Digital twin traffic flow intelligent monitoring method and system

Abstract

The invention discloses a digital twinborn traffic flow intelligent monitoring method and system, and relates to the technical field of intelligent traffic, and the method comprises the steps: obtaining basic configuration information, dividing a road network into a plurality of local regions, building a local digital twinborn body at an edge node, and building a global digital twinborn body at a cloud end; the edge node receives the multi-source traffic perception data, carries out time-space alignment fusion, extracts traffic features, updates local digital twins, carries out short-time traffic state deduction and generates a local control strategy; and the edge nodes upload model updating information to the cloud, the cloud aggregates and updates the global digital twins and issues model parameters, meanwhile, medium and long-term traffic state deduction is carried out, and a global optimization control strategy is generated, decomposed and issued. According to the invention, real-time monitoring and multi-time scale prediction of the traffic flow are realized through a cloud-side collaborative double-layer digital twinborn architecture and a bidirectional feedback mechanism, and the monitoring precision and the response speed are improved.

Description

Technical Field

[0001] This invention relates to the field of intelligent transportation technology, and in particular to a digital twin-based intelligent traffic flow monitoring method and system. Background Technology

[0002] With the acceleration of urbanization, problems such as road traffic congestion and frequent accidents are becoming increasingly prominent, placing higher demands on real-time monitoring and intelligent management of traffic flow. Traditional traffic monitoring systems adopt a centralized architecture, with data processing and decision-making taking place in the traffic control center. When dealing with massive amounts of multi-source data, they face challenges such as limited communication bandwidth, large data transmission latency, and insufficient real-time performance. In recent years, edge computing and digital twin technologies have been increasingly applied to the field of intelligent transportation. Edge computing enables local data processing through roadside nodes, while digital twin technology maps the real road network status in virtual space, providing new technical means for traffic monitoring and decision-making.

[0003] However, existing solutions still have shortcomings in areas such as the collaborative mechanism between edge nodes and the cloud, and the real-time updating of digital twin models. Some solutions focus on local processing at the edge, making it difficult to achieve global optimization; others focus on centralized modeling in the cloud, which limits real-time performance. In addition, the parameter adjustments of traffic prediction models largely rely on offline training with historical data, making it difficult to adaptively adjust according to changes in real-time traffic conditions, thus affecting prediction accuracy and control effectiveness. Summary of the Invention

[0004] In view of the aforementioned existing problems, the present invention is proposed.

[0005] Therefore, this invention provides a digital twin-based intelligent traffic flow monitoring method and system to address the technical problems of insufficient cloud-edge collaboration, difficulty in balancing global and local optimization, and insufficient model adaptive update capability.

[0006] To solve the above-mentioned technical problems, the present invention provides the following technical solution:

[0007] In a first aspect, the present invention provides a digital twin-based intelligent traffic flow monitoring method, comprising:

[0008] Obtain basic configuration information, divide the road network into multiple local areas, establish local digital twins at each edge node, and establish a global digital twin in the cloud;

[0009] Each edge node receives multi-source traffic sensing data within its jurisdiction, performs spatiotemporal alignment and fusion processing on the multi-source traffic sensing data, and extracts traffic features.

[0010] The state variables of the local digital twin are updated based on traffic characteristics, and short-term traffic state is inferred based on the updated local digital twin to generate a local control strategy.

[0011] Each edge node uploads model update information to the cloud according to a preset cycle. The cloud aggregates and processes the uploaded information and updates the global digital twin, then sends the updated model parameters to each edge node.

[0012] The cloud-based system performs medium- to long-term traffic state simulations based on a global digital twin, generates a global optimization control strategy, and decomposes the global optimization control strategy into local sub-strategies and distributes them to the corresponding edge nodes.

[0013] Edge nodes correct local digital twins based on the deviation between predicted and actual observed values, and feed the correction information back to the cloud for global digital twin optimization.

[0014] As a preferred embodiment of the digital twin-based intelligent traffic flow monitoring method of the present invention, the method comprises: dividing the road network into multiple local areas, including:

[0015] Based on the road network topology information, node communication range and node computing capabilities, a graph clustering algorithm is used to divide the road network into multiple local regions and establish a mapping relationship between regions and edge nodes.

[0016] A local digital twin is established at each edge node. The local digital twin contains a local road network model, a local traffic flow model, and a set of state variables.

[0017] A global digital twin is built in the cloud, which includes a full road network model, a global traffic flow model, and an optimization engine.

[0018] As a preferred embodiment of the intelligent traffic flow monitoring method based on digital twins of the present invention, the extraction of traffic features includes:

[0019] Each edge computing node receives at least two types of traffic sensing data within its jurisdiction;

[0020] Spatiotemporal alignment of multi-source traffic perception data and unification to the road network coordinate system;

[0021] Data fusion and abnormal data removal are performed locally at the edge nodes;

[0022] Traffic features are extracted, including vehicle location, speed, acceleration, vehicle type classification, and lane occupancy information.

[0023] As a preferred embodiment of the digital twin-based intelligent traffic flow monitoring method of the present invention, the generation of local control strategies includes:

[0024] The state variables of the local digital twin are updated based on traffic characteristics. The state variables include road segment flow, average speed, lane occupancy rate, and queue length.

[0025] An incremental mapping strategy is adopted to transmit the state variables that have changed compared to the previous moment to the global digital twin in the cloud.

[0026] The timing prediction model runs at the edge node and extrapolates the short-term traffic conditions based on the updated local digital twin.

[0027] Based on the simulation results, local control strategies are generated and distributed to the execution equipment. The local control strategies include signal timing adjustment, variable lane control, and local guidance information.

[0028] As a preferred embodiment of the intelligent traffic flow monitoring method based on digital twins of the present invention, the cloud-based aggregation and updating of uploaded information and the global digital twin includes:

[0029] The cloud receives model update information from multiple edge nodes and calculates the aggregation weight by comprehensively considering the data quality, sample quantity, and road network topology correlation of each edge node.

[0030] A federated learning framework is used to weight and aggregate model update information based on aggregation weights;

[0031] Differential privacy techniques are used to process the aggregated model gradients and update the model parameters of the global digital twin.

[0032] The updated model parameters are distributed to each edge node, and each edge node makes personalized adjustments to the model based on the global model parameters and local features.

[0033] As a preferred embodiment of the intelligent traffic flow monitoring method based on digital twins of the present invention, the cloud-based medium- and long-term traffic state projection based on a global digital twin includes:

[0034] Based on a global digital twin, the cloud uses a traffic condition projection model to project the future medium- and long-term traffic conditions of the entire road network.

[0035] Analyze the cross-regional traffic flow propagation characteristics during the simulation process, and identify cross-regional traffic flow propagation relationships and potential congestion diffusion paths;

[0036] Based on the simulation results and the identified congestion propagation paths, multiple candidate schemes for global optimization control strategies are generated.

[0037] The execution effects of each candidate scheme are simulated in a global digital twin environment to evaluate the degree of improvement in traffic capacity and the degree of reduction in delay, and the optimal strategy is selected based on the evaluation results.

[0038] The optimal strategy is decomposed into local sub-strategies according to the local region boundary, the execution sequence of each sub-strategy is determined and distributed to the corresponding edge nodes.

[0039] As a preferred embodiment of the intelligent traffic flow monitoring method based on digital twins of the present invention, the correction of the local digital twin includes:

[0040] The edge nodes continuously monitor the actual traffic conditions after the control strategy is implemented.

[0041] Calculate the deviation between the predicted values ​​from the local digital twin and the actual observed values;

[0042] When the deviation meets the triggering condition, the edge node corrects the parameters of the local digital twin model;

[0043] The model calibration information and deviation statistics are uploaded to the cloud, and the cloud analyzes the deviation distribution characteristics of multiple edge nodes.

[0044] When the cloud determines that there is a systematic bias, it updates the model parameters of the global digital twin.

[0045] Secondly, the present invention provides a digital twin-based intelligent traffic flow monitoring system, comprising:

[0046] The digital twin construction module is used to obtain basic configuration information, divide the road network into multiple local areas, establish local digital twins at each edge node, and establish a global digital twin in the cloud;

[0047] The multi-source data fusion module is used to receive multi-source traffic sensing data, perform spatiotemporal alignment and fusion processing on the multi-source traffic sensing data, and extract traffic features;

[0048] The local simulation control module is used to update the state variables of the local digital twin based on traffic characteristics, perform short-term traffic state simulation based on the updated local digital twin, and generate local control strategies.

[0049] The cloud-edge collaborative optimization module is used to upload model update information to the cloud at a preset cycle. The cloud aggregates and processes the uploaded information and updates the global digital twin, and then sends the updated model parameters to each edge node.

[0050] The global simulation and control module is used to perform medium- and long-term traffic state simulations based on the global digital twin, generate a global optimization control strategy, and decompose the global optimization control strategy into local sub-strategies and distribute them to the corresponding edge nodes.

[0051] The adaptive correction module is used to correct the local digital twin based on the deviation between the predicted value and the actual observed value, and feeds the correction information back to the cloud for global digital twin optimization.

[0052] Thirdly, the present invention provides a computer device including a memory and a processor, wherein the memory stores a computer program, wherein when the computer program is executed by the processor, it implements any step of the intelligent traffic flow monitoring method of digital twin as described in the first aspect of the present invention.

[0053] Fourthly, the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein, when executed by a processor, the computer program implements any step of the intelligent traffic flow monitoring method based on digital twins as described in the first aspect of the present invention.

[0054] The beneficial effects of this invention are as follows: This invention innovatively constructs a distributed two-level digital twin system, achieving deep collaboration between real-time edge control and global optimization in the cloud. Edge nodes undertake multi-source fusion and short-term extrapolation, while the cloud is responsible for medium- and long-term prediction and cross-regional policy collaboration, forming a complementary control framework across multiple spatiotemporal scales. A cloud-edge collaborative update mechanism based on federated learning, combined with differential privacy protection, ensures data security while reducing communication overhead. Deviation-driven adaptive correction endows the system with continuous evolution capabilities, breaking through the limitations of traditional static models and significantly improving prediction accuracy and control performance in complex traffic scenarios. Attached Figure Description

[0055] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying 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.

[0056] Figure 1 A flowchart for a digital twin-based intelligent traffic flow monitoring method.

[0057] Figure 2 The flowchart shows the initialization process of the distributed digital twin architecture for intelligent traffic flow monitoring based on digital twins.

[0058] Figure 3 A flowchart for generating local control strategies for intelligent traffic flow monitoring methods based on digital twins.

[0059] Figure 4 This is a module connection diagram for a digital twin-based intelligent traffic flow monitoring system. Detailed Implementation

[0060] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0061] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the spirit of the invention. Therefore, the invention is not limited to the specific embodiments disclosed below.

[0062] Secondly, the term "an embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in different places throughout this specification does not necessarily refer to the same embodiment, nor is it a single embodiment or an embodiment selectively excluded from other embodiments.

[0063] Reference Figures 1-4 As one embodiment of the present invention, this embodiment provides a digital twin-based intelligent traffic flow monitoring method, the flowchart of which is shown below. Figure 1 As shown, the method includes the following steps:

[0064] S1: Obtain basic configuration information, divide the road network into multiple local areas, establish local digital twins at each edge node, and establish a global digital twin in the cloud.

[0065] Specifically, the initialization flowchart of the distributed digital twin architecture is as follows: Figure 1 As shown, it includes the following steps:

[0066] S1.1: Obtain road network topology information, edge node deployment information, and sensor deployment information for the target monitoring area.

[0067] The road network topology information includes road segment attributes, intersection attributes, and road network connectivity. Road segment attributes include road segment identification, length, number of lanes, and traffic capacity. Intersection attributes include intersection identification, geographic coordinates, and type. Edge node deployment information includes node identification, deployment location, computing power, and communication coverage. Sensor deployment information includes sensor type, deployment location, and coverage.

[0068] S1.2: Based on the road network topology information, node communication range and node computing capabilities, a graph clustering algorithm is used to divide the road network into multiple local regions and establish a mapping relationship between regions and edge nodes.

[0069] Preferably, a weighted graph model is constructed based on road network topology information. In the graph model, nodes represent intersections, edges represent road segments, and the edge weights comprehensively consider the spatial distance and traffic attributes of the road segments. A graph clustering algorithm is used to perform cluster analysis on the weighted graph model, dividing the road network into multiple local regions. During the clustering process, constraints on node communication range and node computing power are considered to ensure that road segments within each local region are within the communication coverage of the corresponding edge nodes, and that the traffic load of each region matches the computing power of the edge nodes. Based on the deployment location of the edge nodes, each local region is mapped to the nearest or most load-appropriate edge node.

[0070] In one embodiment, the graph clustering algorithm can employ either the Louvain algorithm or the spectral clustering algorithm. The Louvain algorithm is suitable for the rapid partitioning of large-scale road networks, while the spectral clustering algorithm is suitable for scenarios requiring high partitioning quality.

[0071] S1.3: Establish a local digital twin at each edge node.

[0072] The local digital twin includes a local road network model, a local traffic flow model, and a set of state variables.

[0073] Specifically, the local road network model includes road segments, intersections, and their topological connections within the jurisdiction of the edge node. The local traffic flow model describes the vehicle movement patterns within this area, employing micro- or meso-level traffic flow models such as car-following models or cellular transport models. The set of state variables includes traffic state parameters such as real-time traffic flow, average speed, lane occupancy, and queue length for each road segment. After the local digital twin is established, the initial values ​​of each state variable are initialized to prepare for data collection in step S2 and state updates in step S3.

[0074] S1.4: Establish a global digital twin in the cloud.

[0075] The global digital twin includes a full road network model, a global traffic flow model, and an optimization engine.

[0076] Specifically, the full road network model integrates road network information from all local areas to form a complete road network topology covering the entire monitoring area. The global traffic flow model describes the traffic flow propagation patterns at the road network level, employing either a macroscopic traffic flow model or a data-driven model based on graph neural networks. The optimization engine generates global optimization control strategies, including signal coordination optimization, traffic guidance strategy generation, and emergency dispatch. After the global digital twin is established, data interaction interfaces are established with the local digital twins of each edge node, laying the foundation for subsequent cloud-edge collaboration in step S4 and global optimization in step S5.

[0077] Through the above steps, this invention completes the initial deployment of a distributed digital twin architecture, realizing a two-level distribution of local digital twins at the edge and global digital twins at the cloud. This architecture decentralizes traffic monitoring and control capabilities to edge nodes, reducing communication bandwidth consumption and transmission latency; intelligent regional division matches the traffic load of each local area with the computing power of the nodes, improving system scalability; and it retains the cloud's overall modeling and optimization capabilities for the entire road network, balancing real-time and global requirements.

[0078] S2: Each edge node receives multi-source traffic perception data within its jurisdiction, performs spatiotemporal alignment and fusion processing on the multi-source traffic perception data, and extracts traffic features.

[0079] To ensure data quality and real-time performance, edge nodes preprocess and fuse the raw sensing data to reduce transmission overhead.

[0080] Specifically, S2 includes the following sub-steps:

[0081] S2.1: Each edge computing node receives at least two types of traffic perception data within its jurisdiction.

[0082] Traffic sensing data includes video image data, radar detection data, loop detector data, and floating car GPS trajectory data. These different data types are complementary, and multi-source data fusion can improve detection accuracy and reliability.

[0083] Edge nodes perform format parsing and preprocessing on the received raw data: extracting vehicle target and motion information from video data; identifying targets and extracting position and speed from radar data; adding timestamps to coil data; and performing map matching on GPS trajectories. Preprocessing converts data of different formats into a unified data format.

[0084] S2.2: Perform spatiotemporal alignment of multi-source traffic perception data and unify it to the road network coordinate system.

[0085] It should be noted that different sensing devices have different clock references, sampling frequencies, and coordinate systems, so spatiotemporal alignment processing is required.

[0086] S2.3: Perform data fusion and remove abnormal data locally at the edge node.

[0087] Vehicle observation data from different data sources are fused to obtain fused vehicle state information, including position, velocity, and acceleration. The fusion process comprehensively considers the observation values ​​and error characteristics of each data source, and obtains the fusion result through weighted calculation.

[0088] S2.4: Extract traffic features, including vehicle location, speed, acceleration, vehicle type classification, and lane occupancy information.

[0089] Traffic features are extracted from the fused data. Vehicle location includes the road segment, lane number, and location coordinates; vehicle speed and acceleration are derived from the fusion results; vehicle type classification is obtained from video recognition; and lane occupancy information is obtained from vehicle location statistics.

[0090] Furthermore, segment-level aggregated features can be extracted, including macroscopic traffic parameters such as average segment speed, traffic flow, and lane occupancy. The extracted features are stored in a structured format, achieving a significant reduction in data volume compared to the original sensor data.

[0091] Through the above processing, the fusion and feature extraction of multi-source data are completed locally at the edge node, which improves data quality, reduces the amount of data transmitted from the cloud to the edge and the latency, and provides a reliable data foundation for subsequent digital twin updates and traffic simulation.

[0092] S3: Update the state variables of the local digital twin based on traffic characteristics, perform short-term traffic state simulation based on the updated local digital twin, and generate a local control strategy.

[0093] Specifically, the flowchart for generating local control strategies is as follows: Figure 3 As shown, it includes the following sub-steps:

[0094] S3.1: Update the state variables of the local digital twin based on traffic features.

[0095] The state variables include traffic flow, average speed, lane occupancy, and queue length. These state variables are obtained through statistical calculations of the extracted traffic features: traffic flow is obtained by counting the number of vehicles passing through per unit time; average speed is obtained by statistically calculating the speed of vehicles within the road segment; lane occupancy is calculated based on lane occupancy; and queue length is obtained by identifying queues of slow-moving vehicles. Through these calculations, vehicle-level micro-features are transformed into road segment-level macro-state variables, enabling real-time updates of the local digital twin's state.

[0096] S3.2: An incremental mapping strategy is adopted to transmit the state variables that have changed compared to the previous moment to the global digital twin in the cloud.

[0097] It should be noted that incremental transmission only uploads state variables that have changed beyond a threshold, significantly reducing cloud-edge communication overhead compared to full state transmission. State variable uploads are used for the synchronous update of the global digital twin, proceeding in parallel with the local inference of edge nodes.

[0098] S3.3: Run the timing prediction model at the edge node, and extrapolate the short-term traffic conditions based on the updated local digital twin.

[0099] The model input includes the current state and a sequence of historical states. The runtime time-series prediction model extrapolates traffic conditions for the next 5 to 15 minutes, and this model may include a Long Short-Term Memory (LSTM) network or a gated recurrent unit (GRU) network. The output is the predicted traffic conditions for each future time point, providing a basis for congestion identification and strategy generation.

[0100] In one embodiment, the extrapolation process considers the spatial correlation between road segments, and the extrapolation results include state prediction values ​​and confidence intervals.

[0101] S3.4: Generate local control strategies based on the simulation results and distribute them to the execution devices.

[0102] The local control strategies include signal timing adjustment, variable lane control, and local guidance information. Signal timing adjustment is based on projected future traffic flow at each intersection approach; when predicted traffic flow changes, the green light duration for each phase is reallocated to optimize traffic efficiency. Variable lane control is based on projected future traffic flow in each direction; when there is a difference in traffic flow between the two directions, the direction of use of the variable lanes is adjusted. Local guidance information is generated based on projected identification of future congested road sections and disseminated through variable message signs or in-vehicle terminals.

[0103] In one embodiment, the strategy generation considers execution constraints and timing, such as the minimum green light time for signal timing and the safe interval for changing lanes, and determines the timing for strategy issuance based on the predicted time lead. The generated control strategy is issued to the corresponding execution devices, including traffic signals, variable lane signs, and variable message signs, via a communication interface.

[0104] Through the S3 steps, real-time status updates, short-term traffic simulations, and predictive control strategy generation for the local digital twin were completed locally at the edge node. The incremental transmission strategy reduced cloud-edge communication overhead; by predicting future traffic state trends through short-term simulations at the edge, a shift from a passive response to a proactive prevention control mode was achieved, enabling intervention before congestion occurs and improving the predictability and effectiveness of traffic control; local simulations and decision-making at the edge avoided cloud-to-cloud round-trip delays, shortening control response time and providing technical support for intelligent monitoring and optimized control of traffic flow.

[0105] S4: Each edge node uploads model update information to the cloud according to a preset cycle. The cloud aggregates and processes the uploaded information and updates the global digital twin, then sends the updated model parameters to each edge node.

[0106] Specifically, S4 includes the following sub-steps:

[0107] S4.1: Each edge node periodically uploads model update information to the cloud.

[0108] It should be noted that, unlike the real-time upload of state variables in S3.2, the model update information uploaded in this step includes the model parameters and performance metrics of the local digital twin, used for global model aggregation and optimization in the cloud. The upload cycle is determined based on changes in model performance, typically on an hourly or daily basis.

[0109] S4.2: The cloud receives model update information from multiple edge nodes and calculates the aggregation weight by comprehensively considering the data quality, sample quantity, and road network topology correlation of each edge node.

[0110] Preferably, the cloud receives model update information uploaded by multiple edge nodes, including model gradients, the number of training samples, and data quality metrics. Based on the road network topology, the spatial correlation between each edge node and other nodes is calculated. The aggregate weight of each edge node is calculated by combining its data quality metrics, the number of training samples, and the spatial correlation. The aggregate weight is then normalized.

[0111] S4.3: The federated learning framework is used to perform weighted aggregation of model update information based on the aggregation weights.

[0112] S4.4: Differential privacy technology is used to process the aggregated model gradients and update the model parameters of the global digital twin.

[0113] S4.5: The updated model parameters are distributed to each edge node, and each edge node makes personalized adjustments to the model based on the global model parameters and local features.

[0114] Through the S4 steps, a cloud-edge collaborative model optimization mechanism was implemented without affecting the real-time inference and control of edge nodes. Edge nodes periodically upload model update information, and the cloud aggregates and optimizes the data and experience from multiple edge nodes to form a global model, which is then distributed to all edge nodes. This mechanism enables edge nodes to share learning results, avoids model bias caused by data limitations of a single node, and improves the inference accuracy and generalization ability of local digital twins. Simultaneously, differential privacy technology protects the data privacy of each edge node, enabling knowledge sharing and continuous model optimization while ensuring security.

[0115] S5: The cloud performs medium- and long-term traffic state simulations based on a global digital twin, generates a global optimization control strategy, and decomposes the global optimization control strategy into local sub-strategies and distributes them to the corresponding edge nodes.

[0116] Specifically, S5 includes the following steps:

[0117] S5.1: The cloud-based system uses a global digital twin to predict the future medium- and long-term traffic conditions of the entire road network using a traffic condition prediction model.

[0118] In one embodiment, a traffic state projection model is selected to project the traffic state of the entire road network for the next 30 to 60 minutes. The projection model includes a macroscopic traffic flow model or a deep neural network model. The macroscopic traffic flow model includes a cellular transport model or a macroscopic fundamental graph model. The deep neural network model includes a spatiotemporal graph convolutional network or a Transformer model.

[0119] Furthermore, the current traffic state of the global digital twin is used as the initial condition for the simulation. The simulation calculation is performed to output the predicted traffic state values ​​for each future time, forming a spatiotemporal evolution sequence, which provides a basis for subsequent cross-regional propagation analysis and strategy generation.

[0120] S5.2: Analyze the cross-regional traffic flow propagation characteristics during the simulation process, and identify cross-regional traffic flow propagation relationships and potential congestion diffusion paths.

[0121] Specifically, based on spatiotemporal evolution sequences, the traffic flow transmission relationship between adjacent areas is analyzed. By calculating the flow rate changes and speed gradients at regional boundary road segments, the direction and rate of traffic flow propagation are identified. When congestion in one area is detected spreading to adjacent areas, the congestion propagation path and estimated arrival time are marked.

[0122] In one embodiment, traffic wave theory is used to analyze the spatiotemporal propagation characteristics of congestion, calculate the congestion wave velocity and its impact range, and identify key bottleneck sections to provide key intervention targets for strategy generation.

[0123] S5.3: Based on the simulation results and the identified congestion propagation paths, generate multiple candidate schemes for global optimization control strategies.

[0124] The global optimization control strategy includes regional boundary signal coordination schemes, large-scale traffic guidance strategies, and emergency vehicle route planning.

[0125] Furthermore, for the identified congestion spread paths, regional boundary signal coordination schemes are generated. These schemes control cross-regional traffic flow by adjusting the signal timing at intersections in adjacent areas. Large-scale traffic guidance strategies are generated, disseminating route suggestions via variable message signs and navigation systems to guide vehicles to detour around congested areas. Emergency vehicle route planning is generated, prioritizing routes for special vehicles. Multiple schemes of varying intensities are generated for each strategy type, forming a candidate scheme set.

[0126] It should be noted that the number of candidate solutions is preferably 3 to 10, balancing simulation efficiency and solution diversity.

[0127] S5.4: Simulate the execution effect of each candidate scheme in the global digital twin environment, evaluate the degree of improvement in traffic capacity and the degree of delay reduction, and select the optimal strategy based on the evaluation results.

[0128] Furthermore, for each candidate solution, simulations are performed in a global digital twin to model the traffic state evolution after the strategy is implemented. The rate of change of average road network speed, the change in average vehicle delay time, and the percentage increase in road segment capacity before and after strategy implementation are calculated. Taking multiple evaluation indicators into account, a weighted scoring method is used to calculate the comprehensive score of each candidate solution. The optimal strategy is selected based on the comprehensive score.

[0129] In one embodiment, the simulation process considers the time lag effect of policy execution and driver response behavior. For candidate solutions with similar effects, the solution with lower implementation cost and lower execution complexity is preferred.

[0130] S5.5: Decompose the optimal strategy into local sub-strategies according to the local region boundary, determine the execution sequence of each sub-strategy, and distribute it to the corresponding edge nodes.

[0131] Furthermore, based on the division of local regions, the optimal strategy is decomposed into local sub-strategies for each region, each containing specific control instructions for that region. The execution sequence of each local sub-strategy is determined based on the estimated arrival time of congestion spread. For strategies requiring multi-region coordination, an execution synchronization mechanism is set up for each region. The decomposed local sub-strategies are then distributed to the corresponding edge nodes.

[0132] Ideally, S5 achieves cross-regional traffic flow collaborative control through cloud-based global optimization, overcoming the limitations of edge node visibility. Preventative control based on medium- to long-term simulations allows for intervention 15-60 minutes in advance, avoiding reactive delays. The strategy decomposition mechanism eliminates the need for edge nodes to perform repeated global optimization calculations, reducing policy conflicts and computational burden. Together with S3's short-term simulations, it forms a complementary system with dual time scales, balancing strategic control with real-time response.

[0133] S6: Edge nodes correct the local digital twin based on the deviation between the predicted value and the actual observed value, and feed the correction information back to the cloud for global digital twin optimization.

[0134] In one embodiment, edge nodes continuously monitor the actual traffic conditions after the control strategy is implemented. They calculate the deviation between the predicted values ​​from the local digital twin and the actual observed values, including traffic flow prediction errors and speed prediction errors. When the deviation meets a trigger condition, the edge nodes correct the parameters of the local digital twin model. The trigger condition includes the deviation exceeding a preset threshold and the duration exceeding a preset time. Model correction information and deviation statistics are uploaded to the cloud. The cloud analyzes the deviation distribution characteristics of multiple edge nodes, including: calculating the statistical characteristics of the deviations of each edge node, identifying spatial clustering patterns of the deviations, and analyzing the temporal evolution trend of the deviations. When the cloud determines that a systematic deviation exists, it updates the model parameters of the global digital twin.

[0135] This embodiment also provides a digital twin-based intelligent traffic flow monitoring system, the module connection diagram of which is shown below. Figure 4 As shown, the system includes:

[0136] The digital twin construction module is used to obtain basic configuration information, divide the road network into multiple local areas, establish local digital twins at each edge node, and establish a global digital twin in the cloud.

[0137] The multi-source data fusion module is used to receive multi-source traffic sensing data, perform spatiotemporal alignment and fusion processing on the multi-source traffic sensing data, and extract traffic features.

[0138] The local simulation control module is used to update the state variables of the local digital twin based on traffic characteristics, perform short-term traffic state simulation based on the updated local digital twin, and generate local control strategies.

[0139] The cloud-edge collaborative optimization module is used to upload model update information to the cloud at preset cycles. The cloud aggregates and processes the uploaded information and updates the global digital twin, and then sends the updated model parameters to each edge node.

[0140] The global simulation control module is used to perform medium- and long-term traffic state simulations based on the global digital twin, generate a global optimization control strategy, and decompose the global optimization control strategy into local sub-strategies and distribute them to the corresponding edge nodes.

[0141] The adaptive correction module is used to correct the local digital twin based on the deviation between the predicted value and the actual observed value, and feeds the correction information back to the cloud for global digital twin optimization.

[0142] This embodiment also provides a computer device applicable to the intelligent traffic flow monitoring method based on digital twins, comprising: a memory and a processor; the memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions to implement the intelligent traffic flow monitoring method based on digital twins as proposed in the above embodiment.

[0143] The computer device can be a terminal, comprising a processor, memory, communication interface, display screen, and input devices connected via a system bus. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The communication interface is used for wired or wireless communication with external terminals; wireless communication can be achieved through Wi-Fi, carrier networks, NFC (Near Field Communication), or other technologies. The display screen can be an LCD screen or an e-ink screen. The input devices can be a touch layer covering the display screen, buttons, a trackball, or a touchpad on the computer device's casing, or an external keyboard, touchpad, or mouse.

[0144] This embodiment also provides a storage medium storing a computer program, which, when executed by a processor, implements the intelligent traffic flow monitoring method for digital twins as proposed in the above embodiments. The storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as Static Random Access Memory (SRAM), Electrically Erasable Programmable Read-Only Memory (EEPROM), Erasable Programmable Read Only Memory (EPROM), Programmable Red-Only Memory (PROM), Read-Only Memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk.

[0145] In summary, this invention innovatively constructs a distributed two-level digital twin system, achieving deep collaboration between real-time edge control and global optimization in the cloud. Edge nodes handle multi-source fusion and short-term extrapolation, while the cloud is responsible for medium- and long-term prediction and cross-regional policy coordination, forming a complementary control framework across multiple spatiotemporal scales. A cloud-edge collaborative update mechanism based on federated learning, combined with differential privacy protection, ensures data security while reducing communication overhead. Deviation-driven adaptive correction endows the system with continuous evolution capabilities, breaking through the limitations of traditional static models and significantly improving prediction accuracy and control performance in complex traffic scenarios.

[0146] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A digital twin-based intelligent traffic flow monitoring method, characterized in that: The method comprises the following steps: acquiring basic configuration information, dividing the road network into multiple local areas, establishing a local digital twin at each edge node, and establishing a global digital twin in the cloud; each edge node receives multi-source traffic perception data within its jurisdiction, performs spatio-temporal alignment and fusion processing on the multi-source traffic perception data, and extracts traffic features; based on the traffic features, update the state variables of the local digital twin, and based on the updated local digital twin, perform short-term traffic state inference to generate a local control strategy; each edge node uploads model update information to the cloud at a preset period, the cloud aggregates the uploaded information and updates the global digital twin, and the updated model parameters are distributed to each edge node; the cloud performs medium and long term traffic state inference based on the global digital twin, generates a global optimization control strategy, and distributes the local sub-strategies of the global optimization control strategy to the corresponding edge nodes; the edge nodes correct the local digital twin based on the deviation between the predicted value and the actual observed value, and feed back the correction information to the cloud for global digital twin optimization. 2.The digital-twin-based intelligent traffic flow monitoring method of claim 1, wherein: The method comprises the following steps: according to the road network topology information, node communication range and node computing capacity, the road network is divided into multiple local areas by using a graph clustering algorithm, and a mapping relationship between the areas and the edge nodes is established; a local digital twin is established at each edge node, which includes a local road network model, a local traffic flow model and a set of state variables; a global digital twin is established in the cloud, which includes a global road network model, a global traffic flow model and an optimization engine. 3.The digital-twin based intelligent traffic flow monitoring method of claim 1, wherein: The method comprises the following steps: each edge computing node receives at least two types of traffic perception data within its jurisdiction; spatio-temporal alignment is performed on the multi-source traffic perception data and unified to the road network coordinate system; data fusion is completed locally at the edge node and abnormal data is removed; traffic features are extracted, including vehicle position, speed, acceleration, vehicle type classification and lane occupancy information.

4. The digital-twin based intelligent traffic flow monitoring method of claim 1, wherein: The method comprises the following steps: based on the traffic features, update the state variables of the local digital twin, including road segment flow, average speed, lane occupancy rate and queue length; an incremental mapping strategy is used to transmit the state variables that have changed compared with the previous time to the global digital twin in the cloud; a time series prediction model is run at the edge node, and the future short-term traffic state is inferred based on the updated local digital twin; according to the inference result, a local control strategy is generated and distributed to the execution device, including signal timing adjustment, variable lane control and local guidance information. 5.The digital-twin based intelligent traffic flow monitoring method of claim 1, wherein: The method comprises the following steps: the cloud receives model update information from multiple edge nodes, and calculates aggregation weights by comprehensively considering the data quality, sample quantity and road network topology correlation of each edge node; using a federated learning framework, the model update information is weighted and aggregated according to the aggregation weights; the aggregated model gradient is processed using differential privacy technology to update the model parameters of the global digital twin; The updated model parameters are distributed to each edge node, and each edge node performs model individualization adjustment based on the global model parameters and local features.

6. The digital-twin based intelligent traffic flow monitoring method of claim 1, wherein: The cloud side performs mid-to-long-term traffic state deduction based on the global digital twin, including: The cloud side performs mid-to-long-term traffic state deduction based on the global digital twin, including: The cloud side performs mid-to-long-term traffic state deduction based on the global digital twin, including: The cloud side performs mid-to-long-term traffic state deduction based on the global digital twin, including: According to the deduction result and the identified congestion diffusion path, a plurality of global optimization control strategy candidate schemes are generated; The execution effect of each candidate scheme is simulated in the global digital twin environment, and the improvement degree of traffic capacity and the reduction degree of delay are evaluated, and the optimal strategy is selected according to the evaluation result; 7. The digital-twin based intelligent traffic flow monitoring method of claim 1, wherein: The optimal strategy is decomposed into local sub-strategies according to the local area boundary, and the execution time sequence of each sub-strategy is determined and distributed to the corresponding edge node. The correction of the local digital twin includes: The edge node continuously monitors the actual traffic state after the execution of the control strategy; The deviation between the prediction value and the actual observation value of the local digital twin is calculated; When the deviation meets the triggering condition, the edge node corrects the model parameters of the local digital twin; The model correction information and deviation statistical data are uploaded to the cloud side, and the cloud side analyzes the deviation distribution characteristics of multiple edge nodes; 8. A digital-twin based intelligent traffic flow monitoring system based on the digital-twin based intelligent traffic flow monitoring method of any one of claims 1-7. When the cloud side judges that there is a systematic deviation, the model parameters of the global digital twin are updated. It includes: A digital twin construction module is used to obtain basic configuration information, divide the road network into multiple local areas, establish a local digital twin in each edge node, and establish a global digital twin in the cloud side; A multi-source data fusion module is used to receive multi-source traffic perception data, perform spatio-temporal alignment and fusion processing on the multi-source traffic perception data, and extract traffic features; A local deduction control module is used to update the state variables of the local digital twin based on the traffic features, perform short-term traffic state deduction based on the updated local digital twin, and generate a local control strategy; A cloud-edge collaborative optimization module is used to upload model update information to the cloud side at a preset period, the cloud side aggregates and processes the uploaded information and updates the global digital twin, and distributes the updated model parameters to each edge node; A global deduction control module is used to perform mid-to-long-term traffic state deduction based on the global digital twin, generate a global optimization control strategy, and decompose the global optimization control strategy into local sub-strategies and distribute them to the corresponding edge node; 9.A computer device, comprising a memory and a processor, wherein the memory stores a computer program, and the computer device is characterized in that: An adaptive correction module is used to correct the local digital twin based on the deviation between the prediction value and the actual observation value, and feed back the correction information to the cloud side for global digital twin optimization.

10. A computer readable storage medium having stored thereon a computer program, characterized in that: The processor executes the computer program to realize the steps of the digital twin intelligent traffic flow monitoring method of any one of claims 1-7. The computer program is executed by the processor to realize the steps of the digital twin intelligent traffic flow monitoring method of any one of claims 1-7.

Images