distribution in the solution space. After a preset number of iterations, the final non-dominated solution set is extracted as the Pareto optimal solution.

[0130] Substituting each Pareto optimal solution into the mapping weight matrix W, the state response vector R = W × X is calculated, where X is the control command vector and R is the expected state response. Through perturbation analysis, a small perturbation δ is applied to the control command X. X Calculate the change in response δ R =W×δ X The response sensitivity S is obtained. r =|δ R | / |δ X This value reflects the system's sensitivity to changes in control commands.

[0131] Simultaneously, based on the expert rule base, decision-making strategies applicable to the current scenario are extracted, such as "high safety margin priority," "low energy consumption priority," or "rapid response priority." According to the extracted strategies and the node attribute distribution characteristics of the key state subset, the weights α1, α2, and α3 of each objective in the multi-objective optimization function are dynamically adjusted to form a weighted objective function F. w =α1×f1(T)+α2×f2(E)+α3×f3(S), from the Pareto optimal solution set, select the response sensitivity S. r The lowest weighted objective function value F w The optimal solution is used as the final control scheme and is converted into a specific control command sequence output. This command sequence satisfies the requirements of multi-objective optimization and has low response sensitivity, enabling it to maintain stable performance when the system parameters fluctuate slightly.

[0132] In one optional implementation, an initial particle swarm is generated within the constraints of the control command parameters, the particle positions and velocities are iteratively updated, non-dominated sorting is performed on the multi-objective optimization function values ​​and crowding distance is calculated, and the final non-dominated solution set is extracted as the Pareto optimal solution, including:

[0133] Within the constraints of the control command parameters, the particle position and velocity vectors are randomly initialized to form an initial particle swarm. The running adjustment time, total energy consumption and safety margin objective function value of each particle are calculated. Non-dominated sorting is performed to divide the particles into multiple non-dominated levels.

[0134] For particles in the first non-dominated level, calculate the distance between each particle and its neighboring particles in the three target dimensions, and normalize and sum them to obtain the comprehensive crowding distance. Particles whose comprehensive crowding distance exceeds the density distinction threshold are marked as sparse regions, and particles whose comprehensive crowding distance is below the density distinction threshold are marked as dense regions.

[0135] Based on the non-dominated hierarchy and region labeling of particles, the inertial weight of particles in dense regions is dynamically adjusted to enhance the global search, while the inertial weight of particles in sparse regions is reduced to enhance the local search. The velocity update amount is calculated by combining the historical best position and the global best position of each particle.

[0136] The velocity update is fused with the current velocity and superimposed with the position vector to obtain the new position of the particle. After performing boundary mapping correction, the non-dominated sorting iteration is repeated until the convergence limit is reached. The particles of the final first non-dominated level are extracted as the Pareto optimal solution set.

[0137] In this specific embodiment, the particle swarm is randomly initialized within the constraints of the control command parameters. For a population of N particles, each particle is represented as a D-dimensional vector, representing different parameters of the control command. For example, if the control command includes three parameters: velocity, acceleration, and time, then D=3, and the position vector of particle i can be represented as X. i =[x i1 , x i2 , ...,x iD The value range of each dimension is limited by the physical constraints of the system, and the velocity vector V i =[v i1 , v i2 , ..., v iD [] represents the particle's velocity in each dimension. The initial position is constrained by parameters in each dimension [L]. min , L max Randomly generated within [], initial velocity set to [-V] max V max A random value within the range of ], where V max It is usually set to 10%-20% of the parameter range.

[0138] After initialization, calculate multiple objective function values ​​for each particle. For control system optimization, three main objective functions are considered: running settling time f1(X). i ), representing the time required for the system to reach the target state from the initial state; total energy consumption f2(X) i ), representing the energy consumption during the entire control process; safety margin f3(X) i The distance between the system's operating state and the danger boundary is represented by three objectives, which are often in conflict. For example, reducing settling time often increases energy consumption or reduces safety margin.

[0139] After calculating the objective function value, non-dominated sorting is performed to divide the particles into different levels. If particle a is not inferior to particle b on all targets and is superior to b on at least one target, then a is said to dominate b. Domination relationship comparison is performed on all particles. Particles that are not dominated by any other particles form the first non-dominated level. From the remaining particles, non-dominated particles are searched again to form the second non-dominated level, and so on, until all particles are assigned to the corresponding non-dominated level.

[0140] For particles within the first non-dominated level, crowding distances are calculated to maintain solution diversity. For each objective function, particles are first sorted according to the objective value, and the distances between adjacent particles in that objective dimension are calculated. For boundary particles (i.e., particles with the largest or smallest objective value), the crowding distance is set to infinity. For intermediate particles, their crowding distance in objective j is calculated as the normalized difference between two adjacent particles in that objective. The sum of the crowding distances in the three objective dimensions yields the comprehensive crowding distance. A density discrimination threshold D is then set. c (Typically 0.8-1.2 times the average congestion distance), the congestion distance will be greater than D. c The particles are labeled as sparse regions, smaller than D. c The marked area is a dense region.

[0141] Based on the particle non-dominated hierarchy and region labeling dynamic adjustment optimization strategy, for particles in dense regions, their inertia weight w is increased, typically from 0.5 to 0.7-0.9, to enhance global search capability; for particles in sparse regions, the inertia weight is decreased to 0.2-0.4 to enhance local search. The individual learning factor c1 and the social learning factor c2 are typically set to 1.5-2.5. The velocity update formula for particle i considers its historical optimal position P. besti and the nearest position G among the globally non-dominated particles besti The updated speed needs to be limited to [-V] max V max Within the range.

[0142] After the position is updated, check if it exceeds the parameter constraint boundary. If it does, use the boundary mapping method to map the particle position back to the effective area. At the same time, adjust the velocity direction in the opposite direction to prevent the particles from continuously rushing out of the boundary. After updating the position, recalculate the objective function value and perform non-dominated sorting.

[0143] The iterative process continues until the preset maximum number of iterations (usually 100-300 times) is reached or the number of consecutive iterations (e.g., 20 times) shows little change in the particles of the first non-dominated level, indicating that the algorithm has basically converged. Finally, all particles of the first non-dominated level are extracted to form a Pareto optimal solution set, providing decision-makers with a series of non-dominated alternatives.

[0144] In practical applications, algorithm parameters can be adjusted according to the specific system characteristics. For example, for control systems with high response time requirements, the population size can be increased to 50-100 to improve search efficiency; for complex nonlinear systems, the maximum number of iterations can be increased to more than 500 to ensure that the algorithm converges fully; the density discrimination threshold can also be dynamically adjusted according to the characteristics of the problem. A smaller value can be set in the early stage of the search to expand the exploration range, and then gradually increased in the later stage to refine the distribution of the optimal solution.

[0145] This method can effectively balance global search and local refinement. By dynamically adjusting the particle behavior characteristics, it can improve search efficiency while maintaining the diversity of solutions, providing a practical and effective solution for multi-objective optimization of complex industrial control systems.

[0146] This invention relates to a multi-system information interaction control system for rail transit based on vehicle-to-ground network collaboration, comprising:

[0147] The first unit is used to acquire operational data of train onboard equipment and ground infrastructure.

[0148] The second unit is used to extract semantic feature vectors of the vehicle equipment operation data and infrastructure operation data, learn the inter-domain mapping transformation matrix by minimizing the semantic space distribution distance between the source domain and the target domain, and project heterogeneous data fields onto a unified semantic space to obtain standardized operation data.

[0149] The third unit is used to calculate the spatiotemporal offset of vehicle-to-ground data and perform spatiotemporal alignment based on the timestamp and spatial location information in the standardized operation data, and to establish a set of vehicle-to-ground synchronization states.

[0150] The fourth unit is used to construct the vehicle-to-ground synchronization state set into a directed graph structure, establish directed edges between nodes based on temporal mutual information and perform topological sorting to obtain a causal sequence, calculate the rate of change of state parameters with respect to control output, and filter key state subsets.

[0151] The fifth unit is used to construct a multi-objective optimization function based on the key state subset and safety constraint rules, which minimizes the operation adjustment time, energy consumption, and safety margin. The multi-objective optimization function is solved using the particle swarm optimization algorithm to obtain the Pareto optimal solution set. The optimal cooperative control command sequence is then calculated by combining the expert knowledge base rules.

[0152] The sixth unit is used to issue the optimal cooperative control command sequence to the vehicle-to-ground execution unit and collect feedback status.

[0153] A third aspect of the present invention provides an electronic device, comprising:

[0154] processor;

[0155] Memory used to store processor-executable instructions;

[0156] The processor is configured to invoke instructions stored in the memory to execute the aforementioned method.

[0157] A fourth aspect of the present invention provides a computer-readable storage medium having stored thereon computer program instructions that, when executed by a processor, implement the aforementioned method.

[0158] This invention can be a method, apparatus, system, and / or computer program product. The computer program product may include a computer-readable storage medium having computer-readable program instructions loaded thereon for performing various aspects of the invention.

[0159] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A multi-system information interaction control method for rail transit based on vehicle-ground network collaboration, characterized in that, include: Acquire operational data of train onboard equipment and ground infrastructure; Extract semantic feature vectors from the vehicle equipment operation data and infrastructure operation data, learn the inter-domain mapping transformation matrix by minimizing the semantic space distribution distance between the source domain and the target domain, and project heterogeneous data fields onto a unified semantic space to obtain standardized operation data; Based on the timestamps and spatial location information in the standardized operational data, calculate the spatiotemporal offset of vehicle-to-ground data and perform spatiotemporal alignment to establish a set of vehicle-to-ground synchronization states. The vehicle-to-ground synchronization state set is constructed as a directed graph structure. Directed edges between nodes are established based on temporal mutual information, and topological sorting is performed to obtain a causal sequence. The rate of change of state parameters with respect to control output is calculated, and a subset of key states is selected, including: For each state parameter in the vehicle-to-ground synchronization state set, a temporal arrangement is performed, the parameter value sequence is extracted within the sliding time window, and the probability density function of the state parameter is calculated by kernel density estimation. The temporal mutual information is calculated based on the integral of the logarithmic ratio of the joint probability density and the marginal probability density. The temporal mutual information of all state parameter pairs is calculated under different time delays. The maximum value of the temporal mutual information is taken as the causal association strength. When it exceeds the temporal threshold, directed edges are established between the state parameters to form a directed graph structure. The number of incoming connections of each state parameter is counted. State parameters with zero incoming connections are added to the causal sequence in turn, and the number of incoming connections of their associated parameters is updated. The number of incoming connections of each state parameter is counted and updated repeatedly until the causal sorting of all state parameters is completed, and the causal sequence is obtained. Based on the causal sequence, a set of all causal paths from each state parameter to the control output is constructed, and the mapping weight matrix from state to control is calculated. The rate of change of the control output is calculated by applying a perturbation to the state parameters, and the state parameters with a rate of change greater than the change threshold are taken as the key state subset. Based on the aforementioned key state subset and safety constraint rules, a multi-objective optimization function is constructed that minimizes operating adjustment time, energy consumption, and safety margin. The multi-objective optimization function is solved using the particle swarm optimization algorithm to obtain the Pareto optimal solution set. The optimal cooperative control command sequence is then calculated by combining the expert knowledge base rules. The optimal cooperative control command sequence is issued to the vehicle-to-ground execution unit, and feedback status is collected.

2. The method according to claim 1, characterized in that, Semantic feature vectors of the vehicle-mounted equipment operation data and infrastructure operation data are extracted. An inter-domain mapping transformation matrix is ​​learned by minimizing the semantic space distribution distance between the source and target domains. Heterogeneous data fields are then projected onto a unified semantic space to obtain standardized operation data, including: For each data field in the vehicle-mounted equipment operation data, extract the field identifier, data type descriptor, and business context descriptor, input them into the pre-trained semantic encoder, and obtain the source domain semantic feature vector. Perform the same extraction and encoding operations on each data field in the infrastructure operation data to obtain the target domain semantic feature vector. The Frobenius norm distance between the covariance matrices of the source domain semantic feature vector and the target domain semantic feature vector is calculated. The inter-domain mapping transformation matrix is ​​initialized and multiplied with the source domain semantic feature vector to obtain the transformed source domain feature vector. Calculate the maximum mean difference between the transformed source domain feature vector and the target domain semantic feature vector, construct a loss function in combination with the Frobenius norm distance, and iteratively update the inter-domain mapping transformation matrix using gradient descent until the loss function converges to a preset convergence condition. Using the converged inter-domain mapping transformation matrix, a linear transformation is performed on the source domain semantic feature vector to obtain the feature representation in the unified semantic space. Based on the feature representation in the unified semantic space, the standardized format and standardized values ​​of the data fields are reconstructed to obtain standardized running data.

3. The method according to claim 1, characterized in that, Based on the timestamps and spatial location information in the standardized operational data, the spatiotemporal offset of the vehicle-to-ground data is calculated and spatiotemporal alignment is performed to establish a vehicle-to-ground synchronization state set, including: From the standardized operational data, timestamp information of onboard equipment and infrastructure is extracted respectively. The timestamps of onboard equipment are converted into the local time reference of the train, and the timestamps of infrastructure are converted into the time reference of the ground system. The time offset vector is then calculated. Based on the standardized operation data, the track coordinates of the current mileage position of the train and the installation position of the infrastructure are extracted, the spatial distance between the train and the infrastructure at the same moment is calculated, and the spatial offset vector is obtained. The time offset vector and the spatial offset vector are combined to construct the spatiotemporal offset matrix. Based on the spatiotemporal offset matrix, the timestamps of the on-board equipment data are corrected to achieve time alignment, and the spatial reference system of the infrastructure data is transformed to achieve spatial registration. The time-aligned train status data and the spatially registered equipment status data are extracted. The spatiotemporal offset matrix is ​​used to establish vehicle-ground data association rules, and the spatially related ground equipment status is found based on the train position. The status parameters that are the same in time and spatially related are combined into a vehicle-ground synchronization status set.

4. The method according to claim 1, characterized in that, Based on the causal sequence, a set of all causal paths from each state parameter to the control output is constructed, and the mapping weight matrix from state to control is calculated. The rate of change of the control output is calculated by applying a perturbation to the state parameters. State parameters with a rate of change greater than a threshold are selected as a subset of key states, including: Based on the sorting position of the state parameters in the causal sequence, a breadth-first traversal is performed from the source node of the directed graph. The temporal mutual information values ​​of all node sequences and their connecting edges from each state parameter to the control output are recorded. The node sequences and their edge information are stored as causal paths. All nodes that can reach the control output are traversed to form a complete set of causal paths. Calculate the product of the temporal mutual information values ​​on each causal path to obtain the path propagation coefficient. Sum the propagation coefficients of all paths with the state parameter as the starting node to obtain the direct influence factor. Calculate the sum of the product of the path propagation coefficient with the state parameter as the intermediate node and the mutual information of the incoming edge as the indirect influence factor. Combine the direct influence factor to calculate the mapping weight matrix from state to control. The standard deviation of the historical sequence of each state parameter is selected as the disturbance reference. Positive and negative disturbances are superimposed on the current value. The change in control output caused by the disturbance is calculated by mapping weights. The difference between the positive and negative disturbance influence values ​​is extracted to obtain the bidirectional change rate of the state parameters. The interquartile range is calculated based on the bidirectional rate of change of all state parameters. The sum of the third quartile and the multiples of the interquartile range is set as the screening threshold. State parameters whose rate of change exceeds the screening threshold are extracted and rearranged according to the causal sequence to obtain the key state subset.

5. The method according to claim 1, characterized in that, Based on the aforementioned key state subset and safety constraint rules, a multi-objective optimization function is constructed that minimizes operating adjustment time, energy consumption, and maximizes safety margin. The particle swarm optimization algorithm is used to solve the multi-objective optimization function, obtaining a Pareto optimal solution set. The optimal cooperative control command sequence is then calculated using expert knowledge base rules, including: The deviation between the current value and the target value of each state parameter in the key state subset is calculated. The running adjustment time is calculated in combination with the control command adjustment rate. The energy consumption coefficient per unit adjustment is obtained. The total energy consumption value is calculated. The minimum interval between the target value of the state parameter and the safety boundary is used as the safety margin. The running adjustment time, total energy consumption value and safety margin are constructed into a multi-objective optimization function. Within the constraints of the control command parameters, an initial particle swarm is generated, the particle position and velocity are iteratively updated, the multi-objective optimization function value is non-dominated and the crowding distance is calculated, and the final non-dominated solution set is extracted as the Pareto optimal solution. The Pareto optimal solution is substituted into the mapping weight matrix to calculate the state response. The response sensitivity is calculated through perturbation analysis. The current scenario strategy is extracted based on the expert rule base. The node attribute distribution of the key state subset is analyzed. The weights of each objective in the multi-objective optimization function are adjusted. The Pareto optimal solution with the lowest response sensitivity and the largest multi-objective optimization function value is selected as the optimal cooperative control command sequence.

6. The method according to claim 5, characterized in that, Within the constraints of the control command parameters, an initial particle swarm is generated, and the particle positions and velocities are iteratively updated. Non-dominated sorting is performed on the multi-objective optimization function values, and crowding distances are calculated. The final non-dominated solution set is extracted as the Pareto optimal solution, including: Within the constraints of the control command parameters, the particle position and velocity vectors are randomly initialized to form an initial particle swarm. The running adjustment time, total energy consumption and safety margin objective function value of each particle are calculated. Non-dominated sorting is performed to divide the particles into multiple non-dominated levels. For particles in the first non-dominated level, calculate the distance between each particle and its neighboring particles in the three target dimensions, and normalize and sum them to obtain the comprehensive crowding distance. Particles whose comprehensive crowding distance exceeds the density distinction threshold are marked as sparse regions, and particles whose comprehensive crowding distance is below the density distinction threshold are marked as dense regions. Based on the non-dominated hierarchy and region labeling of particles, the inertial weight of particles in dense regions is dynamically adjusted to enhance the global search, while the inertial weight of particles in sparse regions is reduced to enhance the local search. The velocity update amount is calculated by combining the historical best position and the global best position of each particle. The velocity update is fused with the current velocity and superimposed with the position vector to obtain the new position of the particle. After performing boundary mapping correction, the non-dominated sorting iteration is repeated until the convergence limit is reached. The particles of the final first non-dominated level are extracted as the Pareto optimal solution set.

7. A multi-system information interaction control system for rail transit based on vehicle-to-ground network collaboration, used to implement the method as described in any one of claims 1-6, characterized in that, include: The first unit is used to acquire operational data of train onboard equipment and ground infrastructure. The second unit is used to extract semantic feature vectors of the vehicle equipment operation data and infrastructure operation data, learn the inter-domain mapping transformation matrix by minimizing the semantic space distribution distance between the source domain and the target domain, and project heterogeneous data fields onto a unified semantic space to obtain standardized operation data. The third unit is used to calculate the spatiotemporal offset of vehicle-to-ground data and perform spatiotemporal alignment based on the timestamp and spatial location information in the standardized operation data, and to establish a set of vehicle-to-ground synchronization states. The fourth unit is used to construct the vehicle-to-ground synchronization state set into a directed graph structure, establish directed edges between nodes based on temporal mutual information and perform topological sorting to obtain a causal sequence, calculate the rate of change of state parameters with respect to control output, and filter key state subsets. The fifth unit is used to construct a multi-objective optimization function based on the key state subset and safety constraint rules, which minimizes the operation adjustment time, energy consumption, and safety margin. The multi-objective optimization function is solved using the particle swarm optimization algorithm to obtain the Pareto optimal solution set. The optimal cooperative control command sequence is then calculated by combining the expert knowledge base rules. The sixth unit is used to issue the optimal cooperative control command sequence to the vehicle-to-ground execution unit and collect feedback status.

8. An electronic device, characterized in that, include: processor; Memory used to store processor-executable instructions; The processor is configured to invoke instructions stored in the memory to execute the method according to any one of claims 1 to 6.

9. A computer-readable storage medium having computer program instructions stored thereon, characterized in that, When the computer program instructions are executed by the processor, they implement the method described in any one of claims 1 to 6.

Images