Virtual marshalling train cooperative control method based on MB-RTK

Abstract

The invention discloses a virtual marshalling train cooperative control method based on MB-RTK, and the method is focused on a virtual marshalling operation scene, and achieves the high-precision relative positioning, clock synchronization and distributed cooperative control of multiple trains through a dynamic baseline real-time positioning technology, so as to guarantee the dynamic safety distance and operation consistency between the trains. According to the method, a plurality of key technologies such as a satellite navigation system, an inertial measurement unit, wireless communication and distributed control are comprehensively applied, a sensing-synchronization-control closed-loop system is constructed by taking a positioning result as input, and high-precision position and time consistency control between trains is realized, so that the safety and efficiency optimization of virtual marshalling operation are supported.

Description

A cooperative control method for virtual train formations based on MB-RTK Technical Field

[0001] This invention relates to the field of rail transit technology, and in particular to a virtual train formation cooperative control method based on MB-RTK. Background Technology

[0002] With the rapid development of intelligent railways and automatic driving technologies, train operation control systems are evolving from centralized scheduling to autonomous and collaborative systems. Traditional train control systems (such as CTCS-3 and CBTC) mostly adopt a "single-car operation, central monitoring" model, where each train needs to be indirectly controlled through ground-based central equipment. This centralized architecture suffers from problems such as large response delays, complex scheduling, and high system coupling when facing high-density, high-efficiency transportation demands. Virtual coupling technology, through wireless communication and high-precision positioning, enables multiple trains to operate collaboratively in a "virtual coupling" manner while maintaining safe spacing, achieving real-time speed coordination between trains and optimized allocation of track capacity. Compared with traditional physical coupling, virtual coupling has advantages such as flexible coupling and decoupling and increased track capacity, making it an important development direction for future autonomous train operation control systems.

[0003] However, the implementation of virtual train formation places extremely high demands on the relative positioning accuracy, time synchronization accuracy, and communication latency between trains. To ensure that the train spacing is maintained stably within a short distance, the system must achieve high-precision relative positioning between trains and high-time-consistency synchronization control.

[0004] Currently, the main methods for obtaining the relative position of trains in the railway system include: (1) Ground positioning method: relying on ground equipment such as transponders, track circuits, and axle counters for train detection. This method is limited by ground infrastructure and cannot provide direct relative positioning information between trains.

[0005] (2) Indirect positioning method based on "vehicle-to-ground-vehicle" communication: Each train uploads its own positioning information to the ground center system through the vehicle-to-ground-vehicle communication link, and the center estimates the interval between trains based on the absolute coordinate difference. This indirect calculation method has a large communication and processing delay, and the positioning update frequency and timestamp of different trains are different, which can easily lead to asynchronous and drifting of relative information.

[0006] (3) Relative positioning method based on vehicle-to-vehicle communication: Although the relative position calculation method using the vehicle-mounted speed measurement and positioning system combined with vehicle-to-vehicle direct communication can obtain relative displacement information in a short time, due to problems such as the cumulative error of the odometer and inertial navigation, the inability of errors between different trains to cancel each other out, inconsistent time references and communication delays that cannot be compensated, this method is difficult to maintain stable relative position accuracy for a long time and cannot meet the requirements of virtual formation for high-precision, high-reliability and high-consistency relative positioning.

[0007] (4) GNSS (Satellite Navigation System) single-point positioning method: a single vehicle uses satellite signals for absolute positioning, with a typical horizontal positioning error of 1.5 to 5 m, which can reach more than 10 m under multipath or obstruction conditions. This method cannot effectively eliminate common satellite error sources, and the resulting positioning results have strong correlation errors between different trains, making it difficult to guarantee the continuity and stability of relative position calculation, and failing to meet the accuracy requirements of dynamic interval control between virtual train formations.

[0008] Therefore, under the existing communication and positioning system, virtual train formation technology still needs to maintain a large safety margin in terms of train spacing, which limits the improvement of train formation density and restricts its promotion and application in trunk railways and heavy-haul transportation.

[0009] Moving Baseline Real-time Kinematic (MB-RTK) technology achieves high-precision relative positioning through carrier phase differential calculation between two dynamic receivers, and is the core means to realize high-precision relative perception of virtual train formations. This method calculates the real-time baseline vector when both sides are moving, thus not relying on static base stations, and has the advantages of flexible deployment and strong independence. However, existing MB-RTK technologies are mostly designed for general unmanned platforms (such as drones, intelligent vehicles, etc.), and the algorithm objectives are mainly focused on relative position solving and ambiguity fixing, without fully considering the following key requirements in rail transit systems: (1) The time synchronization problem is not solved. Trains need to maintain a unified time base to achieve synchronous control, but existing MB-RTK only outputs relative coordinates and does not have a clock alignment mechanism.

[0010] (2) The control loop has not been established. Traditional RTK positioning is only used as a navigation output, while virtual grouping requires the positioning results to directly drive the vehicle control system (such as speed adjustment and braking coordination) to form a closed-loop system of "perception-synchronization-control".

[0011] (3) Communication delay is not compensated. There are transmission asymmetry and time delay jitter in inter-train wireless communication. If delay compensation is not performed, the relative positioning result will be out of sync with the actual motion state.

[0012] (4) Track constraints were not introduced. Train operation is subject to track geometry constraints. The positioning model should combine linear coordinate system and running direction information to improve the stability of the solution.

[0013] Therefore, although the existing MB-RTK algorithm can theoretically achieve high relative positioning accuracy, there is still a significant system-level adaptation gap in the application scenario of virtual train formation, which is highly dynamic, highly constrained, and has low latency.

[0014] In summary, the realization of virtual train formations not only relies on high-precision relative positioning but also requires a systematic solution in areas such as clock synchronization, communication compensation, control coordination, and anomaly redundancy. Current research urgently needs a technical method that can achieve high-precision relative positioning between dynamic trains, achieve highly consistent clock synchronization through GNSS timing and message timestamps, and combine communication delay modeling and inertial navigation prediction to achieve robust spatiotemporal consistency. Furthermore, it should embed the positioning results into a distributed cooperative control algorithm for trains to achieve a dynamic balance between operational safety and efficiency. Summary of the Invention

[0015] The purpose of this invention is to provide a virtual train formation cooperative control method based on MB-RTK. It focuses on achieving high-precision relative positioning, clock synchronization and distributed cooperative control of multiple trains in a virtual train formation operation scenario through dynamic baseline real-time positioning technology (MB-RTK) to ensure dynamic safety distance and operational consistency between trains.

[0016] The objective of this invention is achieved through the following technical solution: a cooperative control method for virtual train formations based on MB-RTK, comprising: acquiring raw GNSS observations, IMU and odometer information of each train in the virtual formation; based on MB-RTK technology, combined with track constraint models and IMU and odometer information, performing real-time estimation of the relative positions between trains, and obtaining the relative baseline vector and relative speed between trains through optimized solution process; using GNSS time synchronization and timestamps obtained from train-to-train communication for bidirectional alignment, and combining clock deviation estimation and correction algorithms to establish a unified time reference among multiple trains; and under the unified time reference, compensating for the relative baseline vector and relative speed between trains based on train-to-train communication delay; and under the unified time reference, combining the compensated relative baseline vector and relative speed between trains, implementing cooperative control of trains through a distributed consensus algorithm to maintain the spatiotemporal consistency and safe interval of the virtual formation.

[0017] As can be seen from the technical solution provided by the present invention, the comprehensive technical solution, which integrates MB-RTK high-precision relative positioning, GNSS timing and delay compensation, a spatiotemporal synchronization mechanism, and a multi-train cooperative control strategy based on distributed consensus algorithm, constructs a "perception-synchronization-control" closed-loop system to achieve high-precision position and time consistency control between trains, thereby supporting the safety and efficiency optimization of virtual train formation operation. Attached Figure Description

[0018] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the following description of the embodiments will be briefly introduced. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0019] Figure 1 is a flowchart of a virtual train formation cooperative control method based on MB-RTK provided in an embodiment of the present invention.

[0020] Figure 2 is a schematic diagram of the overall architecture of a virtual train formation cooperative control method based on MB-RTK provided in an embodiment of the present invention.

[0021] Figure 3 is a schematic diagram of the overall process of a virtual train formation cooperative control method based on MB-RTK provided in an embodiment of the present invention.

[0022] Figure 4 is a schematic diagram of the information and control flow provided in an embodiment of the present invention. Detailed Implementation

[0023] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the protection scope of the present invention.

[0024] First, the following explanation is provided for terms that may be used in this document: Terms such as "comprising," "including," "containing," "having," or other similar semantic descriptions should be interpreted as non-exclusive inclusion. For example, "comprising a technical feature element (such as raw material, component, ingredient, carrier, dosage form, material, size, part, component, mechanism, device, step, process, method, reaction conditions, processing conditions, parameter, algorithm, signal, data, product or article of manufacture, etc.)" should be interpreted as including not only the explicitly listed technical feature element but also other technical feature elements known in the art that are not explicitly listed.

[0025] The term "composed of" excludes any technical features not expressly listed. When used in a claim, it closes the claim to exclude all technical features other than those expressly listed, except for associated conventional impurities. If the term appears only in a clause of a claim, it limits the claim to the elements expressly listed in that clause; elements recited in other clauses are not excluded from the overall claim.

[0026] The following is a detailed description of a virtual train formation cooperative control method based on MB-RTK provided by this invention. Contents not described in detail in the embodiments of this invention are prior art known to those skilled in the art. Where specific conditions are not specified in the embodiments of this invention, they are performed according to conventional conditions in the art or conditions recommended by the manufacturer. Where the manufacturers of the instruments used in the embodiments of this invention are not specified, they are all conventional products that can be purchased commercially.

[0027] Example 1: This invention provides a real-time cooperative control method for virtual train formations based on MB-RTK, as shown in Figure 1. It mainly includes: Step 1: Obtaining raw GNSS observations, IMU, and odometer information for each train in the virtual formation; Based on MB-RTK technology, combined with a track constraint model and multi-source information (such as IMU and odometer data), real-time estimation of the relative positions between trains is performed. Through optimized solution processes, the relative baseline vectors and relative speeds between trains are obtained.

[0028] In this embodiment of the invention, high-precision relative positioning based on MB-RTK technology is achieved by fusing multi-source data, including raw GNSS observations (including pseudorange and carrier phase observations), IMU and odometer information. Specifically, raw GNSS observations provide distance information between trains, while the IMU and odometer provide motion state information to improve positioning accuracy. During the calculation process, the relative baseline vector between trains is initially calculated using carrier phase differential, and then combined with ionospheric error (… ), multipath error ( ) and integer ambiguity ( After correction, the relative baseline vector and relative speed between trains are finally obtained. Specifically, this can be expressed as: ;in, , These are the three-dimensional coordinate vectors of the two trains. This represents the relative baseline vector between trains.

[0029] In this embodiment of the invention, the relative velocity is the relative baseline vector based on the rate of change over time. The relative velocity can be calculated from the relative baseline vector, and information such as relative acceleration can also be calculated at the same time.

[0030] In this embodiment of the invention, the track constraint model includes: the train operation is constrained by the track structure, and the train trajectory is a one-dimensional linear motion or a known curve in space; thereby, the track constraint model is established through the track centerline coordinate system, and the calculated relative baseline vector is projected onto the track direction vector.

[0031] Step 2: Use the timestamps obtained from GNSS time synchronization and vehicle-to-vehicle communication for bidirectional alignment. Combine clock deviation estimation and correction algorithms to establish a unified time reference among multiple trains. Under the unified time reference, compensate for the relative baseline vector and relative speed between trains based on vehicle-to-vehicle communication delay.

[0032] In this embodiment of the invention, the step of establishing a unified time reference among multiple trains by combining clock deviation estimation and correction algorithms includes: based on clock deviation estimation, combining sliding weighted average and least squares iterative correction methods to continuously correct the clock drift of each train, thereby achieving time alignment among multiple trains and establishing a unified time reference among multiple trains.

[0033] In this embodiment of the invention, the unified time reference includes three layers of time information: the first layer is the local time of each train, the second layer is the total time deviation of each train relative to the reference time, and the third layer is the time deviation between trains; wherein, the local time of each train is used to calculate the total time deviation of each train relative to the reference time; the total time deviation of each train relative to the reference time is used to correct the communication timestamp, thereby separating and estimating the actual train-to-train communication delay, and then compensating for the relative baseline vector and relative speed between trains respectively; the time deviation between trains is used for the coordinated control of trains.

[0034] In this embodiment of the invention, the total time deviation of the train relative to the reference time is obtained through clock deviation estimation, and is expressed as: ;in, This represents the total time deviation of train i relative to the reference time; This indicates the local time obtained by train i via GNSS timing. Indicates the defined reference time base; This represents the amount of time delay compensation introduced by the communication link. The time deviation... Used to correct the local time of each train in order to establish a unified time base for multiple trains.

[0035] In this embodiment of the invention, the time deviation between trains is also calculated using the following formula: ;in, The timestamp for when train i receives the message; The timestamp for when train j sends a message; This is an estimate of communication delay.

[0036] In this embodiment of the invention, the total time deviation of the train relative to the reference time is used to determine the train-to-train communication delay, and the relative baseline vector and relative speed between trains are compensated respectively. Specifically, the relative baseline vector can be directly compensated, and then the compensated relative speed can be calculated using the compensated relative baseline vector.

[0037] Step 3: Under a unified time reference, combining the compensated relative baseline vector and relative speed, a distributed consensus control algorithm is used to adjust the cooperative control between trains. This algorithm uses the state differences between trains (including relative distance, speed difference, and time deviation) as input to achieve precise coordination and maintain safe distances between trains.

[0038] In this embodiment of the invention, the compensated relative baseline vector and relative velocity output in step 2 serve as the basic description of the space-time state between trains. In this step, based on this, according to the track direction and control objective, the relative baseline vector is mapped to a relative distance deviation along the train's running direction. Combined with the relative velocity and the time deviation between trains under the unified time reference in step 2, state variables for coordinated control are constructed. Specifically, this is expressed as follows: ;in, Let be the consistency control law for train i. This represents the relative baseline vector between the two trains after compensation. The relative speed between the two trains after compensation. This refers to the time difference between two trains. These are the control gains for distance, speed, and time deviation, respectively.

[0039] For other trains, the same method is used to calculate the corresponding consistency control law; the consistency control law of each train is used to control the train operation, thereby realizing the coordinated control of the trains.

[0040] In this embodiment of the invention, the relative baseline vector calculation scheme between trains provided in step 1 is a general model that can be applied to the calculation of the relative baseline vector between any two trains in a virtual trainset. That is, for any given train, its relative baseline vector with all other trains can be calculated. Similarly, the time deviation calculation scheme between trains in step 2 is also applicable to the calculation of the time deviation between any two trains in a virtual trainset. However, in the cooperative control scheme involved in step 3, only the relative baseline vector and time deviation between two preceding and following trains are typically considered. Therefore, in the actual control process, only the relative baseline vector and time deviation between two preceding and following trains can be calculated, and after compensating for the relative baseline vector, the compensated relative speed can be calculated. Then, train cooperative control can be implemented based on the scheme in step 3.

[0041] Preferably, in this embodiment of the invention, the status during the train cooperative control process is also monitored, including: monitoring the quality of the train-to-train communication link, the clock synchronization error, and the solution confidence index obtained by bidirectional differential calculation; if the quality of the train-to-train communication link does not meet the set requirements, or the clock synchronization error exceeds the set threshold, or the solution confidence index is lower than the predetermined value, it indicates an abnormal operation, and the operating mode is switched through the operating mode switching mechanism.

[0042] The above-mentioned solution provided by the embodiments of the present invention is a comprehensive technical solution that integrates MB-RTK high-precision relative positioning, GNSS time synchronization and delay compensation, and multi-train cooperative control strategy based on distributed consensus algorithm. It constructs a "perception-synchronization-control" closed-loop system to achieve high-precision position and time consistency control between trains, thereby supporting the safety and efficiency optimization of virtual train formation operation.

[0043] To more clearly demonstrate the technical solution and its effects provided by the present invention, the method provided by the embodiments of the present invention will be described in detail below with reference to specific examples.

[0044] I. Overall Overview of the Plan

[0045] The existing virtual train formation operation control system has significant shortcomings in terms of relative positioning between trains, time synchronization, and communication delay compensation: traditional single-point positioning or ground detection methods are difficult to provide high-precision dynamic relative position information that meets the requirements of collaborative control; each train's local clock operates independently, resulting in asynchronous system sampling and inconsistent information time base; wireless communication links have random delays and jitter, causing delays in control signal transmission; at the same time, the centralized scheduling architecture lacks autonomous and collaborative mechanisms among multiple trains, making it difficult to support the safe and efficient operation of virtual train formations in complex sections.

[0046] This invention provides a virtual train formation cooperative control method based on MB-RTK. Focusing on virtual train formation operation scenarios, it utilizes dynamic baseline real-time positioning technology to achieve high-precision relative positioning, clock synchronization, and distributed cooperative control among multiple trains, ensuring dynamic safety distance and operational consistency between trains. This method comprehensively utilizes multiple key technologies, including GNSS, inertial measurement units (IMU), wireless communication, and distributed control. Using positioning results as input, it establishes a closed-loop architecture for spatiotemporal synchronization and cooperative control, enabling precise cooperative operation of trains even without physical coupling.

[0047] II. Detailed introduction of the plan.

[0048] 1. Overall architecture of the solution.

[0049] In this embodiment of the invention, the core idea of ​​the entire method is to construct a three-layer closed-loop architecture of "perception-synchronization-control" in a multi-train operation system. This architecture uses MB-RTK technology as its core to achieve high-precision relative positioning at the vehicle level; it combines GNSS time synchronization with onboard clock fusion to achieve consistent time synchronization; and based on this, it uses a distributed consistency control algorithm to achieve coordinated control of speed and spacing between trains. The overall architecture is shown in Figure 2.

[0050] The perception layer is responsible for sensing the relative position, speed, and attitude between trains, and includes multi-source sensing modules such as GNSS receivers, inertial measurement units, and odometers. This layer uses MB-RTK technology to calculate the relative baseline vectors between trains in real time, providing high-precision relative position information; it also calculates the relative speed of the trains.

[0051] Synchronization Layer: Responsible for maintaining consistency of time bases between trains. It utilizes GNSS PPS timing signals and train-to-train communication timestamps for dual correction, forming a distributed clock synchronization mechanism. This layer outputs relative state information under a unified time base, ensuring data synchronization across the entire trainset.

[0052] Control Layer: Responsible for speed coordination, distance maintenance, and safety control of multiple trains. This layer uses a distributed consensus algorithm to achieve collaborative control based on real-time information on relative position, speed difference, and time deviation, thereby maintaining the spatiotemporal consistency and safe intervals of train formations.

[0053] The above three layers constitute a unified information flow closed loop, forming a closed-loop autonomous operation.

[0054] 2. Overall process of the plan.

[0055] As shown in Figure 3, the overall steps of the MB-RTK-based virtual train cooperative control method provided by the present invention are as follows: Step S1, Data perception: Collect GNSS raw observations, IMU and odometer information, and exchange GNSS observations and timestamps between trains to form the input data required for relative positioning.

[0056] Step S2, Relative Positioning: Based on MB-RTK technology, perform bidirectional differential calculation on the original GNSS observations of multiple trains, and combine the track constraint model with GNSS+IMU+odometer multi-source fusion to obtain continuous relative baseline vectors and relative speeds.

[0057] Step S3, Time Synchronization: GNSS time synchronization and V2V communication (vehicle-to-vehicle communication) timestamps are used for bidirectional alignment. Combined with clock deviation estimation and correction algorithms, a unified time reference is achieved among multiple trains.

[0058] Step S4, Communication Delay Compensation: The communication delay is obtained by measuring the round-trip delay, and compensation is performed using statistical filtering and a state prediction model. The relative state data after delay correction is output.

[0059] Step S5, Distributed Cooperative Control: Calculate the consistency control law based on relative distance, speed difference and time deviation, generate acceleration / speed commands, and realize cooperative operation and safety interval maintenance among multiple trains.

[0060] Step S6, Anomaly Detection and Mode Switching: Monitor GNSS status, communication quality, and clock error. When the indicators exceed the limits, trigger the corresponding degradation mode, such as inertial navigation prediction mode or virtual decompilation mode, to ensure system security.

[0061] The mathematical formulas involved in the above steps are as follows: (1) MB-RTK solution model.

[0062] The relative baseline vector between train i and train j can be represented as: ;in, , These are the three-dimensional coordinate vectors of the two trains; This is the residual error term of the ionosphere; This is the path error term; It is the carrier phase integer ambiguity term. Let be the relative baseline vector between trains. Common error sources are eliminated through double-difference processing, and dynamic estimation is achieved by combining an ambiguity fixing algorithm with an extended Kalman filter.

[0063] (2) Time synchronization model.

[0064] Workshop time deviation can be expressed as: ;in, The timestamp for when train i receives the message; The timestamp for when train j sends a message; This is a communication delay estimate. Clock consistency is achieved by minimizing the sum of squared time differences between multiple nodes.

[0065] (3) Delay compensation prediction model.

[0066] To address information lag caused by communication delays, a state prediction model is used to handle neighbor vehicle state prediction and compensation during the delay period. Indicates communication delay Under these conditions, for trains The predicted position estimate after delay compensation is used to construct the consistent control input of the control layer to eliminate the impact of communication delay on the stability of cooperative control.

[0067] ;in, It is the actual location at the current moment; This is the current speed; It is the current acceleration.

[0068] The overall logic here is: first, estimate the actual train-to-train communication delay by combining the total time deviation of the train relative to the reference time. Then, the train's position is compensated to obtain the compensated train position. Finally, the relative baseline vector between trains is calculated based on the compensated train positions, and the calculation result is the compensated relative baseline vector.

[0069] (4) Consistency control law.

[0070] The control input of train i is constructed based on the compensated relative state output of the synchronization layer, and its consistency control law is... Represented as: ;in, Let be the consistency control law for train i. This represents the relative baseline vector between the two trains after compensation. The relative speed between the two trains after compensation. This refers to the time difference between two trains. These are the control gains for distance, speed, and time deviation, respectively.

[0071] 3. Signal and information flow process.

[0072] Each train's GNSS receiver acquires raw observations (pseudorange, carrier phase, timestamp), while the IMU acquires acceleration and angular velocity signals. Adjacent trains exchange raw GNSS observations and timestamps via vehicle-to-vehicle communication modules, calculating relative position (relative baseline vector) and relative velocity in real time based on MB-RTK technology. The synchronization layer receives GNSS timing signals, calculates the clock deviation of each train by combining the communication timestamp difference, outputs a unified time base through clock synchronization, and performs communication delay compensation by filtering and correcting the delay error of each status message based on round-trip delay measurement results and historical data to ensure data synchronization. The control layer receives the "spatiotemporal consistent state vector" output by the sensing layer and after time alignment and delay compensation by the synchronization layer, executes the consistency control algorithm to calculate control inputs, and adjusts traction, braking, or speed limit commands through the train traction control unit (TCU). Throughout the process, positioning reliability, clock error, and communication status are monitored in real time. When any indicator exceeds the limit, a mode switch is triggered to achieve operational safety redundancy. The entire process constitutes a closed-loop system as shown in Figure 4.

[0073] 4. Technical details of the solution.

[0074] (1) MB-RTK relative positioning technology.

[0075] MB-RTK relative positioning technology is implemented through an MB-RTK relative positioning module, primarily used for high-precision relative position determination between multiple trains. This module comprises a GNSS dual-frequency receiver, a carrier phase calculation unit, an ambiguity fixing algorithm module, and an inertial navigation data fusion unit. Its working principle is as follows: any one train can serve as a reference node, while the other trains act as moving nodes. Through vehicle-to-vehicle communication, they exchange raw GNSS observations and timestamp information in real time. The relative baseline vector between trains is calculated using carrier phase differential calculation. To ensure the continuity and reliability of the positioning results, this module employs a two-way differential and track constraint model, combined with IMU assistance to achieve dynamic integer ambiguity fixing, thereby maintaining stable calculations even under high-speed train movement and complex environments. The final output includes the relative baseline vector (ΔX, ΔY, ΔZ), relative velocity, and calculation confidence index, providing accurate spatial state input for subsequent clock synchronization and cooperative control modules. This invention further incorporates odometer velocity and displacement information during the baseline calculation process, and uses a multi-source fusion approach of GNSS+IMU+odometer to construct a continuous baseline estimation model, so as to improve the calculation stability and continuity in signal obstruction and handover sections.

[0076] This module takes the raw GNSS observations, IMU motion information, odometer information collected by each train, and the observation data and timestamps of adjacent trains obtained through vehicle-to-vehicle communication as inputs. It performs relative positioning calculation and constraint fusion processing on the input information to obtain the relative baseline vector, relative speed and corresponding calculation confidence between trains.

[0077] (2) Clock synchronization and timestamp correction.

[0078] Clock synchronization and timestamp correction are primarily achieved through a clock synchronization and timestamp correction module, used to ensure highly consistent time synchronization among trains in a trainset. Each train obtains its own absolute time base using GNSS timing signals and receives status information from adjacent trains via the train-to-train communication link. The link transmission delay is estimated by calculating the difference between the transmitted and received timestamps, and a distributed clock consistency algorithm is used to dynamically correct the local clock, thereby ensuring that the multi-train system operates under a unified time base.

[0079] To further improve synchronization accuracy and stability, a time deviation estimation model is introduced: ;in, This represents the total time deviation of train i relative to the reference time; This indicates the local time obtained by train i via GNSS timing. Indicates the defined reference time base; This represents the amount of time delay compensation introduced by the communication link. The model calculates the difference between the GNSS time of each train and the system reference time, and compensates for the communication link delay to obtain the actual deviation of the train's local clock.

[0080] Based on the above formula, a sliding weighted average and least squares iterative optimization method is used to continuously correct the clock drift of each node, achieving high-precision time alignment in a multi-vehicle network. The final output includes a unified synchronization time base, time drift compensation parameters, and standardized message timestamps, providing time-consistent input data for subsequent control stages. The time drift compensation parameters are used to continuously correct local clock offsets during subsequent time synchronization cycles and communication delay estimation to avoid interference from clock drift in link delay estimation. The standardized message timestamps serve as a unified time base, provided to the communication delay compensation module and the distributed cooperative control module for time alignment of relative state information, prediction compensation, and consistency control calculations.

[0081] (3) Communication delay estimation and compensation.

[0082] Communication delay estimation and compensation are primarily implemented through a dedicated module, used to identify and correct delay jitter in inter-train communication links in real time. Communication delay data is acquired through a round-trip time measurement mechanism, and statistical filtering algorithms (such as Kalman filtering or unscented Kalman filtering, UKF) are employed for dynamic delay estimation and prediction. Combined with an IMU prediction model, the relative position changes of trains are calculated to compensate for signal lag caused by communication network delays, thereby eliminating the impact of random communication network jitter on the timeliness of control information. To enhance the algorithm's adaptability, a weight adjustment mechanism based on communication quality indicators (including Received Signal Strength Index (RSSI), packet loss rate, etc.) is introduced. The delay prediction model is dynamically updated based on real-time network conditions, and the statistical filtering algorithm is used to estimate the delay change trend of the communication link. In the event of communication interruption, short-term state maintenance can be achieved through inertial navigation calculations and track geometric constraints, ensuring the continuity and safety of the control link. The final output includes delay-compensated relative state data and link quality indicators, providing timely and consistent input for distributed control.

[0083] This module takes the GNSS timing information of each train and the message timestamps obtained through vehicle-to-vehicle communication as input, performs time deviation estimation and correction processing on the input information, and forms a unified system time reference. The module output includes a synchronization time reference, time deviation compensation parameters, and standardized timestamps, which are used to provide time-consistent input information for the communication delay compensation module and the distributed cooperative control module.

[0084] (4) Distributed collaborative control.

[0085] Distributed cooperative control is mainly achieved through a distributed cooperative control module, which coordinates and controls the train speed, acceleration, and braking process based on spatiotemporally synchronized relative state information to maintain safe distances and consistency in train formation operation.

[0086] The control algorithm takes the relative state difference between adjacent trains as input and employs a consistency control law: ;in, Let be the consistency control law for train i. This represents the relative baseline vector between the two trains after compensation. The relative speed between the two trains after compensation. This refers to the time difference between two trains. These are the control gains for distance, speed, and time deviation, respectively.

[0087] This section can embed consensus algorithms (including improved Raft voting mechanisms or reinforcement learning strategies) as needed to achieve flexible control optimization. It supports dynamic lead car election and automatic train formation / disgrouping operations, and can automatically adjust the control topology based on the status of each node during operation. When local communication fails or a node malfunctions, the module can maintain control stability and safety through redundant information and self-healing mechanisms. Finally, the consensus control algorithm calculates train acceleration control commands, speed adjustment signals, and safety status indicators, providing real-time decision-making basis for the coordinated operation of the assembled trains.

[0088] This module takes the relative distance, relative speed, and time deviation information between trains after time synchronization and communication delay compensation as input, and calculates the train's control input according to the consistency control strategy. The module output includes traction, braking, or speed limit control commands and corresponding safety status flags, which are used to drive the train traction control unit to perform coordinated control.

[0089] (5) Track constraints.

[0090] The track constraint module introduces track geometric constraints during MB-RTK calculation and inertial navigation fusion to improve the stability and accuracy of train relative positioning. Train operation is strictly constrained by the track structure, and its trajectory can be approximated as a one-dimensional linear motion or a known curve in space. Based on this, a track constraint model is established using the track centerline coordinate system, and the calculated relative baseline vector is projected onto the track direction vector, thereby suppressing lateral drift and vertical noise errors.

[0091] The track constraint module outputs track constraint residuals and corrected position estimates. These results are fed back to the MB-RTK relative positioning module and the state fusion unit (which combines track constraint and positioning information) to further optimize positioning accuracy and ensure positioning stability in GNSS signal obstruction and complex environments. The final output relative baseline vector and relative velocity serve as key inputs for subsequent cooperative control and safety strategies, supporting the safe operation of virtual train formations.

[0092] This module takes the relative baseline vector and orbital geometry information output by the MB-RTK relative positioning module as input, applies orbital geometric constraints to the relative positioning results, and corrects the lateral and vertical errors in the positioning solution. The module outputs the positioning results after orbital constraint correction and constraint residual information, and feeds them back to the MB-RTK relative positioning module to improve the overall solution stability.

[0093] (6) Anomaly detection and mode switching.

[0094] Anomaly detection and mode switching are primarily implemented through an anomaly detection and mode switching module. This module is responsible for monitoring the operational status of key system components, including calculating confidence indices, communication link quality, and clock synchronization errors. This information assists in anomaly judgment and operational mode switching. The confidence indices are derived from the output of the MB-RTK relative positioning module and are used for status reliability assessment. Clock synchronization error refers to the deviation between the local clock of each train and the system reference clock, specifically calculated using the unified time base established in step 2.

[0095] When the clock synchronization error exceeds a set threshold, the system will trigger a mode switching mechanism. Specifically, when the timestamp difference between trains exceeds a preset threshold, indicating that the clock asynchrony has exceeded the allowable range, the system will immediately enter the "inertial navigation prediction mode," where the IMU and odometer calculate and maintain short-term positioning output. When the quality of the train-to-train communication link does not meet the set requirements (e.g., communication delay or packet loss rate exceeds the threshold), the system will execute the "virtual decoupling mode," temporarily disabling the coordinated control linkage between trains and switching to independent control. When the confidence index of the MB-RTK relative positioning calculation is lower than a set threshold, the system will also trigger a corresponding degraded operation mode (i.e., enter the "inertial navigation prediction mode," where the IMU and odometer calculate and maintain short-term positioning output) to ensure the safety and stability of system operation.

[0096] Once communication and computation confidence index conditions return to normal, the system can automatically rebuild the RTK baseline and restore spatiotemporal synchronization and cooperative control. This module ultimately outputs the system health status, operating mode flags, and switching commands, realizing the self-detection, self-adjustment, and safety redundancy functions of the virtual train formation control system.

[0097] This module takes the positioning confidence index, communication link status, and clock synchronization error as inputs to comprehensively evaluate the system's operating status. When an abnormal condition is detected and meets a preset threshold, it outputs the corresponding operating mode switching command and system health status flag, which are used to trigger the inertial navigation prediction mode, virtual decompression mode, or restore the cooperative control mode.

[0098] Compared with existing technologies, the MB-RTK-based virtual train cooperative control method proposed in this invention has significant innovation and practicality in terms of overall architecture, algorithm mechanism and operation performance. It can effectively solve the long-standing problems of large positioning error, clock asynchrony, communication delay and lack of distributed autonomous control in the operation of virtual trains. Its main beneficial effects are as follows: (1) It achieves high-precision relative positioning accuracy and significantly improves the perception capability between trains.

[0099] Traditional CTCS and CBTC systems rely on ground equipment for train detection, failing to provide high-precision relative position information at the vehicle-to-vehicle level, thus hindering the reduction of safe intervals between virtual train formations. This invention employs a two-way differential and track-constraint model to eliminate atmospheric and satellite common errors, and combines inertial navigation information to achieve dynamic integer ambiguity fixation, maintaining continuous and stable high-precision baseline estimation even under high-speed train operation and complex electromagnetic environments. By introducing an MB-RTK relative positioning mechanism, this invention not only cancels errors through two-way differential but also achieves high-precision relative position calculation between trains without a static reference station. This invention significantly improves the robustness and real-time performance of relative position calculation, supporting the safe operation requirements of smaller dynamic intervals between trains and providing core technical support for the "precise sensing" of virtual train formations.

[0100] (2) Achieve time synchronization of multiple trains and ensure spatiotemporal consistency of the multiple train system.

[0101] In multi-train cooperative operation scenarios, if the clocks of each train are not synchronized, it will lead to asynchronous state sampling, delayed control response, and error accumulation, which will seriously affect safety and stability.

[0102] This invention proposes a distributed clock synchronization mechanism based on the fusion of GNSS time synchronization and communication timestamps to achieve time alignment. By introducing a multi-node clock synchronization and consistency algorithm, the local clock deviation of each train is corrected in real time to ensure that all train clocks are synchronized. Combined with a communication delay compensation model, clock differences caused by communication delays are corrected to ensure that data transmission between trains takes place under a unified time reference, thereby maintaining the spatiotemporal consistency of the system. Unlike traditional single-train time synchronization methods, this invention achieves a closed-loop spatiotemporal synchronization of the entire train formation at the system level, ensuring that the status information of all trains is updated and controlled under a unified time coordinate, fundamentally eliminating the impact of asynchronous sampling on control accuracy.

[0103] (3) It has the ability to predict and compensate for communication delays, which significantly improves the real-time performance and robustness of control.

[0104] Train-to-train communication under 5G-R or LTE-M networks still suffers from time-varying delays and random jitter. Without compensation, this can lead to control command lag and system oscillations. This invention designs a communication delay modeling mechanism based on adaptive filtering and predictive compensation, which can estimate and correct delay errors introduced by the communication link in real time. This mechanism uses round-trip time measurements and historical delay sequences to establish a statistical model and predicts the dynamic delay distribution through UKF / EKF filtering. When communication is interrupted, it automatically switches to inertial navigation calculation mode to maintain short-term control continuity. This method can control delay compensation errors to a minimum, significantly improving the system's real-time performance and stability, enabling virtual train formations to maintain stable control performance even under complex network conditions.

[0105] (4) Establish a distributed collaborative control mechanism to realize the autonomous and coordinated operation of multiple trains.

[0106] Existing train control systems primarily rely on centralized dispatching by a ground control center (CTC / RBC), lacking direct collaboration and autonomy between trains. This invention constructs a distributed consensus control algorithm based on relative position, speed difference, and time deviation to achieve autonomous collaboration between trains. Through the consensus control law, each train can adjust its speed and spacing in real time based on the status of neighboring trains under local communication conditions, achieving full trainset consistency without a central node. This mechanism supports dynamic master train election and disbanding operations, and can quickly reorganize the control network in the event of train failure or communication interruption, exhibiting strong self-healing and fault tolerance capabilities.

[0107] (5) Construct a closed-loop architecture of "perception-synchronization-control" to improve the overall reliability and security of the system.

[0108] Traditional virtual train formation research often focuses on single improvements to positioning algorithms or communication protocols, lacking system-level integration and closed-loop verification. This invention deeply integrates four key aspects—MB-RTK, clock synchronization, communication compensation, and control coordination—to form a complete "perception-synchronization-control" closed-loop system, as shown in Figure 2. This multi-layered fusion architecture enables the perception layer to continuously provide accurate relative train status, the synchronization layer to ensure system clock consistency and predictable communication delays, and the control layer to perform coordinated control and safety decisions based on unified spatiotemporal information. Under this closed-loop mechanism, the system can maintain stable operation under various abnormal conditions, supporting real-time safe distance maintenance, dynamic train formation and disassembly control, and energy-optimized scheduling for virtual train formations, achieving highly safe and efficient intelligent operation.

[0109] (6) It has engineering feasibility and scalability.

[0110] The system architecture of this invention is compatible with existing GNSS, IMU, and train control communication platforms, and can be directly deployed in onboard equipment without modifying the ground signaling system. Its modular design allows for scalability to various train types and operating scenarios (heavy-haul railways, high-speed railways, urban rail transit, etc.). Furthermore, this invention can be combined with existing virtual train formation control algorithms, reinforcement learning controllers, or model predictive control frameworks to achieve energy consumption optimization and improved operational efficiency. Therefore, this invention not only possesses significant theoretical innovation but also demonstrates direct engineering feasibility and widespread application value.

[0111] Through the above description of the embodiments, those skilled in the art can clearly understand that the above embodiments can be implemented by software, or by using software plus necessary general-purpose hardware platforms. Based on this understanding, the technical solutions of the above embodiments can be embodied in the form of a software product, which can be stored in a non-volatile storage medium (such as a CD-ROM, USB flash drive, mobile hard drive, etc.), including several instructions to cause a computer device (such as a personal computer, server, or network device, etc.) to execute the methods described in the various embodiments of the present invention.

[0112] The above description is merely a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims. The information disclosed in the background section is intended only to enhance the understanding of the overall background technology of the present invention and should not be construed as an admission or implication in any way that such information constitutes prior art known to those skilled in the art.

Claims

1. A cooperative control method for virtual train formations based on MB-RTK, characterized in that, include: Obtain raw GNSS observations, IMU and odometer information for each train in the virtual train formation; Based on MB-RTK technology, combined with track constraint model and IMU and odometer information, the relative position between trains is estimated in real time. By optimizing the solution process, the relative baseline vector and relative speed between trains are obtained. By using GNSS time synchronization and timestamps obtained from vehicle-to-vehicle communication for bidirectional alignment, and combining clock deviation estimation and correction algorithms, a unified time base is established among multiple trains. Under the unified time base, the relative baseline vector and relative speed between trains are compensated based on the vehicle-to-vehicle communication delay. Under the unified time base, and combining the compensated relative baseline vector and relative speed between trains, a distributed consensus algorithm is used to achieve coordinated control of trains, maintaining the spatiotemporal consistency and safe interval of the virtual train formation.

2. The method for cooperative control of virtual train formations based on MB-RTK according to claim 1, characterized in that, The process of optimizing the solution to obtain the relative baseline vector and relative speed between trains includes: initially calculating the relative baseline vector between trains using carrier phase differential, and then combining it with ionospheric error. Multipath error and whole week blur The correction is made as follows: ;in, It is the relative baseline vector between train i and train j. ， Let be the three-dimensional coordinate vectors of train i and train j; This is the residual error term of the ionosphere; This is the path error term; It is the carrier phase integer ambiguity term; then the relative speed is calculated using the relative baseline vector between the trains.

3. The method for cooperative control of virtual train formations based on MB-RTK according to claim 1, characterized in that, The track constraint model includes: train operation is constrained by the track structure, and the train trajectory is a one-dimensional linear motion or a known curve in space; thereby, the track constraint model is established through the track centerline coordinate system, and the calculated relative baseline vector is projected onto the track direction vector.

4. The method for cooperative control of virtual train formations based on MB-RTK according to claim 1, characterized in that, The unified time reference includes three layers of time information: the first layer is the local time of each train, the second layer is the total time deviation of each train relative to the reference time, and the third layer is the time deviation between trains. Among them, the local time of each train is used to calculate the total time deviation of each train relative to the reference time; the total time deviation of each train relative to the reference time is used to correct the communication timestamps to separate and estimate the actual train-to-train communication delay, and then to compensate for the relative baseline vector and relative speed between trains respectively; the time deviation between trains is used for the coordinated control of trains.

5. The method for cooperative control of virtual train formations based on MB-RTK according to claim 4, characterized in that, The total time deviation of the train relative to the reference time is obtained by clock deviation estimation and is expressed as: ;in, This represents the total time deviation of train i relative to the reference time; This indicates the local time obtained by train i via GNSS timing. Indicates the defined reference time base; This indicates the amount of time delay compensation introduced by the communication link.

6. The method for cooperative control of virtual train formations based on MB-RTK according to claim 4, characterized in that, The time deviation between trains is calculated using the following formula: ;in, The timestamp for when train i receives the message; The timestamp for when train j sends a message; This is an estimate of communication delay.

7. The method for cooperative control of virtual train formations based on MB-RTK according to claim 1, characterized in that, Under a unified time reference, and combining the compensated relative baseline vectors and relative speeds between trains, coordinated train control is achieved through a distributed consensus algorithm. This includes: generating a consensus control law by combining the time deviations between trains under the unified time reference and the compensated relative baseline vectors and relative speeds; and using the consensus control laws of each train to control train operation, thus achieving coordinated train control. For train i, a consensus control law is generated. , is represented as: ;in, Let be the consistency control law for train i. This represents the relative baseline vector between the two trains after compensation. The relative speed between the two trains after compensation. This refers to the time difference between two trains. These are the control gains for distance, speed, and time deviation, respectively.

8. The method for cooperative control of virtual train formations based on MB-RTK according to claim 1, characterized in that, Also includes: The status of the train cooperative control process is monitored, including: monitoring the quality of the train-to-train communication link, clock synchronization error, and the solution confidence index obtained by bidirectional differential calculation; if the quality of the train-to-train communication link does not meet the set requirements, or the clock synchronization error exceeds the set threshold, or the solution confidence index is lower than the predetermined value, it indicates an abnormal operation, and the operating mode is switched through the operation mode switching mechanism.

9. The method for cooperative control of virtual train formations based on MB-RTK according to claim 8, characterized in that, The working mode switching mechanism includes: when the clock synchronization error exceeds a set threshold, the mode switching mechanism will be triggered to enter the inertial navigation prediction mode, and the positioning output will be maintained by the IMU and odometer calculation; when the calculated confidence index is lower than the set threshold, the corresponding degraded operating mode will be triggered, that is, the inertial navigation prediction mode will be entered, and the positioning output will be maintained by the IMU and odometer calculation; when the quality of the vehicle-to-vehicle communication link does not meet the set requirements, the virtual decoupling mode will be executed to temporarily release the cooperative control linkage between trains and switch to independent control state.

Images