Heavy haul train cooperative control method and device, computer device and storage medium

Abstract

The application relates to a heavy-haul train cooperative control method and device, computer equipment and a storage medium. The method first acquires a real-time physical connection state between carriages, updates a time-varying topology matrix and reconstructs a multi-particle model, and dynamically adjusts a coupling force action logic. Then, a communication delay between trains is acquired, a predicted compensation is made on train state information in combination with the reconstructed model, and a current time state prediction value is output. Track environment parameters are sensed by an on-board sensor, a train running resistance model parameter is dynamically updated, and a dynamic driving safety constraint is generated. Finally, according to the time-varying topology matrix, the state prediction value, the updated resistance model and the dynamic driving safety constraint, a future train running state is predicted, potential running conflicts are detected by exchanging control intentions, weight parameters of an optimization objective function are adjusted when a conflict occurs, and optimal control instructions are solved and executed under the premise of meeting various constraints. The method can improve running efficiency and safety.

Description

Technical Field

[0001] This application relates to the field of railway train control technology, and in particular to a method, device, computer equipment, computer-readable storage medium, and computer program product for coordinated control of heavy-haul trains. Background Technology

[0002] Heavy-haul rail transport plays a crucial role in long-distance bulk freight transportation. To meet the growing transportation demand and improve line capacity, the rail transport industry is actively exploring more efficient train operation organization models.

[0003] Virtual train formation technology has emerged to address this need. Through car-to-car communication and cooperative control, it enables multiple trains to track each other with minimal safety intervals, as if they were physically linked together, thereby significantly improving transportation efficiency without altering the physical railway line. The core of this technology lies in establishing accurate train dynamics models and control strategies. Current mainstream methods are based on multi-mass models and train formation models built upon the fixed formation assumption, treating the coupling relationships between trains as constant.

[0004] However, in actual heavy-haul train operation scenarios, trains may need to detach some carriages and reassemble them at marshalling yards due to local faults, or dynamically adjust their formation to cope with environmental factors such as temporary speed limits and changes in track gradient. Mainstream control technologies cannot account for these formation changes, leading to a severe mismatch between control strategies and the actual physical state of the train, resulting in reduced operational efficiency and safety. Summary of the Invention

[0005] Therefore, it is necessary to provide a method, device, computer equipment, computer-readable storage medium, and computer program product for the cooperative control of heavy-haul trains that can improve operational efficiency in response to the above-mentioned technical problems.

[0006] Firstly, this application provides a cooperative control method for heavy-haul trains, including:

[0007] The system acquires the physical connection status between carriages in each train formation in real time in the heavy-haul train system, and updates the time-varying topology matrix that characterizes the real-time connection relationship between carriages in each train formation based on the changes in the acquired physical connection status; and reconstructs the multi-mass model describing the longitudinal dynamics of the train formation according to the time-varying topology matrix, so as to dynamically adjust the action logic of the coupler force between adjacent carriages in the multi-mass model.

[0008] The communication delay between each train group at the current time is obtained, and based on the reconstructed multi-mass model and the communication delay, the received train status information of each train group is predicted and compensated, and the compensated current time status prediction value is output.

[0009] The vehicle uses onboard sensors to perceive track environment parameters in real time, and based on the perceived track environment parameters, dynamically updates the corresponding parameters in the train running resistance model and generates dynamic driving safety constraints.

[0010] Based on the time-varying topology matrix, the current state prediction value, the updated train running resistance model, and the dynamic driving safety constraints, the operating states of each train formation in the future time domain are predicted, enabling each train formation to obtain control intentions from the predicted operating states and exchange control intentions with each other. This detects potential operating conflicts between train formations. If a potential operating conflict is detected, the weight parameters corresponding to the safety distance and speed tracking terms in the preset optimization objective function are dynamically adjusted. Under the premise of satisfying the dynamic constraints determined by the variable topology matrix, the dynamic driving safety constraints, and the preset actuator physical constraints, the optimal control commands for each train are solved with the goal of minimizing the dynamically adjusted objective function and avoiding operating conflicts. Each optimal control command is output to the execution control system of each train to achieve safe and coordinated operation of each train formation.

[0011] In one embodiment, the predicted operating status of each train group in the future time domain includes the predicted position and predicted speed of each train in the future time domain.

[0012] The method for detecting potential operational conflicts between different train sets includes:

[0013] The control system obtains the control intention for each train group in the future time domain based on the predicted position and speed of each train in the future time domain, and exchanges the control intentions with each other.

[0014] After the trains in each formation exchange control intentions, the trains in each formation are controlled to perform collaborative simulation prediction to obtain the predicted relative distance of each train in the future time domain.

[0015] The predicted relative distance is compared with a safe distance threshold determined jointly by the time-varying topology matrix and the dynamic driving safety constraints. If the predicted relative distance is less than the safe distance threshold, a potential operational conflict is determined.

[0016] In one embodiment, the step of dynamically adjusting the weight parameters corresponding to the safety distance term and the speed tracking term in the preset optimization objective function when a potential operational conflict is detected includes:

[0017] If a potential operational conflict is detected, the weight coefficient of the safety distance term in the optimization objective function is increased, and the weight coefficient of the speed tracking term in the optimization objective function is decreased.

[0018] In one embodiment, the physical connection state includes a coupler locking state; the steps for constructing the time-varying topology matrix include:

[0019] Sensors deployed in each carriage are used to sense the relative position of adjacent carriages and the locking status of the couplers between adjacent carriages, and the communication unit is used to sense the connection status of the communication link.

[0020] A time-varying topology matrix is ​​constructed using graph neural networks and digital twin technology; wherein, the nodes in the time-varying topology matrix represent carriages, and the edges represent physical connections of buffers, communication links, and preset safety distance constraints.

[0021] In one embodiment, the step of obtaining the communication delay between each train group at the current time, and based on the reconstructed multi-mass model and the communication delay, performing prediction compensation on the received train state information of each train group, and outputting the compensated current time state prediction value, includes:

[0022] Data is transmitted using a multipath redundancy communication protocol, and the communication delay of each communication link between each train group is obtained in real time by parsing the timestamp in the data packet.

[0023] A Kalman filter based on the delay differential equation is used as the state predictor. Starting from the received train state information, and combined with the reconstructed multi-mass model, the communication delay time length is predicted forward to predict the state of each train group at the current moment, which is used as the compensated current moment state prediction value.

[0024] In one embodiment, the step of sensing track environment parameters in real time through onboard sensors, dynamically updating corresponding parameters in the train running resistance model based on the sensed track environment parameters, and generating dynamic driving safety constraints includes:

[0025] By integrating inertial measurement units, lidar, and meteorological sensors, the track's slope, curvature, unevenness, and ambient wind speed data are collected in real time.

[0026] Machine learning algorithms are used to fuse and analyze the collected data on slope, curvature, unevenness, and environmental wind speed, and to calculate and update the track irregularity index and air resistance coefficient in the train running resistance model in real time.

[0027] Based on the real-time perceived curvature of the track, a lateral acceleration safety constraint is generated when the train passes through a curve, which serves as the dynamic driving safety constraint.

[0028] Secondly, this application also provides a heavy-haul train cooperative control device, comprising:

[0029] The dynamic topology sensing and modeling module is used to acquire the physical connection status between carriages in each train formation in the heavy-haul train system in real time, and update the time-varying topology matrix that represents the real-time connection relationship between carriages in each train formation based on the changes in the acquired physical connection status; and reconstruct the multi-mass model describing the longitudinal dynamics of the train formation according to the time-varying topology matrix, so as to dynamically adjust the action logic of the coupler force between adjacent carriages in the multi-mass model.

[0030] The communication delay compensation and robust predictive control module is used to obtain the communication delay between each train group at the current time, and based on the reconstructed multi-mass model and the communication delay, to predict and compensate the received train status information of each train group, and output the compensated current time status prediction value.

[0031] The environmental perception and dynamic parameter update module is used to perceive track environmental parameters in real time through on-board sensors, and based on the perceived track environmental parameters, dynamically update the corresponding parameters in the train running resistance model and generate dynamic driving safety constraints.

[0032] The conflict resolution and adaptive optimization module is used to predict the operating state of each train formation in the future time domain based on the time-varying topology matrix, the current state prediction value, the updated train running resistance model, and the dynamic traffic safety constraints. This allows each train formation to obtain control intentions from the predicted operating states and exchange control intentions with each other, thereby detecting potential operating conflicts between the train formations. If potential operating conflicts are detected, the module dynamically adjusts the weight parameters corresponding to the safety distance term and speed tracking term in the preset optimization objective function. Under the premise of satisfying the dynamic constraints determined by the variable topology matrix, the dynamic traffic safety constraints, and the preset actuator physical constraints, the module solves for the optimal control commands for each train, aiming to minimize the dynamically adjusted optimization objective function and avoid operating conflicts. The module outputs each of the optimal control commands to the execution control system of each train to achieve safe and coordinated operation of each train formation.

[0033] Thirdly, this application also provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps of the method described in the first aspect.

[0034] Fourthly, this application also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the method described in the first aspect.

[0035] Fifthly, this application also provides a computer program product, including a computer program that, when executed by a processor, implements the steps of the method described in the first aspect.

[0036] The aforementioned heavy-haul train cooperative control method, device, computer equipment, computer-readable storage medium, and computer program products, by acquiring the physical connection status in real time and updating the time-varying topology matrix, combined with matrix reconstruction of a multi-mass model, enable the dynamic model to accurately match the dynamic disassembly / reassembly state of the train. This achieves dynamic adaptation of the heavy-haul train topology, solving the problem that traditional fixed topology models cannot adapt to changes in train formation, and avoiding the control performance degradation caused by dynamic model mismatch. By acquiring real-time communication delays and predictively compensating for train status information, the accuracy of status perception is improved, providing accurate basic data for subsequent conflict detection and control command solving, and reducing control deviations caused by communication delays. By sensing track environment parameters through onboard sensors and dynamically updating the model and constraints, the train running resistance model and driving safety constraints are matched with the actual track environment in real time, allowing the control strategy to adapt to complex and changing track environments, thus improving the environmental adaptability of the control strategy. By predicting future operational states in the time domain and exchanging control intentions, potential operational conflicts can be identified in advance. Combining weight adjustment and constraint optimization to solve for the optimal control command, proactive conflict detection and resolution of heavy-haul trains is achieved, preventing derailments, collisions, and other safety accidents at the source and improving the safety of coordinated operation of heavy-haul trains. By dynamically adjusting the weights of the objective function, the efficiency objective of speed tracking is guaranteed when there are no conflicts, while the safety objective of safe distance is prioritized when there are conflicts. Simultaneously, the optimal control command is solved under all constraints, balancing the safety and transportation efficiency of coordinated operation of heavy-haul trains, achieving a dynamic balance between safety and efficiency. Attached Figure Description

[0037] To more clearly illustrate the technical solutions in the embodiments of this application or related technologies, the drawings used in the description of the embodiments of this application or related technologies will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0038] Figure 1 This is a flowchart illustrating a heavy-haul train cooperative control method in one embodiment;

[0039] Figure 2 This is a flowchart illustrating the steps for detecting potential runtime conflicts in one embodiment;

[0040] Figure 3 This is a schematic diagram of the composition of a virtual train formation cooperative control system based on a dynamic topology network in another embodiment;

[0041] Figure 4 This is a structural block diagram of a heavy-haul train cooperative control device in one embodiment;

[0042] Figure 5 This is an internal structural diagram of a computer device in one embodiment. Detailed Implementation

[0043] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0044] Against the backdrop of ever-increasing demand for heavy-haul railway transportation, complex and ever-changing line conditions, and continuously rising requirements for transportation efficiency and safety, this project aims to achieve efficient collaborative control of multiple heavy-haul trains in a virtual formation mode by constructing a dynamic topology network structure that can be adjusted in real time according to the operating environment. By optimizing train operation organization, it promotes the efficient allocation and utilization of railway transportation resources, helps build a green, low-carbon, safe, and efficient modern comprehensive transportation system, and ultimately achieves the organic unity of economic, social, and environmental benefits. This has irreplaceable value in promoting the sustainable development of the heavy-haul railway transportation industry.

[0045] Traditional technologies establish multi-mass models and train platoon models based on fixed topology assumptions. However, heavy-haul trains may dynamically disassemble or reassemble in actual operation due to faults, scheduling requirements, or environmental changes (such as temporary speed-limited sections). This leads to the reconstruction of inter-car connections, buffer operation logic, and communication links. Existing models do not explicitly consider the impact of topological changes on dynamic equations and control variables, which can cause control strategies to be unable to adapt to changes in train formation, potentially resulting in dynamic model mismatch, decreased control performance, or even derailment or collisions.

[0046] To adapt the control strategy to changes in grouping and thus improve operational efficiency, in an exemplary embodiment, such as Figure 1 As shown, a cooperative control method for heavy-haul trains is provided. This method is executed by computer equipment, specifically by a terminal or server alone, or by both a terminal and a server. In this embodiment, the method includes the following steps:

[0047] Step S102: Real-time acquisition of the physical connection status between carriages in each train group in the heavy-haul train system; and based on the changes in the acquired physical connection status, updating the time-varying topology matrix used to characterize the real-time connection relationship between carriages in each train group; and reconstructing the multi-mass model describing the longitudinal dynamics of the train group according to the time-varying topology matrix, so as to dynamically adjust the action logic of the coupler force between adjacent carriages in the multi-mass model.

[0048] Among them, the heavy-haul train system: from the perspective of application field, it can refer to the train operation system composed of multiple heavy-haul trains and applied to heavy-haul railway transportation scenarios, which can realize the operation of single train multi-car formation or multiple train virtual formation; from the perspective of structure, it can refer to the core components including the train body, on-board sensing equipment, communication equipment, control execution system and collaborative control backend, which is the application carrier of this method.

[0049] Among them, a train formation can refer to a train unit formed by combining multiple carriages or multiple trains according to scheduling requirements to adapt to heavy-haul transportation needs. It can be dynamically decomposed / reassembled according to faults and scheduling needs. It is the basic unit of collaborative control, and each train formation must follow a unified collaborative control strategy.

[0050] Among them, the physical connection status refers to the actual mechanical connection characteristics between the carriages in the train, with the core being the locking / unlocking status of the couplers, and also including the connection status of the buffers; it is the core basis for judging the train topology and directly determines the force transmission relationship between the carriages.

[0051] Among them, the time-varying topology matrix can refer to a matrix-type mathematical model that changes dynamically with time. The matrix elements represent the connection relationship between carriages at different times, such as 1 for connection and 0 for disconnection. It is a digital expression of the dynamic topology structure of heavy-haul trains and serves as the core data basis for dynamic model reconstruction and coordinated control.

[0052] The multi-mass model refers to a dynamic model that treats each carriage as an independent point mass, describing the longitudinal motion of the train. It characterizes the forces and motion relationships along the train's longitudinal direction and can include parameters such as carriage mass, acceleration, traction, resistance, and coupler force. Longitudinal dynamics refers to the branch of dynamics that studies the forces acting on the train along the track and the changes in its motion state. It can include the interaction of longitudinal forces such as traction, braking, air resistance, gradient resistance, and coupler force, as well as the changing patterns of motion parameters such as train speed, acceleration, and displacement. The logic of coupler force refers to the rules governing force transmission between carriages via couplers. When connected, the coupler force interacts between adjacent carriages; when disconnected, the coupler force disappears. It is the calculation rule for the coupler force term in the multi-mass model, determining the coupling relationship of forces in the model.

[0053] The overall execution entity of this method can be the heavy-haul train collaborative control center (which can be deployed as a cloud center or a distributed edge computing node), and the execution entities of each sub-step can include the collaborative control center and the on-board controllers of each train formation.

[0054] For example, onboard connectivity sensing sensors collect the physical connection status between carriages of each train formation in real time and upload it to the collaborative control center of the heavy-haul train system. The collaborative control center monitors changes in the physical connection status in real time. When topological changes such as decoupling / reassembly are detected, the time-varying topology matrix representing the real-time connection relationship between carriages is updated. Based on the time-varying topology matrix, the multi-mass model is reconstructed, and the logic of the coupler force between adjacent carriages in the model is dynamically adjusted to match the model with the current topology. For example, the time-varying topology matrix is ​​associated with the coupler force calculation term of the multi-mass model. When the topology matrix element is 0, the coupler force term between the corresponding carriages is removed from the model. When the element is 1, the coupler force term is retained and the force coupling relationship is maintained. The reconstructed multi-mass model is then sent to the onboard controller of each train formation in real time.

[0055] Step S104: Obtain the communication delay between each train group at the current time, and based on the reconstructed multi-mass model and the communication delay, perform prediction compensation on the received train status information of each train group, and output the compensated current time status prediction value.

[0056] Communication delay can refer to the time difference caused by network transmission and data processing when data is transmitted between trains or between a train and the roadside via a communication network; as a time-varying parameter, it changes dynamically due to the quality of the communication link, network load, and transmission distance.

[0057] Train status information refers to the core parameters that characterize the real-time operating status of a train, including position, speed, acceleration, traction, braking force, and carriage formation information; as dynamic data collected in real time, it is the basic input data for collaborative control.

[0058] Predictive compensation can refer to a technique that uses a dynamic model to predict the delayed state information in advance in response to the time delay of data transmission, in order to estimate the actual state at the current moment; it can be used to offset the impact of communication delay on control and improve the accuracy of state perception.

[0059] For example, the train communication unit collects communication link data between each train formation in real time, and the collaborative control center obtains the communication delay at the current moment based on the link data; the on-board controller of each train formation uploads the collected train status information to the collaborative control center, and the collaborative control center predicts and compensates for the lagging train status information based on the reconstructed multi-mass model and the communication delay, and outputs the compensated current moment status prediction value by extrapolating forward through the dynamic model.

[0060] Step S106: Real-time sensing of track environment parameters through onboard sensors, and dynamic updating of corresponding parameters in the train running resistance model based on the sensed track environment parameters, and generation of dynamic driving safety constraints.

[0061] Among them, on-board sensors refer to sensing devices installed at the train's front, carriages, couplers, etc.; for example, they can include environmental sensing sensors, state sensing sensors, connection sensing sensors, etc., serving as the carriers for collecting track environment and train status data. Track environment parameters refer to parameters characterizing the track line and its surrounding environment, such as track gradient, curvature, irregularity, ambient wind speed, and temperature; these are the core basis for calculating train running resistance and formulating driving safety constraints. The train running resistance model can be a mathematical model that calculates various resistances during train operation by specifying quantities; it can include sub-models such as gradient resistance, curve resistance, air resistance, and additional resistance from track irregularities, with core parameters such as the track irregularity index and air resistance coefficient. Dynamic driving safety constraints refer to rigid rules that are dynamically generated based on real-time track environment parameters to ensure train operation safety; for example, they can include lateral acceleration constraints, safety distance constraints, and speed constraints, which are adjusted in real time according to changes in the track environment.

[0062] For example, onboard sensors perceive track environment parameters in real time and transmit them to the vehicle's onboard controller. Based on the perceived track environment parameters, the onboard controller dynamically updates the core parameters in the train running resistance model, and simultaneously generates dynamic driving safety constraints in combination with track environment characteristics. For instance, by substituting track environment parameters into the sub-terms of the train running resistance model, core parameters such as track irregularity index and air resistance coefficient are updated to match the calculation results of the resistance model with the actual environment. Based on the curvature, gradient, and other characteristics of the track line, safety constraints adapted to the current environment are generated, and the resistance model and safety constraints are synchronized to the collaborative control center.

[0063] Step S108: Based on the time-varying topology matrix, the current state prediction value, the updated train running resistance model, and dynamic traffic safety constraints, predict the running state of each train formation in the future time domain. This allows each train formation to obtain control intentions from the predicted running states and exchange control intentions with each other. Based on this, detect whether there are potential running conflicts between the train formations. If potential running conflicts are detected, dynamically adjust the weight parameters corresponding to the safety distance term and speed tracking term in the preset optimization objective function. Under the premise of satisfying the dynamic constraints, dynamic traffic safety constraints, and preset actuator physical constraints determined by the variable topology matrix, solve for the optimal control command of each train with the goal of minimizing the optimization objective function after dynamic weight adjustment and avoiding running conflicts. Output each optimal control command to the execution control system of each train to achieve safe and coordinated operation of each train formation.

[0064] The future time domain can refer to a time interval predicted forward based on the current moment (rolling time domain); it can also refer to the time range for predicting train operation status and formulating control strategies in collaborative control. The length of the time domain can be set according to actual control requirements.

[0065] The operating status can refer to the train's position, speed, acceleration, traction force, and other motion and force parameters in the future time domain; it can be derived from the current status, dynamic model, and track environment parameters.

[0066] Control intent refers to the control strategy plan formulated by each train group in the future time domain based on its predicted operating state. This can include, for example, acceleration sequences, traction / braking force adjustment plans, and gear shifting plans. It is the core information exchanged between trains and can be used for conflict detection. Potential operational conflicts refer to safety risks that may occur to trains in the future time domain. These can be distance conflicts (relative distance less than a safety threshold) or speed conflicts (rear-end collisions / grabs caused by speed mismatch). These conflicts may arise from inconsistent train control intent, changes in track environment constraints, or adjustments to the train topology.

[0067] The optimization objective function can be a mathematical function used to measure the effectiveness of coordinated control, and may include multiple sub-items corresponding to control objectives. For example, it may include safe distance tracking, speed tracking, and energy consumption optimization, with each sub-item having its priority adjusted by weight parameters. The safe distance / speed tracking term can be the core sub-item of the optimization objective function, where the safe distance term represents the deviation between the actual relative distance of the train and the ideal safe distance, and the speed tracking term represents the deviation between the actual speed of the train and the reference speed; these are core indicators for quantifying safety and efficiency control objectives. Dynamic constraints can be constraints based on train dynamics principles, determined by a time-varying topology matrix; they may include force coupling constraints between carriages, train acceleration / deceleration constraints, coupler force bearing constraints, etc., dynamically matched with the train topology. Actuator physical constraints can refer to the physical performance limitations of train traction, braking, and other actuators; specifically, they may include maximum traction force, maximum braking force, gear adjustment rate constraints, etc., which are hard physical conditions that control commands must satisfy. The optimal control command can be the control command obtained by minimizing the optimization objective function under all constraints; it may include the train's optimal traction / braking force, optimal gear, acceleration / deceleration commands, etc.

[0068] The execution control system can refer to a system that receives optimal control commands and drives the train's traction, braking, gear adjustment, and other actuators. Safe and coordinated operation refers to the coordinated matching of speed and position among train formations while meeting all safety constraints; it possesses the characteristics of dynamically adapting to topology and environmental changes, avoiding operational conflicts, and balancing efficiency and safety.

[0069] For example, the collaborative control center receives a time-varying topology matrix, the current state prediction, the updated train running resistance model, and dynamic traffic safety constraints. Based on this data, it predicts the running state of each train formation in the future time domain through dynamic simulation. Each train's onboard controller formulates its own control intent based on the predicted running state and exchanges these intents with each other via vehicle-to-vehicle communication. The onboard controller then uploads the exchanged control intents to the collaborative control center. The collaborative control center combines all control intents to detect potential running conflicts between train formations. If a potential conflict is detected, the collaborative control center dynamically adjusts the weight parameters of the safety distance and speed tracking terms in the objective function. Under the premise of satisfying dynamic constraints, dynamic traffic safety constraints, and actuator physical constraints, the collaborative control center aims to minimize the adjusted objective function and avoid running conflicts by solving for the optimal control commands for each train using a mathematical programming algorithm. For example, a mathematical optimization problem containing all constraints can be constructed, and a mixed-integer linear programming algorithm can be used to solve the objective function to obtain the optimal control command sequence for each train formation in the future time domain. Finally, the collaborative control center issues the optimal control commands to the execution control systems of each train formation.

[0070] In the aforementioned heavy-haul train cooperative control method, by acquiring the physical connection status in real time and updating the time-varying topology matrix, combined with matrix reconstruction of the multi-mass model, the dynamic model can accurately match the dynamic disassembly / reassembly state of the train, realizing dynamic adaptation of the heavy-haul train topology structure. This solves the problem that traditional fixed topology models cannot adapt to changes in train formation and avoids the control performance degradation caused by dynamic model mismatch. By acquiring real-time communication delays and predictively compensating for train status information, the accuracy of status perception is improved, providing accurate basic data for subsequent conflict detection and control command solving, and reducing control deviations caused by communication delays. By sensing track environment parameters through onboard sensors and dynamically updating the model and constraints, the train running resistance model and driving safety constraints are matched with the actual track environment in real time, allowing the control strategy to adapt to complex and ever-changing track environments and improving the environmental adaptability of the control strategy. By predicting future time-domain operating states and exchanging control intentions, potential operating conflicts can be identified in advance. Combined with weight adjustment and constraint optimization to solve for the optimal control command, active conflict detection and resolution of heavy-haul trains are realized, preventing safety accidents such as derailment and collisions from the source and improving the safety of cooperative operation of heavy-haul trains. By dynamically adjusting the weights of the objective function, the efficiency objective of speed tracking is guaranteed when there is no conflict, while the safety objective of safe distance is prioritized when there is a conflict. At the same time, the optimal control command is solved under all constraints, taking into account both the safety and transportation efficiency of the coordinated operation of heavy-haul trains, and achieving a dynamic balance between safety and efficiency.

[0071] In an exemplary embodiment, the predicted operating status of each train formation in the future time domain includes the predicted position and predicted speed of each train in the future time domain; such as Figure 2 As shown, detecting potential operational conflicts between different train formations can specifically include the following steps S202 to S206. Wherein:

[0072] Step S202: Control each train group to obtain the control intention in the future time domain based on the predicted position and predicted speed of each train in the future time domain, and exchange control intentions with each other.

[0073] Among them, the predicted position / predicted speed can be the position and speed parameters of the train at different times in the future time domain, which are predicted based on the current state, dynamic model, and track environment parameters. These are dynamically changing time series data. They can be the core indicators for predicting the train's operating status and detecting operational conflicts, and can directly determine the calculation results of the relative distance.

[0074] Step S204: After the trains in each train formation exchange control intentions, control each train formation to perform collaborative simulation prediction to obtain the predicted relative distance of each train formation in the future time domain.

[0075] Collaborative simulation prediction refers to the dynamic simulation jointly performed by each train group based on mutually exchanged control intentions. This aims to accurately predict the relative motion states of each train in the future time domain, avoiding the limitations of single-train simulation. Predicted relative distance refers to the real-time distance difference between trains calculated based on their predicted positions. This can be a parameter that dynamically changes over time in the future time domain and serves as a direct basis for determining whether distance conflicts exist between trains. The safe distance threshold refers to the minimum relative distance that ensures safe train operation, determined by a combination of time-varying topology matrices (train group relationships) and dynamic traffic safety constraints (track environment, speed, etc.). This can be a dynamic threshold that adjusts in real-time according to changes in train topology and track environment, serving as a criterion for detecting train operation conflicts.

[0076] Step S206: The predicted relative distance is compared with a safe distance threshold determined by a time-varying topology matrix and dynamic driving safety constraints. If the predicted relative distance is less than the safe distance threshold, a potential operational conflict is identified.

[0077] For example, the collaborative control center sends the predicted positions and speeds of each train in the future time domain to the onboard controllers of each train formation. Each onboard controller, based on the received predicted positions and speeds and its updated train running resistance model, formulates its control intentions for the future time domain and exchanges these intentions with other trains in the formation via the vehicle-to-vehicle communication network. After receiving the control intentions from all other trains, each onboard controller performs collaborative simulation prediction based on the time-varying topology matrix and dynamic traffic safety constraints to simulate the joint operation state of all trains in the future time domain and calculate the... The predicted relative distance of train formations at different times in the future time domain; the collaborative control center determines the safe distance threshold in the current scenario based on the current time-varying topology matrix (to determine the train formation relationship, and to set stricter distance thresholds for trains in the same formation) and dynamic driving safety constraints (such as increasing the distance threshold for curves and slopes) and sends it to each on-board controller; each on-board controller compares the calculated predicted relative distance with the safe distance threshold in real time. If the predicted relative distance is less than the safe distance threshold at any time in the future time domain, it immediately determines that there is a potential operational conflict and uploads the conflict information to the collaborative control center.

[0078] In this embodiment, by using predicted position and speed as the basis for control intent formulation, and combining this with co-simulation prediction to obtain predicted relative distance, and using the quantitative comparison of relative distance and safe distance threshold as the conflict judgment criterion, precise and quantitative detection of operational conflicts is achieved, improving the accuracy of conflict detection. By requiring each train formation to exchange control intents and perform co-simulation prediction, the operation plans of all trains are integrated, improving the comprehensiveness of conflict detection and avoiding missed conflicts caused by a single train predicting based solely on its own state. This approach can identify complex operational conflicts between multiple trains. By using time-varying topology matrices and dynamic traffic safety constraints together as the basis for determining the safe distance threshold, the threshold can be adapted to the dynamic formation state of trains and complex track environments, achieving dynamic matching of the safe distance threshold. This avoids the problem of fixed thresholds being too lenient or too strict in different scenarios, improving the rationality of conflict detection. Conflict detection based on future time-domain prediction data can identify potential risks before a conflict actually occurs, reserving sufficient time for subsequent weight adjustments and control command optimization, achieving early warning of conflicts, and improving the effectiveness of conflict resolution.

[0079] In an exemplary embodiment, when a potential operational conflict is detected, the weight parameters corresponding to the safety distance term and the speed tracking term in the preset optimization objective function are dynamically adjusted, including: when a potential operational conflict is detected, increasing the weight coefficient of the safety distance term in the optimization objective function and decreasing the weight coefficient of the speed tracking term in the optimization objective function.

[0080] The weighting coefficient can refer to the proportional coefficient of each sub-item in the objective function, which is used to adjust the importance of each sub-item in the objective function; the magnitude of the weighting coefficient can represent the priority of the corresponding control objective, and the larger the coefficient, the more important the corresponding control objective; the weighting coefficient can be dynamically adjusted according to the operating scenario.

[0081] For example, the onboard controller of each train group uploads the detected potential operational conflict information (such as conflict type, future time of conflict occurrence, and conflicting train number) to the collaborative control center. Upon receiving the potential conflict information, the collaborative control center initiates a dynamic adjustment mechanism for weight parameters, increasing the weight coefficient corresponding to the safety distance term in the optimization objective function while decreasing the weight coefficient corresponding to the speed tracking term. The collaborative control center then substitutes the adjusted weight coefficients into the optimization objective function, updating the function and preparing for the subsequent solution of the optimal control command.

[0082] In this embodiment, by increasing the weight coefficient of the safety distance term and decreasing the weight coefficient of the speed tracking term, the safety objective of train operation is placed above the efficiency objective when there is a potential conflict. This ensures that the solution of control commands is centered on resolving conflicts and ensuring safety, thereby guaranteeing train safety from the objective level and realizing the dynamic switching of control objective priority.

[0083] In an exemplary embodiment, the physical connection state includes the coupler locking state; the construction steps of the time-varying topology matrix include: sensing the relative position of adjacent cars and the coupler locking state between adjacent cars through sensors deployed in each car, and sensing the connection state of the communication link using a communication unit; constructing the time-varying topology matrix using graph neural network and digital twin technology; wherein, the nodes in the time-varying topology matrix represent cars, and the edges represent the physical connection of the buffer, the connection of the communication link, and the preset safety distance constraints.

[0084] The coupler locking status refers to the locking / unlocking state of the coupler. Locking indicates a valid physical connection between the carriages, while unlocking indicates a connection failure. This is a mechanical characteristic that can be detected by sensors and serves as the core indicator of the physical connection status. The communication unit refers to the equipment that enables data transmission between trains and between the train and the roadside. It may include wireless communication modules, antennas, data processing units, etc., and may support various communication methods such as 5G, UWB, and direct train-to-train connections. It serves as a carrier for sensing the connection status of communication links and transmitting various types of data.

[0085] Graph neural networks (GNNs) can refer to deep learning algorithms for graph-structured data, capable of feature learning and relationship modeling of nodes and edges in a graph. They can be used to model the topology of trains, mining the connections and features between carriages. They possess the ability to process dynamic graph-structured data, adapting to the time-varying characteristics of train topologies. Digital twin technology refers to constructing virtual mirrors of physical entities through digital means, achieving real-time mapping between physical entities and virtual mirrors. It can be used to construct virtual topology models of heavy-haul trains, reflecting the physical connection status and communication link status of the train in real time. It can provide visualization and modeling support for the construction of time-varying topology matrices. Buffer physical connection refers to the connection status of mechanical connection devices between carriages used to buffer force impacts. Buffers work with couplers to achieve force transmission and buffering between carriages, and their physical connection status is synchronized with the coupler locking status; it is one of the core representations of "edges" in a graph structure.

[0086] For example, visual sensors and lidar are deployed at key locations such as the couplers in each carriage. These sensors collect real-time data on the relative positions of adjacent carriages and the coupler locking status between them, uploading the data to the collaborative control center. Communication units deployed on the train and roadside monitor the signal quality and data transmission status of the communication links in real time, sensing the connection status of communication links between carriages and trains, and uploading the link status data to the collaborative control center. After receiving the physical connection status and communication link status data, the collaborative control center uses graph neural networks and digital twin technology to construct a topology model of the heavy-haul train: each carriage is used as a node in the topology model, and the physical connections of buffers, communication link connections, and preset safety distance constraints are used as edges. Based on the constructed topology model, the collaborative control center generates and updates a time-varying topology matrix in real time. For example, elements in the matrix... Represents the connection status between the i-th and j-th carriages at time t, when the connection is valid. =1, when the connection fails (coupler disconnected / communication interrupted). =0.

[0087] In this embodiment, by deploying dedicated sensors at key locations of the coupler, the locking status of the coupler and the relative positions of adjacent cars are directly sensed, achieving precise perception of the physical connection status. Compared with indirect detection methods, this improves the accuracy and real-time performance of physical connection status detection. Combining graph neural networks and digital twin technology, the train topology is modeled as graph-structured data, enabling intelligent construction of a time-varying topology matrix. This accurately uncovers the connection relationships and dynamic changes between cars, making the time-varying topology matrix more closely reflect the actual topology of the train and improving the modeling accuracy. Incorporating buffer physical connections, communication link connections, and safety distance constraints into the construction of the topology matrix enriches its representation, allowing the matrix to represent not only mechanical connections but also communication connections and safety constraints, providing a more comprehensive basis for subsequent dynamic model reconstruction and conflict detection.

[0088] In an exemplary embodiment, the communication delay between each train group at the current moment is obtained, and based on the reconstructed multi-mass model and the communication delay, the received train state information of each train group is predicted and compensated, and the compensated current moment state prediction value is output. This includes: transmitting data using a multi-path redundancy communication protocol, and obtaining the communication delay of each communication link between each train group in real time by parsing the timestamp in the data packet; using a Kalman filter based on the delay differential equation as a state predictor, starting from the received train state information, and combining the reconstructed multi-mass model, predicting the time length of the communication delay in advance, and predicting the state of each train group at the current moment as the compensated current moment state prediction value.

[0089] Among them, multipath redundancy communication protocols refer to communication protocols that transmit data simultaneously across multiple communication links, achieving data transmission reliability through link redundancy; they can select the link with the best quality for data transmission in real time and have automatic switching capability in case of link failure; they can be used for critical data transmission between trains to improve communication stability. Timestamps are time stamps added to each data packet, representing the sending / receiving time of the data packet; they serve as the core basis for calculating communication delay, and the communication delay of a single data packet can be obtained by the difference between the sending and receiving timestamps. Delay differential equations are differential equations containing unknown functions and their delay derivatives, capable of describing dynamic systems with time delay characteristics; they can be used to construct train dynamics models considering communication delay, accurately characterizing the impact of communication delay on train state; they can quantitatively describe the dynamic characteristics of time-varying delays. Kalman filters are optimal estimation algorithms that achieve optimal estimation of the state of a dynamic system through state equations and observation equations, combined with prediction and update steps; as state predictors, they can be used for predictive compensation of train state. A state predictor can be a device / algorithm that estimates the current or future state of a system based on mathematical models and observation data; it can be used to offset the effects of communication delays and estimate the actual state of a train at the current moment.

[0090] For example, each train formation uses a multi-path redundancy communication protocol for data transmission between trains and between trains and the collaborative control center. A high-precision timestamp is added to each transmitted data packet to record the transmission time. After receiving the data packet, the communication unit at the data receiving end records the reception time. By parsing the timestamp in the data packet, the difference between the transmission time and the reception time is calculated to obtain the communication delay of a single link. The communication unit collects the communication delay of each link in real time and uploads it to the collaborative control center. The collaborative control center performs fusion processing on the communication delay of each link to obtain the comprehensive communication delay between each train formation at the current time. The collaborative control center uses a Kalman filter based on the delay differential equation as the state predictor, transforming the reconstructed multi-mass model into a delay differential equation that includes the communication delay, which serves as the state equation of the Kalman filter. The collaborative control center uses the received lagging train state information as the observation value of the Kalman filter. With the comprehensive communication delay as the time offset, starting from the acquisition time of the lagging state information, and combining the forward prediction time length of the delay differential equation, the optimal state estimate of each train formation at the current time is obtained through the prediction-update step of the Kalman filter, which serves as the compensated state prediction value for the current time.

[0091] In this embodiment, by employing a multi-path redundancy communication protocol and adding timestamps to data packets, the communication delay of each link can be directly calculated, achieving real-time and accurate acquisition of communication delay. Compared with indirect estimation methods, this improves the accuracy and real-time performance of communication delay detection, while multi-path redundancy ensures communication stability. A Kalman filter based on the delay differential equation is used as the state predictor. This not only accurately characterizes the impact of communication delay on train dynamics through the delay differential equation, but also offsets the effects of sensor noise and communication interference through the optimal estimation capability of the Kalman filter. This makes the state prediction value more closely match the actual current state of the train, improving the accuracy of state prediction compensation. The delay differential equation can quantitatively describe the characteristics of time-varying communication delay, and the Kalman filter can perform prediction compensation in real time based on the latest communication delay and state information. This allows the prediction compensation process to adapt to the dynamic changes in communication delay, improving the adaptability of the solution.

[0092] In one exemplary embodiment, onboard sensors are used to perceive track environment parameters in real time. Based on the perceived track environment parameters, the corresponding parameters in the train running resistance model are dynamically updated, and dynamic driving safety constraints are generated. This includes: collecting track gradient, curvature, irregularity, and ambient wind speed data in real time by integrating an inertial measurement unit, lidar, and meteorological sensors; using machine learning algorithms to fuse and analyze the collected gradient, curvature, irregularity, and ambient wind speed data, and calculating and updating the track irregularity index and air resistance coefficient in the train running resistance model in real time; and generating lateral acceleration safety constraints for the train when cornering based on the real-time perceived track curvature, which serve as dynamic driving safety constraints.

[0093] In this context, an Inertial Measurement Unit (IMU) can refer to a combination of sensors capable of measuring the angular velocity and acceleration of an object, including gyroscopes, accelerometers, etc.; used for sensing the train's attitude, motion state, and track gradient and curvature. A lidar (Light Detection and Ranging) device can refer to a device that senses the distance, shape, and position of a target by emitting a laser beam and receiving reflected signals; used for high-precision sensing of track irregularities, curvature, and the surrounding environment. A meteorological sensor can be a sensor used to collect environmental meteorological parameters, including wind speed sensors, temperature sensors, and barometric pressure sensors; used to collect environmental wind speed data to provide a basis for calculating air resistance. A machine learning algorithm can be an algorithm that builds models through data learning and achieves data prediction and analysis; in this embodiment, it is used to fuse and analyze multi-source track environmental data to calculate the track irregularity index and air resistance coefficient; it can uncover complex correlations between multi-source data and improve the accuracy of parameter calculations. The track irregularity index can be a comprehensive indicator that quantitatively characterizes the degree of unevenness of the track surface in terms of elevation and lateral movement; it is a core calculation parameter for the additional resistance caused by track irregularities in the train running resistance model; as a parameter that dynamically changes with the track position, it needs to be updated in real time. Lateral acceleration safety constraints refer to the maximum acceleration limits set to prevent trains from overturning or derailing due to excessive lateral acceleration when cornering. These limits are determined by the curve radius and train speed; the smaller the curve radius, the smaller the allowable lateral acceleration. It is one of the core components of dynamic train safety constraints.

[0094] For example, an inertial measurement unit (IMU), a multi-line lidar, and a meteorological sensor can be integrated and deployed at the front of the train to form an integrated environmental perception system. This system collects track environment parameters in real time: the IMU and lidar collaboratively collect data on track gradient, curvature, and irregularity, while the meteorological sensor collects environmental wind speed data in the direction of train travel in real time. The onboard controller of each train set receives multi-source track environment data collected by the environmental perception system. After preprocessing the data (such as noise reduction and normalization), machine learning algorithms (such as gradient boosting trees and neural networks) are used to perform fusion analysis on the multi-source data to uncover the correlation features between the data. Based on the fusion analysis results, the onboard controller calculates and updates the track irregularity index and air resistance coefficient in the train running resistance model in real time, and substitutes the updated parameters into the calculation formula of the resistance model to complete the dynamic update of the train running resistance model. The onboard controller extracts the real-time perceived track curvature, sets the maximum value of lateral acceleration based on the lateral acceleration calculation formula and the train's safe operation requirements, and generates lateral acceleration safety constraints when the train passes through curves. These constraints serve as dynamic train operation safety constraints under the current track environment. At the same time, the updated resistance model and safety constraints are uploaded to the collaborative control center.

[0095] In this embodiment, an integrated environmental perception system is constructed by integrating an inertial measurement unit, lidar, and meteorological sensors. This system can simultaneously collect track geometric parameters and environmental meteorological parameters. Compared to a single sensor, it provides more comprehensive and accurate perception, achieving high-precision, multi-dimensional perception of track environmental parameters. Machine learning algorithms are used to fuse and analyze multi-source environmental data, uncovering complex correlations between data points and accurately calculating the track irregularity index and air resistance coefficient. This ensures that the parameters of the resistance model are highly matched to the actual environment, improving the accuracy of resistance calculations and providing a reliable mechanical basis for solving control commands. Lateral acceleration safety constraints are generated based on the real-time perceived track curve radius, enabling precise adaptation of safety constraints to the curve characteristics of the track. This achieves scenario-based generation of dynamic train safety constraints, preventing train overturning and derailment during curves from a mechanical perspective and improving the operational safety of trains when curves.

[0096] In an exemplary embodiment, based on the real-time perception of track environment parameters and updating of the train running resistance model by the on-board sensors, a self-correcting mechanism for the entire train running resistance model process—data fusion, benchmark calibration, error correction, and trend prediction—can be constructed by fusing multi-source heterogeneous data from the roadside and road network levels. This mechanism enables high-precision real-time updates and trend predictions of the resistance model parameters, making the resistance model more closely aligned with actual track conditions, weather, and operating conditions. This provides a more accurate mechanical basis for solving collaborative control commands. Specific steps may include:

[0097] Multi-source data acquisition and preprocessing: A multi-source data interaction channel is established between vehicles, roads, and networks. Based on the acquisition of track gradient, curvature, unevenness, and real-time wind speed by onboard integrated inertial measurement units, lidar, and meteorological sensors, three types of data are simultaneously accessed: First, high-precision full-cycle track inspection data from the track inspection vehicle, including baseline values ​​for track unevenness and curve radius calibration values ​​for fixed sections of the track; second, large-scale data on the resistance operation of similar heavy-load trains on the same line, including actual resistance values ​​under different operating conditions and the matching relationship between track parameters and air resistance coefficients; and third, regional weather forecast data from meteorological stations, including wind speed, wind direction, and temperature trends for the next 1-3 hours. All accessed data undergoes temporal and spatial alignment, noise reduction, and normalization preprocessing to form a standardized dataset.

[0098] Dual-benchmark calibration and error modeling: High-precision data from the track inspection vehicle is used as a static benchmark to calibrate static parameters such as track irregularity index and track curvature collected by onboard sensors in real time, eliminating measurement deviations caused by sensor hardware errors and local deformation of the track. The resistance operation big data of trains on the same line is used as a dynamic benchmark. Combined with the actual traction / braking force and speed data of the train, the actual operating resistance value is inferred and compared with the theoretical resistance value calculated by the onboard model. Through big data regression analysis, an error correction model for the core parameters of the resistance model (such as track irregularity index and air resistance coefficient) is established to clarify the parameter error rules under different track, speed and weather conditions.

[0099] Real-time parameter correction and model update: The pre-processed onboard real-time sensing data is input into the error correction model. Based on the current operating conditions, the error compensation values ​​of the track irregularity index and air resistance coefficient are automatically calculated to accurately correct the parameters calculated by the onboard sensors in real time. The corrected parameters are then immediately input into the train running resistance model to complete the dynamic update of the model, so that the total resistance value calculated by the model is highly matched with the resistance experienced by the train in actual operation.

[0100] Trend prediction and forward adjustment: Combining regional meteorological forecast data from meteorological stations, the characteristics of wind speed, wind direction, and temperature changes in the next 1-3 hours are extracted and input into the extension module of the error correction model to predict the trend of air drag coefficient changes at different time points in advance. Based on this trend prediction, the train running resistance model is updated in advance with forward parameters, providing accurate resistance trend basis for the collaborative control system to formulate control strategies in the future time domain in advance, making control commands more forward-looking and adaptable.

[0101] The above embodiments compensate for the perception limitations of a single onboard sensor by fusing multi-source data, enabling static benchmark calibration and dynamic error correction of resistance model parameters, and significantly improving the model calculation accuracy. At the same time, by combining weather forecasts, parameter trend prediction and forward-looking model adjustment are achieved, allowing the resistance model to not only adapt to the current operating conditions, but also provide forward-looking mechanical basis for train cooperative control, further improving the accuracy and foresight of heavy-haul train cooperative control commands.

[0102] In one exemplary embodiment, such as Figure 3 As shown, this application also provides a virtual formation heavy-haul train cooperative control system based on a dynamic topology network, including:

[0103] The dynamic topology sensing and modeling module is used to monitor the connection status changes of train formations in real time and output a dynamically changing topology matrix T(t). More specifically, lidar and vision sensors are deployed at key nodes (couplers) in each car to sense the relative positions of adjacent cars and the locking status of the couplers. Simultaneously, 5G / UWB communication units deployed on the train and roadside are used to sense the connection status of communication links. All sensing data is transmitted in real time to a central processing unit (or distributed edge computing node). The central processing unit (or distributed edge computing node) runs a digital twin model based on a graph neural network (GNN). In the GNN digital twin model, each car or the entire train is considered a node, while buffer connections, communication links, and virtual coupling distances reserved for safety are considered edges. Based on the real-time sensing data, a topology matrix is ​​dynamically updated. For example, when the i-th car is connected to the j-th car, =1; When the coupler disconnects or communication is interrupted. =0, when the module detects that the coupler is disconnected (e.g.) When the equation changes from 1 to 0, it immediately triggers a reconstruction of the dynamics model. For example, for the third car that becomes the last car in the new formation after being disconnected, its dynamics equation ( Topological matrix , of which elements Indicates whether the i-th and j-th carriages are connected at time t, and the buffer tension. (Only effective if the connection exists) From the original Automatically update to Because it is no longer subjected to the pulling force of the front buffer Fcw2. Updated topology matrix. The reconstructed model is sent to all subsequent modules in real time.

[0104] Communication delay compensation and robust predictive control module: for use based on topology matrix and real-time communication latency The system predicts and compensates for train status and outputs the predicted status value after delay compensation. More specifically, it adopts a multi-path redundant communication protocol that combines vehicle-to-vehicle direct communication based on 5G network slicing with roadside unit relay. By adding a timestamp to each data packet, it estimates the communication delay of each link in real time. The communication delay compensation and robust predictive control module uses delay differential equations (DDEs). , The system dynamics are described using time-varying communication delays (represented by a Kalman filter), and a Kalman filter is used as the state predictor. For example, the controller receives the train's delay at time t- at time t- The state at any given moment (e.g., velocity, position). The predictor uses the received lagged states and the known model to make forward predictions. The time frame is used to estimate the most probable state of the train at the current time t. This estimated most probable state is then used for subsequent optimal control. Simultaneously, a delay-sensitive term is incorporated into the objective function of Model Predictive Control (MPC). , ,in It is an estimated current speed. This takes into account the reference speed after the delay. This makes the optimization target more realistic, thus generating control commands that are more able to offset the effects of the delay.

[0105] Environmental perception and dynamic parameter update module: used to acquire orbital environmental parameters in real time and output updated orbital irregularity index R(s) and air resistance coefficient. More specifically, a multi-line lidar and inertial measurement unit (IMU) are deployed at the front of the vehicle to accurately sense the slope, curvature, and track irregularities (such as elevation changes) ahead of the track. Simultaneously, integrated weather sensors acquire real-time wind resistance data. Furthermore, periodic line data from a track inspection vehicle is received. Machine learning algorithms (gradient boosting tree) are used to fuse and analyze this multi-source data, calculating the track irregularity index R(s) and air resistance coefficient in real time. Subsequently, the train running resistance model was updated: ,in, The total resistance of the train; For slope resistance; For curvilinear resistance; For air resistance; Added resistance due to track irregularities. , Let A be the air density and A be the cross-sectional area of ​​the vehicle body. The updated and more accurate drag model is fed into the optimization solution module. At the same time, based on the real-time sensed curve radius r, a lateral acceleration constraint is generated. , This constraint is passed to the optimization module as a hard constraint to ensure that the train will not overturn due to excessive speed when cornering.

[0106] The conflict resolution and adaptive optimization solution module communicates with the dynamic topology sensing and modeling module, the communication delay compensation and robust predictive control module, and the environmental sensing and dynamic parameter update module for the following purposes:

[0107] Receive the topology matrix T(t), the delayed-compensated state prediction, and the updated orbital environment parameters and constraints;

[0108] Based on the received data, control intentions are exchanged with other trains in the trainset via car-to-car communication to detect potential operational conflicts. More specifically, each train locally calculates a preliminary set of control intentions (the expected acceleration sequence in the next few seconds) based on the information it receives. Then, all trains broadcast their preliminary intentions through the car-to-car communication network. After receiving the intentions of other trains, each train performs collaborative simulation prediction. It calculates the predicted relative distance between itself and the train ahead in the future and compares it with a safe distance threshold that comprehensively considers the current topology (such as whether they are in the same train group) and track conditions (such as needing to increase the distance on curves). If the predicted distance is less than the safe distance threshold, a potential conflict is determined.

[0109] When a potential conflict is detected, the weight parameters related to safe distance and speed tracking in the objective function are dynamically adjusted. More specifically, once a conflict is detected (for example, the following vehicle is predicted to accelerate while the preceding vehicle needs to decelerate due to curve constraints, resulting in the distance being too close), the weight adaptive adjustment mechanism will be activated to automatically increase the priority of the safe distance term (corresponding to weight K_s) in the objective function, while appropriately decreasing the priority of the speed tracking term (corresponding to weight K_v).

[0110] Based on the dynamically adjusted objective function, topology matrix T(t), and safety constraints composed of track environment parameters, the mixed-integer linear programming problem is solved to calculate the optimal control gear for each train. Under the adjusted objective function and all constraints (including dynamic constraints, environmental constraints, and topology constraints), the MILP (Mixed Integer Linear Programming) solver is recalculated to find a new optimal control gear sequence that can avoid conflicts.

[0111] The optimal control gear is output to the train execution control system to drive the train.

[0112] In this embodiment:

[0113] Imagine this scenario: A heavily loaded train decouples due to a malfunction, and the following train needs to accelerate to catch up and recouple. They are entering a curve at this point. The specific process is as follows:

[0114] Topology Adaptive Activation: The dynamic topology module detects a decoupling event, updates the train fleet model (from one long train to two independent short trains), and notifies the entire system.

[0115] Environment model update: The track environment module detected a curve, injecting greater curve drag and lateral acceleration constraints into the dynamics model.

[0116] Delay compensation and state prediction: The rear vehicle controller predicts the future state of the preceding and following vehicles based on a new topology and environment model and taking into account the current communication delay.

[0117] Conflict detection and resolution:

[0118] Conflict: The objective function of the following car requires it to accelerate to catch up, but the constraints of the curve environment require it not to go too fast, while the distance to the car in front is decreasing.

[0119] Resolution: The conflict resolution module works within a unified objective function. Safe distance item ( , (For safety distance) and curve constraints will strongly inhibit the intention of following vehicles to accelerate blindly. At the same time, This ensures that the decision is made based on the latest, delayed-compensated state information.

[0120] Result: The MPC solver calculates a compromise optimal strategy: the following car accelerates gently on the straight section and decelerates smoothly in advance before entering the curve, which ensures safety while striving towards the goal of recoupling.

[0121] In one embodiment, the conflict resolution and adaptive optimization solution module, when detecting potential operational conflicts, specifically performs the following: predicting the position and speed of each train in the future rolling time domain based on the state prediction value after delay compensation; calculating the predicted relative distance between any two trains based on the predicted position and speed; comparing the predicted relative distance with a safety distance threshold determined based on the topology matrix T(t) and track environment parameters; if the predicted relative distance is less than the safety distance threshold, it is determined that there is a potential conflict.

[0122] In one embodiment, when dynamically adjusting the weight parameters of the objective function, the conflict resolution and adaptive optimization solution module specifically performs the following: when a potential conflict is determined to exist, it increases the weight coefficient K_s of the safety distance term in the objective function and simultaneously decreases the weight coefficient K_v of the velocity tracking term.

[0123] In one embodiment, the conflict resolution and adaptive optimization solution module uses distributed game theory and a rolling time-domain negotiation mechanism to resolve conflicts. Specifically, it includes: each train's local conflict resolution and adaptive optimization solution module calculates the initial control intention based on local information; broadcasts the initial control intention to other trains in the train formation through the vehicle-to-vehicle communication network; receives the initial control intentions of other trains and conducts multilateral negotiation to reach a consensus on the conflict resolution strategy.

[0124] In one embodiment, the dynamic topology perception and modeling module uses graph neural network (GNN) and real-time digital twin technology to construct a dynamic topology model, where nodes represent carriages or trains, and edges represent buffer connections, communication links, and safety distance constraints.

[0125] In one embodiment, the communication delay compensation and robust predictive control module uses a Kalman filter based on delay differential equation (DDE) for state prediction compensation and employs a multipath redundancy communication protocol to transmit data.

[0126] In one embodiment, the environmental perception and dynamic parameter update module integrates an inertial measurement unit (IMU), lidar, and meteorological sensors, and uses machine learning algorithms to predict the track irregularity index R(s) and air drag coefficient C_d in real time.

[0127] In this embodiment, a dynamic topology sensing module monitors the train formation status in real time. A graph theory model is used to represent the formation topology, where nodes represent carriages and edges represent buffer connections and communication links. When the topology changes, the module automatically updates the dynamic equations of the multi-mass model, introducing topology-dependent constraints and variables into the optimization problem. This scheme introduces real-time topology sensing and an adaptive dynamic model, enabling the control strategy to dynamically respond to formation changes. Based on game theory or negotiation algorithms, each train calculates control commands locally, exchanges intentions through car-to-car communication, detects potential conflicts (such as speed conflicts and distance conflicts), and adjusts control commands when conflicts occur using priority strategies or rolling time-domain negotiation. Through distributed decision-making and conflict penalty terms, cooperative safety is ensured.

[0128] It should be understood that although the steps in the flowcharts of the embodiments described above are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the flowcharts of the embodiments described above may include multiple steps or multiple stages. These steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these steps or stages is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the steps or stages in other steps. It is understood that the steps in different embodiments can be freely combined as needed, and all non-contradictory solutions formed by such combinations are within the scope of protection of this application.

[0129] Based on the same inventive concept, this application also provides a heavy-haul train cooperative control device for implementing the aforementioned heavy-haul train cooperative control method. The solution provided by this device is similar to the implementation described in the above method; therefore, the specific limitations in one or more embodiments of the heavy-haul train cooperative control device provided below can be found in the limitations of the heavy-haul train cooperative control method described above, and will not be repeated here.

[0130] In one exemplary embodiment, such as Figure 4 As shown, a heavy-haul train cooperative control device 400 is provided, including: a dynamic topology sensing and modeling module 401, a communication delay compensation and robust predictive control module 402, an environmental sensing and dynamic parameter updating module 403, and a conflict resolution and adaptive optimization solution module 404, wherein:

[0131] The dynamic topology perception and modeling module 401 is used to acquire the physical connection status between carriages in each train group in the heavy-haul train system in real time, and update the time-varying topology matrix that represents the real-time connection relationship between carriages in each train group based on the changes in the acquired physical connection status; and reconstruct the multi-mass model describing the longitudinal dynamics of the train group according to the time-varying topology matrix, so as to dynamically adjust the action logic of the coupler force between adjacent carriages in the multi-mass model.

[0132] The communication delay compensation and robust prediction control module 402 is used to obtain the communication delay between each train group at the current time, and based on the reconstructed multi-mass model and the communication delay, to predict and compensate the received train status information of each train group, and output the compensated current time status prediction value.

[0133] The environmental perception and dynamic parameter update module 403 is used to perceive track environmental parameters in real time through on-board sensors, and dynamically update the corresponding parameters in the train running resistance model based on the perceived track environmental parameters, and generate dynamic driving safety constraints.

[0134] The conflict resolution and adaptive optimization solution module 404 is used to predict the operating state of each train formation in the future time domain based on the time-varying topology matrix, the current state prediction value, the updated train running resistance model, and dynamic traffic safety constraints. This allows each train formation to obtain control intentions from the predicted operating states and exchange control intentions with each other, thereby detecting whether there are potential operating conflicts between the train formations. If potential operating conflicts are detected, the weight parameters corresponding to the safety distance term and speed tracking term in the preset optimization objective function are dynamically adjusted. Under the premise of satisfying the dynamic constraints, dynamic traffic safety constraints, and preset actuator physical constraints determined by the variable topology matrix, the optimal control command for each train is solved with the goal of minimizing the dynamically adjusted optimization objective function and avoiding operating conflicts. The optimal control commands are then output to the execution control system of each train to achieve safe and coordinated operation of each train formation.

[0135] In an exemplary embodiment, the predicted operating status of each train formation in the future time domain includes the predicted position and predicted speed of each train in the future time domain. The conflict resolution and adaptive optimization solution module 404 is further configured to control each train formation to obtain the control intention in the future time domain based on the received predicted position and predicted speed of each train in the future time domain, and exchange control intentions with each other. After the train formations exchange control intentions, the module controls each train formation to perform collaborative simulation prediction to obtain the predicted relative distance of each train formation in the future time domain. The predicted relative distance is compared with a safety distance threshold jointly determined based on a time-varying topology matrix and dynamic traffic safety constraints. If the predicted relative distance is less than the safety distance threshold, a potential operating conflict is determined.

[0136] In an exemplary embodiment, the conflict resolution and adaptive optimization solution module 404 is further configured to increase the weight coefficient of the safety distance term in the optimization objective function and decrease the weight coefficient of the speed tracking term in the optimization objective function when a potential operational conflict is detected.

[0137] In an exemplary embodiment, the physical connection state includes the coupler locking state; the aforementioned dynamic topology sensing and modeling module 401 is further configured to sense the relative position of adjacent carriages and the coupler locking state between adjacent carriages through sensors deployed in each carriage, and to sense the connection state of the communication link using the communication unit; and to construct a time-varying topology matrix using graph neural network and digital twin technology; wherein, the nodes in the time-varying topology matrix represent carriages, and the edges represent the physical connection of the buffer, the connection of the communication link, and the preset safety distance constraints.

[0138] In an exemplary embodiment, the aforementioned communication delay compensation and robust prediction control module 402 is further configured to transmit data using a multipath redundancy communication protocol, and to obtain the communication delay of each communication link between each train group in real time by parsing the timestamp in the data packet; using a Kalman filter based on the delay differential equation as a state predictor, starting from the received train state information, and combining it with the reconstructed multi-mass model, to predict the time length of the communication delay in advance, and to predict the state of each train group at the current moment, as the compensated current moment state prediction value.

[0139] In an exemplary embodiment, the aforementioned environmental perception and dynamic parameter update module 403 is used to collect track gradient, curvature, irregularity, and ambient wind speed data in real time by integrating an inertial measurement unit, lidar, and meteorological sensors; to perform fusion analysis on the collected gradient, curvature, irregularity, and ambient wind speed data using machine learning algorithms; to calculate and update the track irregularity index and air resistance coefficient in the train running resistance model in real time; and to generate lateral acceleration safety constraints for the train when cornering based on the real-time perceived track curvature, as dynamic driving safety constraints.

[0140] Each module in the aforementioned heavy-haul train cooperative control device can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in or independent of the processor in a computer device, or stored in the memory of a computer device as software, so that the processor can call and execute the corresponding operations of each module.

[0141] In one exemplary embodiment, a computer device is provided, which may be a server, and its internal structure diagram may be as follows: Figure 5 As shown, the computer device includes a processor, memory, input / output (I / O) interfaces, and a communication interface. The processor, memory, and I / O interfaces are connected via a system bus, and the communication interface is also connected to the system bus via the I / O interfaces. The processor provides computational and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system, computer programs, and a database. The internal memory provides the environment for the operating system and computer programs stored in the non-volatile storage media to run. The I / O interfaces are used for exchanging information between the processor and external devices. The communication interface is used for communicating with external terminals via a network connection. When the computer program is executed by the processor, it implements a heavy-haul train cooperative control method.

[0142] Those skilled in the art will understand that Figure 5 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device to which the present application is applied. Specific computer devices may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0143] In one embodiment, a computer device is also provided, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps in the above method embodiments.

[0144] In one embodiment, a computer-readable storage medium is provided having a computer program stored thereon that, when executed by a processor, implements the steps in the above method embodiments.

[0145] In one embodiment, a computer program product is provided, including a computer program that, when executed by a processor, implements the steps in the above method embodiments.

[0146] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. Any references to memory, databases, or other media used in the embodiments provided in this application can include at least one of non-volatile memory and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can take many forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM). The databases involved in the embodiments provided in this application may include at least one type of relational database and non-relational database. Non-relational databases may include, but are not limited to, blockchain-based distributed databases. The processors involved in the embodiments provided in this application may be general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic devices, quantum computing-based data processing logic devices, artificial intelligence (AI) processors, etc., and are not limited to these.

[0147] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this application.

[0148] The embodiments described above are merely illustrative of several implementation methods of this application, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of this patent application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this application should be determined by the appended claims.

Claims

1. A method for coordinated control of heavy-haul trains, characterized in that, The method includes: The system acquires the physical connection status between carriages in each train formation in real time in the heavy-haul train system, and updates the time-varying topology matrix that characterizes the real-time connection relationship between carriages in each train formation based on the changes in the acquired physical connection status; and reconstructs the multi-mass model describing the longitudinal dynamics of the train formation according to the time-varying topology matrix, so as to dynamically adjust the action logic of the coupler force between adjacent carriages in the multi-mass model. The communication delay between each train group at the current time is obtained, and based on the reconstructed multi-mass model and the communication delay, the received train status information of each train group is predicted and compensated, and the compensated current time status prediction value is output. The vehicle uses onboard sensors to perceive track environment parameters in real time, and based on the perceived track environment parameters, dynamically updates the corresponding parameters in the train running resistance model and generates dynamic driving safety constraints. Based on the time-varying topology matrix, the current state prediction value, the updated train running resistance model, and the dynamic driving safety constraints, the operating states of each train formation in the future time domain are predicted, enabling each train formation to obtain control intentions from the predicted operating states and exchange control intentions with each other. This detects potential operating conflicts between train formations. If a potential operating conflict is detected, the weight parameters corresponding to the safety distance and speed tracking terms in the preset optimization objective function are dynamically adjusted. Under the premise of satisfying the dynamic constraints determined by the variable topology matrix, the dynamic driving safety constraints, and the preset actuator physical constraints, the optimal control commands for each train are solved with the goal of minimizing the dynamically adjusted objective function and avoiding operating conflicts. Each optimal control command is output to the execution control system of each train to achieve safe and coordinated operation of each train formation.

2. The method according to claim 1, characterized in that, The predicted operating status of each train formation in the future time domain includes the predicted position and speed of each train in the future time domain; The method for detecting potential operational conflicts between different train sets includes: The control system obtains the control intention for each train group in the future time domain based on the predicted position and speed of each train in the future time domain, and exchanges the control intentions with each other. After the trains in each formation exchange control intentions, the trains in each formation are controlled to perform collaborative simulation prediction to obtain the predicted relative distance of each train in the future time domain. The predicted relative distance is compared with a safe distance threshold determined jointly by the time-varying topology matrix and the dynamic driving safety constraints. If the predicted relative distance is less than the safe distance threshold, a potential operational conflict is determined.

3. The method according to claim 2, characterized in that, The step of dynamically adjusting the weight parameters corresponding to the safety distance term and the speed tracking term in the preset optimization objective function when a potential operational conflict is detected includes: If a potential operational conflict is detected, the weight coefficient of the safety distance term in the optimization objective function is increased, and the weight coefficient of the speed tracking term in the optimization objective function is decreased.

4. The method according to claim 1, characterized in that, The physical connection state includes the coupler locking state; the steps for constructing the time-varying topology matrix shown include: Sensors deployed in each carriage are used to sense the relative position of adjacent carriages and the locking status of the couplers between adjacent carriages, and the communication unit is used to sense the connection status of the communication link. A time-varying topology matrix is ​​constructed using graph neural networks and digital twin technology; wherein, the nodes in the time-varying topology matrix represent carriages, and the edges represent physical connections of buffers, communication links, and preset safety distance constraints.

5. The method according to claim 1, characterized in that, The process of obtaining the communication delay between each train group at the current time, and based on the reconstructed multi-mass model and the communication delay, predicting and compensating the received train state information of each train group, and outputting the compensated predicted state value at the current time, includes: Data is transmitted using a multipath redundancy communication protocol, and the communication delay of each communication link between each train group is obtained in real time by parsing the timestamp in the data packet. A Kalman filter based on the delay differential equation is used as the state predictor. Starting from the received train state information, and combined with the reconstructed multi-mass model, the communication delay time length is predicted forward to predict the state of each train group at the current moment, which is used as the compensated current moment state prediction value.

6. The method according to claim 1, characterized in that, The process of using onboard sensors to perceive track environment parameters in real time, dynamically updating corresponding parameters in the train running resistance model based on the perceived track environment parameters, and generating dynamic driving safety constraints includes: By integrating inertial measurement units, lidar, and meteorological sensors, the track's slope, curvature, unevenness, and ambient wind speed data are collected in real time. Machine learning algorithms are used to fuse and analyze the collected data on slope, curvature, unevenness, and environmental wind speed, and to calculate and update the track irregularity index and air resistance coefficient in the train running resistance model in real time. Based on the real-time perceived curvature of the track, a lateral acceleration safety constraint is generated when the train passes through a curve, which serves as the dynamic driving safety constraint.

7. A heavy-haul train cooperative control device, characterized in that, The device includes: The dynamic topology sensing and modeling module is used to acquire the physical connection status between carriages in each train formation in the heavy-haul train system in real time, and update the time-varying topology matrix that represents the real-time connection relationship between carriages in each train formation based on the changes in the acquired physical connection status; and reconstruct the multi-mass model describing the longitudinal dynamics of the train formation according to the time-varying topology matrix, so as to dynamically adjust the action logic of the coupler force between adjacent carriages in the multi-mass model. The communication delay compensation and robust predictive control module is used to obtain the communication delay between each train group at the current time, and based on the reconstructed multi-mass model and the communication delay, to predict and compensate the received train status information of each train group, and output the compensated current time status prediction value. The environmental perception and dynamic parameter update module is used to perceive track environmental parameters in real time through on-board sensors, and based on the perceived track environmental parameters, dynamically update the corresponding parameters in the train running resistance model and generate dynamic driving safety constraints. The conflict resolution and adaptive optimization module is used to predict the operating state of each train formation in the future time domain based on the time-varying topology matrix, the current state prediction value, the updated train running resistance model, and the dynamic traffic safety constraints. This allows each train formation to obtain control intentions from the predicted operating states and exchange control intentions with each other, thereby detecting potential operating conflicts between the train formations. If potential operating conflicts are detected, the module dynamically adjusts the weight parameters corresponding to the safety distance term and speed tracking term in the preset optimization objective function. Under the premise of satisfying the dynamic constraints determined by the variable topology matrix, the dynamic traffic safety constraints, and the preset actuator physical constraints, the module solves for the optimal control commands for each train, aiming to minimize the dynamically adjusted optimization objective function and avoid operating conflicts. The module outputs each of the optimal control commands to the execution control system of each train to achieve safe and coordinated operation of each train formation.

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 the steps of the method according to any one of claims 1 to 6.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 6.

10. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 6.

Images