A train control and management system based on vehicle-mounted multi-bus fusion
By deploying three independent sensing paths and combining Bayesian confidence assessment and digital twin models, the problems of sensing blind spots and energy consumption in train control systems under complex operating conditions were solved, achieving a dynamic balance between safety and energy consumption, and improving the sensing integrity and real-time performance of the train.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- BEIJING QIFAN FUTURE TECH CO LTD
- Filing Date
- 2026-03-31
- Publication Date
- 2026-06-23
AI Technical Summary
Existing train control systems are prone to sensor path failure and data quality degradation under complex conditions such as high-speed operation, sudden load changes, or strong electromagnetic/vibration interference. They also lack dynamic confidence quantification and forward-looking prediction, making it difficult to balance safety assurance and energy consumption optimization.
Three independent sensing paths are deployed. Through Bayesian confidence assessment and cross-validation, combined with a digital twin model for Monte Carlo simulation, dynamic acquisition cycle adjustment and adaptive switching of bus fusion strategy are achieved. A state-space-based train dynamics model is constructed to predict safety risks and energy consumption optimization.
It improves the completeness and real-time performance of perception under complex operating conditions, reduces system power consumption, and enhances safety margin and energy utilization efficiency in scenarios with high-speed changes, drastic load fluctuations, or strong interference.
Smart Images
Figure CN122268906A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of rail transit vehicle-mounted network control technology, specifically to a vehicle-mounted multi-bus integrated train control and management system. Background Technology
[0002] With the rapid development of rail transit towards higher speeds, greater intelligence, and greener operating systems, train control and management systems are placing increasingly higher demands on the accuracy, real-time performance, and reliability of environmental perception. Existing train control systems generally employ multiple sensors to collect train operating status data and perform data transmission and fusion processing via onboard networks. Some research introduces digital twin technology to construct train dynamics models, achieving state prediction through simulation, and attempts to improve system intelligence through multi-sensor fusion sensing or vehicle-to-ground wireless communication. However, existing technologies still have the following major shortcomings:
[0003] The lack of independent dedicated sensing paths for speed changes, load fluctuations and external interference means that under complex operating conditions such as high-speed operation, sudden load changes or strong electromagnetic / vibration interference, the failure of a single path can easily lead to overall sensing blind spots or data quality degradation.
[0004] Relying solely on fixed thresholds or empirical weights makes it impossible to achieve dynamic and continuous confidence quantification and discrete quality grading based on data variance. This makes it difficult to effectively distinguish between high-confidence and low-confidence paths, resulting in fusion results being susceptible to noise or anomalous data contamination.
[0005] The lack of forward-looking prediction of safety risks and energy consumption optimization in the next control cycle means that digital twin models are mostly used for offline simulation or passive monitoring. They are not combined with real-time confidence to perform Monte Carlo probability simulation, and cannot actively drive the adjustment of control strategies based on the probability of safety risks and energy consumption reduction.
[0006] Most vehicle-mounted multi-bus fusion strategies are configured with static weights and fixed sampling rates, which cannot dynamically adjust weights, adaptively switch sampling rates, or isolate low-confidence paths or shut down redundant channels according to operating conditions. This makes it difficult for the system to achieve a dynamic balance between safety assurance and energy consumption optimization, resulting in high overall power consumption and easy jitter during the switching process.
[0007] The information disclosed in the background section is only intended to enhance the understanding of the background of this disclosure, and therefore may include information that does not constitute prior art known to those skilled in the art. Summary of the Invention
[0008] The purpose of this invention is to provide a vehicle-mounted multi-bus integrated train control and management system to solve the problems mentioned in the background art.
[0009] To achieve the above objectives, the present invention provides the following technical solution: a vehicle-mounted multi-bus integrated train control and management system, specifically including: a multi-path acquisition module, a confidence assessment module, a cross-validation module, and a simulation prediction module;
[0010] Multi-path acquisition module: Deploys three independent sensing paths, which are uploaded in parallel to the fusion processor through different vehicle-mounted multi-buses at a period of 100ms;
[0011] Confidence assessment module: Calculates confidence scores for the estimated data output by each sensing path based on Bayesian methods, determines whether the path is credible, and maps it to discrete quality levels;
[0012] Cross-validation module: Performs three-way voting through weighted K-fold cross-validation, calculates the consistency index, and compares the threshold judgment instructions;
[0013] Simulation prediction module: Constructs a digital twin model, performs Monte Carlo simulation for the next cycle in case of conflict, and outputs the probability of the result to switch the bus fusion management strategy.
[0014] As a preferred embodiment of the on-board multi-bus integrated train control and management system described in this invention, wherein:
[0015] The three independent sensing paths include:
[0016] Path 1: Real-time acquisition of longitudinal acceleration and wheel speed pulse counts of the train through inertial measurement units and multi-axis wheel speed sensor arrays;
[0017] The estimated data of the actual speed change of the train is calculated by fusing longitudinal acceleration and wheel speed pulse counts according to the acquisition period using Kalman filtering;
[0018] Path 2: Real-time acquisition of coupler force and axle load during train operation via longitudinal and lateral force sensors and axle load sensor arrays;
[0019] Estimated data equivalent to train operating load are calculated by weighted average fusion.
[0020] Path 3: Real-time acquisition of electromagnetic interference intensity and vibration acceleration of train traffic through electromagnetic interference sensor and vibration acceleration sensor group;
[0021] Estimated data on train operation interference intensity were calculated by normalizing and fusing the data.
[0022] The acquisition cycle is 100ms. When any of the above paths changes drastically, the acquisition cycle is increased to 120-150ms based on the dynamic instructions of the fusion processor.
[0023] The drastic changes in any path include: estimated actual train speed changes > 5 km / h / s, estimated equivalent train load > 10%, and estimated train disturbance intensity > 7.
[0024] As a preferred embodiment of the on-board multi-bus integrated train control and management system described in this invention, wherein:
[0025] The confidence score is calculated for the estimated data output by each sensing path using a precision-based Bayesian approach. The specific calculation formula is as follows:
[0026]
[0027] in, Let represent the confidence score of the estimated data output by the i-th perception path. This represents the variance of the estimated data corresponding to the i-th perception path within the current evaluation window. , These represent the preset empirical hyperparameters, where =2.5, =0.8;
[0028] when When the value is ≥0.75, the path is considered a reliable path, indicating that the estimated data quality of the path is acceptable.
[0029] when When the value is less than 0.75, the path is judged as a suspicious path, indicating that the estimated data quality of the path is low. Its weight is reduced during voting and it is temporarily isolated in the priority protection mode.
[0030] correspond The rate of rapid credit default begins when the value is approximately 0.90 to 1.0.
[0031] The hyperparameters , The selection of [the option] makes the confidence function more accurate. It exhibits rapid transition characteristics within the interval ≈[0.3, 1.5];
[0032] The The mapping is to the following discrete quality levels, specifically including:
[0033] when If the value is ≥0.90, the rating is: Highly reliable.
[0034] When 0.75≤ <0.90, then the grade is: qualified;
[0035] When 0.50≤ If the value is less than 0.75, the level is: suspicious.
[0036] when If the value is less than 0.50, the rating is: Low Trust / Failure.
[0037] As a preferred embodiment of the on-board multi-bus integrated train control and management system described in this invention, wherein:
[0038] Using the data from path 1 as the validation set and the data from path 2 and path 3 as the training set, we can train to obtain the cross-validation state estimate for path 1.
[0039] Using the data from path 2 as the validation set and the data from path 1 and path 3 as the training set, we can train to obtain the cross-validation state estimate for path 2.
[0040] Using the data from path 3 as the validation set and the data from path 1 and path 2 as the training set, we can train a cross-validation state estimate for path 3.
[0041] Through the above three rounds of mutual verification, we obtain the estimated data of the steady state of each path under the supervision of other paths;
[0042] The consistency index is calculated based on the average of the estimated steady-state data of each path monitored by other paths. The specific formula is as follows:
[0043]
[0044] in, Indicators of consistency Let represent the confidence score of the estimated data output by the i-th perception path. The average value represents the estimated steady-state data of each path under the supervision of other paths. This is the tolerance threshold;
[0045] when When the value is ≥2.1, the estimated data of the three paths are determined to be consistent, a consistent result is output, and the current control command is executed.
[0046] when When the value is less than 2.1, it is determined that there is a conflict in the estimated data of the three-hop path, the conflict result is output and the digital twin prediction instruction is executed.
[0047] As a preferred embodiment of the on-board multi-bus integrated train control and management system described in this invention, wherein:
[0048] Constructing a digital twin model based on state-space representation;
[0049] The state transition matrix, control input matrix, and output matrix of the digital twin model are identified offline based on the train's dynamic parameters, mass, moment of inertia, and air resistance.
[0050] The covariance matrix of the process noise and measurement noise of the digital twin model is adjusted in real time based on the confidence scores of the three paths.
[0051] Using the estimated data from the fusion of the three paths at the current moment as the initial state, 500 independent random simulations are performed in Monte Carlo mode for the next control cycle, specifically including:
[0052] The current control command applied to the train at the current moment is generated and mapped by the actuator according to the current control command;
[0053] The estimated data for the next cycle time is calculated iteratively based on the state-space equation;
[0054] Analyze whether the results of each independent random simulation meet the following conditions:
[0055] Safety risk indicators are less than 5%;
[0056] Energy consumption can be reduced by more than 10% compared to the current benchmark;
[0057] The results of 500 independent random simulations are statistically analyzed, and the percentage of simulations with a safety risk of less than 5%, the percentage of simulations that can reduce energy consumption by more than 10%, and the remaining probability corresponding to the equilibrium mode are output.
[0058] The above probability results are sent to the bus fusion management strategy switching module as the direct driving basis for mode switching:
[0059] When the percentage of simulations with a safety risk of less than 5% is greater than 0.7, switch to priority protection mode;
[0060] When the percentage of simulations that can reduce energy consumption by more than 10% is greater than 0.6, switch to energy-saving optimization mode;
[0061] Otherwise, maintain the equilibrium mode.
[0062] On the other hand, the present invention provides a computer device including a memory and a processor, wherein the memory stores a computer program, wherein when the computer program is executed by the processor, it implements the steps of an on-board multi-bus integrated train control and management system as described above.
[0063] On the other hand, the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein: when the computer program is executed by a processor, it implements the steps of an on-board multi-bus integrated train control and management system as described above.
[0064] The technical effects and advantages provided by the present invention in the above technical solution are as follows:
[0065] (1) Three independent sensing paths are deployed to collect data on speed changes, load fluctuations and external interference respectively. The acquisition cycle is adaptively adjusted according to drastic changes. Combined with the differentiated priority upload of MVB, ETB and CAN three buses, the system blind spots caused by the failure of a single sensor or bus are effectively avoided, and the sensing integrity and real-time performance under complex working conditions are significantly improved.
[0066] (2) By using a specific Sigmoid form of Bayesian confidence calculation formula, the path data precision is mapped to a continuous score of [0,1] and further discretized into four levels: high confidence, qualified, doubtful, low confidence / failure. This automatically increases the weight of high-precision paths, effectively suppresses the influence of low-precision paths, and provides strong rapid transition characteristics for confidence assessment.
[0067] (3) By using K=3 cross-validation with three paths of data serving as training / validation sets, a weighted consistency index is calculated. When the consistency index is ≥2.1, the control command is executed directly, and when the consistency index is <2.1, a prediction is triggered. In abnormal situations, conflicts are quickly identified, taking into account both decision reliability and response speed, which is superior to traditional fixed threshold or simple majority voting methods.
[0068] (4) Construct a digital twin model of train dynamics based on state-space equations, and adjust the covariance matrix of process noise and measurement noise in real time according to path confidence. In case of conflict, use the current fusion state as the initial value and perform 500 Monte Carlo simulations for the next cycle, upgrading passive fusion to active decision-making, fundamentally improving the safety margin and energy utilization efficiency of trains in high-speed sudden changes, drastic load fluctuations or strong interference scenarios.
[0069] (5) Based on the digital twin prediction probability, the system intelligently switches between the balanced mode, the priority protection mode and the energy-saving optimization mode, and uses linear interpolation to achieve a smooth transition of ≤200ms. This realizes the dynamic adaptive fusion of the three buses MVB, ETB and CAN in terms of weight, sampling rate and channel switching. While ensuring the determinism and low latency of the safety-critical bus, it significantly reduces the overall power consumption of the system. Attached Figure Description
[0070] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in this invention. For those skilled in the art, other drawings can be obtained based on these drawings.
[0071] Figure 1 This is a flowchart of a train control and management system based on the present invention, which integrates multiple on-board buses.
[0072] Figure 2This is a schematic diagram of a vehicle-mounted multi-bus integrated train control and management system according to the present invention. Detailed Implementation
[0073] Exemplary embodiments will now be described more fully with reference to the accompanying drawings. However, these exemplary embodiments can be implemented in many forms and should not be construed as limited to the examples set forth herein; rather, they are provided so that the description of this disclosure will be more complete and fully convey the concept of the exemplary embodiments to those skilled in the art.
[0074] Example 1, referring to Figure 1 and Figure 2 This is the first embodiment of the present invention. This embodiment provides a vehicle-mounted multi-bus integrated train control and management system, specifically including: a multi-path acquisition module, a confidence assessment module, a cross-validation module, and a simulation prediction module;
[0075] Multi-path acquisition module: Deploys three independent sensing paths, which are uploaded in parallel to the fusion processor through different vehicle-mounted multi-buses at a period of 100ms;
[0076] The three independent sensing paths include:
[0077] Path 1: Real-time acquisition of longitudinal acceleration and wheel speed pulse counts of the train through inertial measurement units and multi-axis wheel speed sensor arrays;
[0078] The estimated data of the actual speed change of the train is calculated by fusing longitudinal acceleration and wheel speed pulse counts according to the acquisition period using Kalman filtering;
[0079] Path 2: Real-time acquisition of coupler force and axle load during train operation via longitudinal and lateral force sensors and axle load sensor arrays;
[0080] Estimated data equivalent to train operating load are calculated by weighted average fusion.
[0081] Path 3: Real-time acquisition of electromagnetic interference intensity and vibration acceleration of train traffic through electromagnetic interference sensor and vibration acceleration sensor group;
[0082] Estimated data on train operation interference intensity were calculated by normalizing and fusing the data.
[0083] The acquisition cycle is 100ms. When any of the above paths changes drastically, the acquisition cycle is increased to 120-150ms based on the dynamic instructions of the fusion processor.
[0084] The drastic changes in any path include: estimated actual train speed changes > 5 km / h / s, estimated equivalent train load > 10%, and estimated train disturbance intensity > 7.
[0085] The first path is to upload via the multi-function vehicle bus (MVB is a train safety critical bus with strong determinism and low latency <10ms).
[0086] Path 2 prioritizes uploading via the twisted-wire / Ethernet train backbone network (suitable for larger bandwidth and train-level data).
[0087] Data from path 3 is preferentially uploaded via the redundant controller local area network bus (which has strong anti-electromagnetic interference capabilities and is suitable for data in noisy environments).
[0088] Confidence assessment module: Calculates confidence scores for the estimated data output by each sensing path based on Bayesian methods, determines whether the path is credible, and maps it to discrete quality levels;
[0089] The confidence score is calculated for the estimated data output by each sensing path using a precision-based Bayesian approach. The specific calculation formula is as follows:
[0090]
[0091] in, Let represent the confidence score of the estimated data output by the i-th perception path. This represents the variance of the estimated data corresponding to the i-th perception path within the current evaluation window. , These represent the preset empirical hyperparameters, where =2.5, =0.8;
[0092] when When the value is ≥0.75, the path is considered a reliable path, indicating that the estimated data quality of the path is acceptable.
[0093] when When the value is less than 0.75, the path is judged as a suspicious path, indicating that the estimated data quality of the path is low. Its weight is reduced during voting and it is temporarily isolated in the priority protection mode.
[0094] correspond The rate of rapid credit default begins when the value is approximately 0.90 to 1.0.
[0095] The hyperparameters , The selection of [the option] makes the confidence function more accurate. It exhibits rapid transition characteristics within the interval ≈[0.3, 1.5];
[0096] The The mapping is to the following discrete quality levels for direct use by the downstream cross-validation voting module and the bus fusion strategy switching module, specifically including:
[0097] when If the value is ≥0.90, the rating is: Highly reliable.
[0098] When 0.75≤ <0.90, then the grade is: qualified;
[0099] When 0.50≤ If the value is less than 0.75, the level is: suspicious.
[0100] when If the value is less than 0.50, the rating is: Low Trust / Failure.
[0101] Cross-validation module: Performs three-way voting through weighted K-fold cross-validation, calculates the consistency index, and compares the threshold judgment instructions;
[0102] Using the data from path 1 as the validation set and the data from path 2 and path 3 as the training set, we can train to obtain the cross-validation state estimate for path 1.
[0103] Using the data from path 2 as the validation set and the data from path 1 and path 3 as the training set, we can train to obtain the cross-validation state estimate for path 2.
[0104] Using the data from path 3 as the validation set and the data from path 1 and path 2 as the training set, we can train a cross-validation state estimate for path 3.
[0105] Through the above three rounds of mutual verification, we obtain the estimated data of the steady state of each path under the supervision of other paths;
[0106] The consistency index is calculated based on the average of the estimated steady-state data of each path monitored by other paths. The specific formula is as follows:
[0107]
[0108] in, Indicators of consistency Let represent the confidence score of the estimated data output by the i-th perception path. The average value represents the estimated steady-state data of each path under the supervision of other paths. This is the tolerance threshold;
[0109] when When the value is ≥2.1, the estimated data of the three paths are determined to be consistent, a consistent result is output, and the current control command is executed.
[0110] when When the value is less than 2.1, it is determined that there is a conflict in the estimated data of the three-hop path, the conflict result is output and the digital twin prediction instruction is executed;
[0111] The tolerance threshold is 0.05. Based on a large number of real-vehicle statistics, this value can make 92% of the cycles consistent under normal train operation and quickly identify conflicts under abnormal operation.
[0112] The threshold is 2.1. If the confidence level of all three paths is 1, then at least two paths need to agree in order to pass the threshold. An additional margin of 0.1 is left so that a path with a low confidence level is allowed to deviate slightly without affecting the overall judgment.
[0113] Simulation prediction module: Constructs a digital twin model, performs Monte Carlo simulation for the next cycle in case of conflict, and outputs the probability of the result to switch the bus fusion management strategy;
[0114] Constructing a digital twin model based on state-space representation;
[0115] The state transition matrix, control input matrix, and output matrix of the digital twin model are identified offline based on the train's dynamic parameters, mass, moment of inertia, and air resistance.
[0116] The covariance matrix of the process noise and measurement noise of the digital twin model is adjusted in real time based on the confidence scores of the three paths.
[0117] Using the estimated data from the fusion of the three paths at the current moment as the initial state, 500 independent random simulations are performed in Monte Carlo mode for the next control cycle, specifically including:
[0118] The current control command applied to the train at the current moment is generated and mapped by the actuator according to the current control command;
[0119] The estimated data for the next cycle time is calculated iteratively based on the state-space equation;
[0120] Analyze whether the results of each independent random simulation meet the following conditions:
[0121] Safety risk indicators are less than 5%;
[0122] Energy consumption can be reduced by more than 10% compared to the current benchmark;
[0123] The results of 500 independent random simulations are statistically analyzed, and the percentage of simulations with a safety risk of less than 5%, the percentage of simulations that can reduce energy consumption by more than 10%, and the remaining probability corresponding to the equilibrium mode are output.
[0124] The above probability results are sent to the bus fusion management strategy switching module as the direct driving basis for mode switching:
[0125] When the percentage of simulations with a safety risk of less than 5% is greater than 0.7, switch to priority protection mode;
[0126] When the percentage of simulations that can reduce energy consumption by more than 10% is greater than 0.6, switch to energy-saving optimization mode;
[0127] Otherwise, maintain the equilibrium mode.
[0128] Three independent sensing paths are deployed to collect data specifically for speed changes, load fluctuations, and external interference. The data collection cycle is adaptively adjusted according to drastic changes. Combined with differentiated priority uploading of data via MVB, ETB, and CAN buses, the system blind spots caused by the failure of a single sensor or bus are effectively avoided, and the sensing integrity and real-time performance under complex operating conditions are significantly improved.
[0129] By using a specific Sigmoid-form Bayesian confidence calculation formula, the path data precision is mapped to a continuous score of [0,1] and further discretized into four levels: high confidence, qualified, doubtful, and low confidence / failure. This automatically increases the weight of high-precision paths, effectively suppresses the influence of low-precision paths, and provides strong rapid transition characteristics for confidence assessment.
[0130] By using K=3 cross-validation with three paths of data serving as training / validation sets, a weighted consensus index is calculated. When the consensus index is ≥2.1, control commands are executed directly, and when the consensus index is <2.1, a prediction is triggered. In abnormal situations, conflicts are quickly identified, balancing decision reliability and response speed, which is superior to traditional fixed threshold or simple majority voting methods.
[0131] A digital twin model of train dynamics based on state-space equations is constructed, and the covariance matrices of process noise and measurement noise are adjusted in real time according to path confidence. In the event of a conflict, the current fusion state is used as the initial value to perform 500 Monte Carlo simulations for the next cycle, upgrading passive fusion to active decision-making, fundamentally improving the safety margin and energy utilization efficiency of the train under high-speed sudden changes, drastic load fluctuations, or strong interference scenarios.
[0132] Based on the digital twin's predictive probability, the system intelligently switches between the balanced mode, the priority protection mode, and the energy-saving optimization mode, and uses linear interpolation to achieve a smooth transition of ≤200ms. This enables dynamic adaptive fusion of the MVB, ETB, and CAN buses in terms of weight, sampling rate, and channel switching. While ensuring the determinism and low latency of the safety-critical bus, this significantly reduces the overall power consumption of the system.
[0133] The foregoing has only described certain exemplary embodiments of the present invention by way of illustration. Undoubtedly, those skilled in the art can modify the described embodiments in various ways without departing from the spirit and scope of the present invention. Therefore, the foregoing drawings and descriptions are illustrative in nature and should not be construed as limiting the scope of protection of the claims of the present invention.
Claims
1. A vehicle-mounted multi-bus integrated train control and management system, characterized in that, Specifically, it includes: Multi-path acquisition module, confidence assessment module, cross-validation module, simulation prediction module; Multi-path acquisition module: Deploys three independent sensing paths, which are uploaded in parallel to the fusion processor through different vehicle-mounted multi-buses at a period of 100ms; Confidence assessment module: Calculates confidence scores for the estimated data output by each sensing path based on Bayesian methods, determines whether the path is credible, and maps it to discrete quality levels; Cross-validation module: Performs three-way voting through weighted K-fold cross-validation, calculates the consistency index, and compares the threshold judgment instructions; Simulation prediction module: Constructs a digital twin model, performs Monte Carlo simulation for the next cycle in case of conflict, and outputs the probability of the result to switch the bus fusion management strategy.
2. The on-board multi-bus integrated train control and management system according to claim 1, characterized in that: In the multi-path acquisition module, the three independent sensing paths include: Path 1: Real-time acquisition of longitudinal acceleration and wheel speed pulse counts of the train through inertial measurement units and multi-axis wheel speed sensor arrays; The estimated data of the actual speed change of the train is calculated by fusing longitudinal acceleration and wheel speed pulse counts according to the acquisition period using Kalman filtering; Path 2: Real-time acquisition of coupler force and axle load during train operation via longitudinal and lateral force sensors and axle load sensor arrays; Estimated data equivalent to train operating load are calculated by weighted average fusion. Path 3: Real-time acquisition of electromagnetic interference intensity and vibration acceleration of train traffic operation through electromagnetic interference sensor and vibration acceleration sensor group.
3. The on-board multi-bus integrated train control and management system according to claim 1, characterized in that: In the confidence assessment module, a confidence score is calculated for the estimated data output by each sensing path based on the accuracy of the Bayesian formula. when When the value is ≥0.75, the path is considered a reliable path, indicating that the estimated data quality of the path is acceptable. when When the value is less than 0.75, the path is judged as a suspicious path, indicating that the estimated data quality of the path is low. Its weight is reduced during voting and it is temporarily isolated in the priority protection mode. The The mapping is to the following discrete quality levels, specifically including: when If the value is ≥0.90, the rating is: Highly reliable. When 0.75≤ <0.90, then the grade is: qualified; When 0.50≤ If the value is less than 0.75, the level is: suspicious. when If the value is less than 0.50, the rating is: Low Trust / Failure.
4. The on-board multi-bus integrated train control and management system according to claim 1, characterized in that: The specific formula for calculating the confidence score is as follows: in, Let represent the confidence score of the estimated data output by the i-th perception path. This represents the variance of the estimated data corresponding to the i-th perception path within the current evaluation window. , These represent the preset empirical hyperparameters, where =2.5, =0.
8.
5. The on-board multi-bus integrated train control and management system according to claim 1, characterized in that: In the cross-validation module, the data of path 1 is used as the validation set, and the data of path 2 and path 3 are used as the training set to train and obtain the cross-validation state estimate for path 1. Using the data from path 2 as the validation set and the data from path 1 and path 3 as the training set, we can train to obtain the cross-validation state estimate for path 2. Using the data from path 3 as the validation set and the data from path 1 and path 2 as the training set, we can train a cross-validation state estimate for path 3. Through the above three rounds of mutual verification, we obtain the estimated data of the steady state of each path under the supervision of other paths; The consistency index is calculated based on the average of the estimated steady state data of each path monitored on other paths. when When the value is ≥2.1, the estimated data of the three paths are determined to be consistent, a consistent result is output, and the current control command is executed. when When the value is less than 2.1, it is determined that there is a conflict in the estimated data of the three-hop path, the conflict result is output and the digital twin prediction instruction is executed.
6. The on-board multi-bus integrated train control and management system according to claim 1, characterized in that: The specific formula for calculating the consistency index is as follows: in, Indicators of consistency Let represent the confidence score of the estimated data output by the i-th perception path. The average value represents the estimated steady-state data of each path under the supervision of other paths. This is the tolerance threshold.
7. The on-board multi-bus integrated train control and management system according to claim 1, characterized in that: In the simulation prediction module, a digital twin model is constructed based on state-space representation; Using the estimated data from the fusion of the three paths at the current moment as the initial state, 500 independent random simulations are performed in Monte Carlo mode for the next control cycle, specifically including: The current control command applied to the train at the current moment is generated and mapped by the actuator according to the current control command; The estimated data for the next cycle time is calculated iteratively based on the state-space equation; The results of 500 independent random simulations are statistically analyzed, and the percentage of simulations with a safety risk of less than 5%, the percentage of simulations that can reduce energy consumption by more than 10%, and the remaining probability corresponding to the equilibrium mode are output. The above probability results are sent to the bus fusion management strategy switching module as the direct driving basis for mode switching: When the percentage of simulations with a safety risk of less than 5% is greater than 0.7, switch to priority protection mode; When the percentage of simulations that can reduce energy consumption by more than 10% is greater than 0.6, switch to energy-saving optimization mode; Otherwise, maintain the equilibrium mode.
8. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that: When the processor executes the computer program, it implements a module of the on-board multi-bus integrated train control and management system according to any one of claims 1 to 7.
9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by the processor, it implements a module of the on-board multi-bus integrated train control and management system as described in any one of claims 1 to 7.