Intelligent city traffic dynamic optimization system and method based on digital twinning

Abstract

The application relates to the technical field of smart city traffic and discloses a smart city traffic dynamic optimization system and method based on digital twinning. The system acquires multi-dimensional data of a city traffic network from multiple heterogeneous data sources through a traffic multi-dimensional data acquisition module and integrates the data into a traffic-related data warehouse; a traffic digital twinning model construction module extracts features from the data warehouse to generate a traffic-related feature matrix, and accordingly constructs a digital twinning traffic dynamic model to output a theoretical traffic state value; a traffic flow graph construction module determines a dimension linkage graph of each dimension and constructs a traffic flow linkage graph; a traffic core feature screening module screens a traffic core feature sequence based on the graph; a traffic multi-dimensional optimization analysis module performs multi-dimensional difference analysis on the theoretical traffic state value and real-time measured traffic data, and generates a regional difference coefficient matrix in combination with the core feature sequence; and a traffic event correlation positioning module can realize accurate management and dynamic optimization of city traffic.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of smart city traffic, in particular to a smart city traffic dynamic optimization system and method based on digital twinning. BACKGROUND

[0002] With the acceleration of urbanization, the urban traffic network is becoming increasingly complex, and problems such as the surge of traffic flow and the sudden occurrence of traffic events have brought severe challenges to urban traffic management. Currently, urban traffic management relies mainly on traditional data collection and analysis methods, which have obvious limitations in the comprehensiveness and timeliness of data acquisition. Most traffic data collection methods can only collect data from a single data source, such as road monitoring cameras that can only obtain image data of a specific area and traffic induction coils that can only collect vehicle passing number data, making it difficult to effectively integrate multiple heterogeneous data sources, resulting in serious data fragmentation and the inability to build a data set that fully reflects the state of the urban traffic network.

[0003] In terms of traffic state analysis and simulation, existing technologies mostly use static models to analyze traffic conditions, which cannot capture dynamic changes in the traffic network in real time, such as fluctuations in vehicle speed and temporary changes in road capacity, resulting in a large deviation between the output traffic state assessment results and the actual traffic conditions, making it difficult to meet the real-time and accuracy requirements of traffic management. Meanwhile, in the traffic feature analysis process, existing technologies lack in-depth exploration of the correlation between traffic-related features, making it difficult to clearly sort out the internal relationships between different traffic feature dimensions, resulting in the need to rely on experience to select traffic core features, which makes it difficult to accurately identify core feature sequences that have a significant impact on traffic conditions, thereby affecting the relevance and effectiveness of subsequent traffic optimization analysis.

[0004] In terms of abnormal traffic event processing, existing technologies usually rely on manual inspection or single monitoring device alarms to detect problems after the occurrence of traffic events, which not only slows down the response speed, but also lacks systematic analysis methods that combine road network topology and traffic node location information to locate the propagation path and impact range of abnormal traffic events, making it difficult to quickly generate an accurate abnormal event impact area distribution map, which prevents traffic management departments from taking effective control measures in a timely manner, leading to the spread of traffic congestion and further exacerbating urban traffic operation pressure. SUMMARY

[0005] The present application aims to provide a smart city traffic dynamic optimization system based on digital twinning to solve the problems raised in the background.

[0006] To achieve the above-mentioned purpose, the present application provides a smart city traffic dynamic optimization system based on digital twinning, which comprises:

[0007] a traffic multidimensional data collection module configured to acquire multidimensional data of a city traffic network from a plurality of heterogeneous data sources, perform data integration processing on the multidimensional data, and obtain a traffic-related data warehouse;

[0008] a traffic digital twin model construction module configured to perform feature extraction on the traffic-related data warehouse, generate a traffic-related feature matrix, construct a digital twin traffic dynamic model based on the traffic-related feature matrix, and output a theoretical traffic state value through the digital twin traffic dynamic model;

[0009] a traffic flow graph construction module configured to determine a dimension link graph of each dimension in the traffic-related feature matrix, and construct a flow link graph atlas related to city traffic through the dimension link graphs of each dimension;

[0010] a traffic core feature screening module configured to screen a traffic core feature sequence based on the flow link graph atlas;

[0011] a traffic multidimensional optimization analysis module configured to perform multidimensional difference analysis on the theoretical traffic state value and real-time collected measured traffic data, combine the traffic core feature sequence, and generate a difference coefficient matrix at a regional level;

[0012] a traffic event correlation positioning module configured to input the difference coefficient matrix into a spatial topology analysis network, combine road network topology parameters and traffic node location information, generate an abnormal probability thermal distribution map of an abnormal traffic event propagation path, and position an abnormal traffic physical area.

[0013] Preferably, the traffic multidimensional data collection module specifically includes:

[0014] a data preprocessing unit configured to perform normalization processing on the multidimensional data to eliminate data dimension differences;

[0015] a semantic mapping unit configured to perform semantic mapping conversion on the preprocessed multidimensional data to generate standardized traffic data of each dimension;

[0016] a data warehouse construction unit configured to aggregate the standardized traffic data of all dimensions to construct the traffic-related data warehouse.

[0017] Preferably, the traffic digital twin model construction module specifically includes:

[0018] a historical data feature mining unit configured to perform multidimensional decomposition processing on historical traffic operation data to extract time series features and spatial distribution features;

[0019] The dynamic model construction unit inputs the processed historical traffic operation data into a time series prediction network to generate a basic traffic state prediction value, embeds a feature compensation network with a spatial attention mechanism to correct prediction deviations caused by environmental disturbances, and dynamically adjusts prediction weights based on a real-time traffic data event feature adapter.

[0020] The theoretical value generation unit inputs real-time traffic data into the digital twin traffic dynamic model to obtain a theoretical traffic state value.

[0021] Preferably, the traffic flow graph construction module specifically includes:

[0022] The dimension linkage graph generation unit selects a dimension in the traffic-related feature matrix, obtains a traffic feature vector of the selected dimension, determines the feature linkage degree between each traffic-related feature in the traffic feature vector, and constructs a dimension linkage graph of the selected dimension according to the feature linkage degree.

[0023] The flow linkage graph integration unit identifies key traffic-related features of each dimension, determines the cross-dimension feature linkage degree between the key traffic-related features, connects the dimension linkage graphs of each dimension according to the cross-dimension feature linkage degree, and generates a flow linkage graph related to urban traffic.

[0024] Preferably, the traffic multi-dimensional optimization analysis module specifically includes:

[0025] The time-domain cumulative deviation calculation unit compares the theoretical traffic state value and the measured traffic data in a preset time window, aligns the non-synchronous sampled traffic data sequences using a sequence alignment algorithm, calculates the cumulative deviation amount in each time window, and generates a time-domain deviation vector.

[0026] The spatial distribution offset detection unit performs spatial domain decomposition processing on the theoretical traffic state value and the measured traffic data, calculates the district energy distribution ratio, extracts the spatial offset index, and constructs a spatial offset vector.

[0027] The event sequence matching degree evaluation unit matches the event sequence distribution of the theoretical traffic state value and the measured traffic data based on a pattern matching algorithm, calculates the time synchronization error of the event trigger point, quantifies the divergence value of the event interval distribution, and generates an event matching degree vector.

[0028] The difference coefficient matrix generation unit performs tensor fusion operation on the time-domain deviation vector, the spatial offset vector, and the event matching degree vector, performs standardization processing through feature weight weighting, and outputs a three-order difference coefficient matrix with dimensions of region number multiplied by time stamp multiplied by difference type.

[0029] Preferably, the traffic event correlation positioning module specifically includes:

[0030] The road topology modeling unit constructs a road network node connection relationship topology graph according to traffic node position information, labels path impedance parameters between nodes, superimposes reverse propagation constraint conditions of traffic event trigger points in the topology graph, and generates a traffic topology model including an impedance matrix and a node admittance matrix.

[0031] The anomaly propagation deduction unit maps the difference coefficient matrix to corresponding nodes of the traffic topology model, performs an anomaly propagation simulation operation based on a graph structure network, and the anomaly propagation simulation operation includes calculating an anomaly event attenuation factor according to path impedance parameters, capturing cross-region anomaly correlation features through a multi-head attention mechanism, and simulating an abnormal event diffusion trajectory in the topology network by using a random path simulation algorithm.

[0032] The probability distribution generation unit calculates anomaly event residence probability values by combining path impedance parameters, and generates an anomaly probability heat distribution graph covering the entire network.

[0033] The physical region demarcation unit performs a clustering analysis operation on the anomaly probability heat distribution graph, identifies anomaly probability aggregation areas, and demarcates abnormal traffic physical boundaries according to traffic node positions and road topology connection relationships.

[0034] Preferably, the system further includes an adaptive traffic control strategy generation module configured according to the anomaly probability heat distribution graph.

[0035] The adaptive traffic control strategy generation module specifically includes a monitoring mode switching unit that issues a mode switching instruction to a traffic monitoring terminal belonging to a region whose anomaly probability value exceeds a preset anomaly residence probability threshold in the anomaly probability heat distribution graph, performs an operation of increasing a traffic data sampling frequency to a multiple of an original frequency, enables an event sequence real-time tracking mode, and deploys a transient event detector to capture traffic state mutation fragments.

[0036] The traffic flow disturbance test unit applies a multi-frequency traffic flow disturbance test operation to adjacent road nodes with the largest anomaly probability gradient change in the anomaly probability heat distribution graph.

[0037] Preferably, the system further includes a traffic state early warning module that performs a traffic state early warning operation based on the anomaly probability heat distribution graph.

[0038] The traffic data transmission module specifically includes a position information acquisition unit that acquires position information of intelligent traffic devices and deployment position information of wireless monitoring elements in a city traffic network.

[0039] The data transmission link establishment unit establishes a communication transmission link based on the position information, including determining an intermediate forwarding node according to a distance threshold, and generating a transmission path from a source node to a target node.

[0040] Preferably, the system further comprises a traffic data transmission module for establishing a communication transmission link based on traffic node position information, for transmitting control instructions and monitoring data.

[0041] The traffic state early warning module specifically comprises a core feature extraction unit for extracting a key abnormal feature vector from the abnormal probability heat map.

[0042] The early warning model execution unit inputs the key abnormal feature vector into a state detection model to generate a traffic state early warning report.

[0043] Preferably, the application further comprises a smart city traffic dynamic optimization method based on digital twinning, including all modules and method processes of the smart city traffic dynamic optimization system based on digital twinning.

[0044] Compared with the prior art, the application has the following beneficial effects:

[0045] The smart city traffic dynamic optimization system based on digital twinning can obtain multi-dimensional data of the city traffic network from multiple heterogeneous data sources at the data acquisition and integration level, and effectively integrates and processes the data to form a traffic-related data warehouse, breaking the limitations of single data source and data fragmentation in the traditional traffic data acquisition process, making the obtained traffic data more comprehensive and complete, and more fully reflecting the overall operation of the city traffic network, providing a more abundant data basis for subsequent traffic analysis and optimization.

[0046] In terms of traffic model construction, the traffic-related feature matrix is generated by feature extraction on the traffic-related data warehouse, and the digital twinning traffic dynamic model is constructed based on the matrix, which can capture the dynamic changes in the traffic network in real time. Compared with the traditional static traffic model, the theoretical traffic state value output by the digital twinning traffic dynamic model is more in line with the actual traffic conditions, and the operation law of the traffic network can be more accurately simulated, providing a reliable reference for traffic management departments to understand the traffic operation trend.

[0047] The traffic flow graph construction module constructs the city traffic-related flow link graph by determining the dimension link graph of each dimension in the traffic-related feature matrix, and clearly sorts out the internal correlation between different traffic feature dimensions, so that traffic management personnel can intuitively understand the mutual influence between traffic features, avoiding the problem of fuzzy understanding of traffic feature correlation in traditional technology, and laying a clear relationship foundation for subsequent core feature screening.

[0048] The traffic core feature screening module screens out a traffic core feature sequence based on a process link map, breaks away from the limitation of screening core features by relying on experience judgment in traditional technologies, can accurately identify core features that have a significant impact on traffic state, enables subsequent traffic optimization analysis to focus on key influencing factors, avoids interference of invalid features on the analysis process, and improves the pertinence of traffic optimization analysis.

[0049] The traffic multi-dimensional optimization analysis module performs multi-dimensional difference analysis operation on the theoretical traffic state value and the real-time collected measured traffic data, and generates a regional level difference coefficient matrix in combination with the traffic core feature sequence, can comprehensively and accurately capture the difference between the actual running condition and the theoretical simulation condition of different regional traffic, and compared with traditional single-dimensional difference analysis, the multi-dimensional difference analysis is more in-depth and can reflect the problems in traffic operation from multiple angles.

[0050] The traffic event correlation positioning module inputs the difference coefficient matrix into a spatial topology analysis network, generates an abnormal probability heat distribution map of an abnormal traffic event propagation path in combination with road network topology parameters and traffic node position information, and then locates an abnormal traffic physical area, can quickly and accurately determine the influence range and propagation path of an abnormal traffic event, changes the situation that the abnormal traffic event is discovered late and positioned inaccurately in traditional technologies, enables the traffic management department to timely master the specific situation of the abnormal traffic event, so that corresponding control measures can be taken quickly, effectively avoids the spread of traffic congestion, and guarantees the smooth operation of the urban traffic network. BRIEF DESCRIPTION OF DRAWINGS

[0051] Figure 1 is a timing diagram of the smart city traffic dynamic optimization system based on digital twinning described in the application;

[0052] Figure 2 is a working principle diagram of the traffic multi-dimensional data collection module;

[0053] Figure 3 is a working principle diagram of the traffic flow map construction module;

[0054] Figure 4 is a working principle diagram of the adaptive traffic control strategy generation module. DETAILED DESCRIPTION

[0055] The technical solutions in the embodiments of the application will be described clearly and completely below with reference to the drawings in the embodiments of the application. Obviously, the described embodiments are only part of the embodiments of the application, rather than all the embodiments of the application. Based on the embodiments in the application, all other embodiments obtained by those skilled in the art without creative labor fall within the protection scope of the application.

[0056] Referring to Figure 1 The application provides a smart city traffic dynamic optimization system and method based on digital twinning, the system comprising:

[0057] A traffic multi-dimensional data acquisition module acquires multi-dimensional data of a city traffic network from multiple heterogeneous data sources, the multi-dimensional data including traffic flow data, vehicle position data, signal light state data, road sensor data, and environmental meteorological data. The multi-dimensional data is processed through data integration to form a traffic-related data warehouse. A traffic digital twinning model construction module performs feature extraction operations on the traffic-related data warehouse to generate a traffic-related feature matrix, and constructs a digital twinning traffic dynamic model based on the matrix, which outputs a theoretical traffic state value. A traffic flow graph construction module determines a dimension link graph for each dimension in the traffic-related feature matrix, and constructs a flow link graph atlas related to city traffic through the dimension link graphs of each dimension. A traffic core feature screening module screens out a traffic core feature sequence based on the flow link graph atlas. A traffic multi-dimensional optimization analysis module performs multi-dimensional difference analysis operations on the theoretical traffic state value and real-time collected measured traffic data, and generates a regional-level difference coefficient matrix in combination with the traffic core feature sequence. A traffic event correlation positioning module inputs the difference coefficient matrix into a spatial topology analysis network, generates an abnormal probability heat distribution map of an abnormal traffic event propagation path in combination with road network topology parameters and traffic node position information, and locates an abnormal traffic physical area.

[0058] Example 1: Referring to Figure 2, involves the specific operation process of the traffic multi-dimensional data acquisition module and the traffic digital twin model construction module. In the traffic multi-dimensional data acquisition module, the system obtains multi-dimensional data of the urban traffic network from multiple heterogeneous data sources. These data sources include loop detectors, microwave radars, high-definition cameras, floating car GPS data, phase state data of traffic signal control systems, and visibility and precipitation intensity information provided by the meteorological department deployed on the road. These raw data have different sampling frequencies, data formats, and dimensions, and need to be uniformly processed. The data preprocessing unit first performs normalization operation, and adopts corresponding standardization strategies for different types of numerical features. Traffic flow data is usually in units of vehicles / hour, which is mapped to the interval of zero to one through linear transformation. Vehicle position data uses latitude and longitude coordinates, which are converted to coordinate values under standard normal distribution through Gaussian normalization method. Signal light state is a categorical variable, which is converted to binary vector representation through one-hot encoding. Road sensor collected vehicle speed and occupancy data have different orders of magnitude, and are respectively processed by Z-score standardization to make their mean value zero and standard deviation one. Environmental meteorological data such as visibility and precipitation may present skewness characteristics, and are processed by quantile normalization method to make them conform to uniform distribution. After these processes, the data in each dimension eliminates the dimensional difference and has comparable numerical characteristics.

[0059] The semantic mapping unit integrates the preprocessed multi-dimensional data at the semantic level. This unit maintains a traffic domain ontology library, which defines the concepts, attributes, and mutual relationships of various traffic entities. The field names in the original data may differ due to different data sources, for example, some systems mark vehicle speed as "speed", while others use "velocity". Through semantic mapping, all these homonyms and synonyms are converted to unified standardized identifiers. The mapping process is based on a pre-defined rule base and semantic similarity calculation to match the input data fields with the concepts in the ontology library. For example, all fields representing vehicle speed are mapped to "speed dimension", and all fields representing traffic capacity are mapped to "flow dimension". This step ensures the consistency of data semantics in subsequent processing.

[0060] The data warehouse construction unit is responsible for aggregating and storing the multi-dimensional data that has been semantically standardized. Due to the possible lack of synchronization of timestamps from different data sources and the differences in spatial reference systems, the unit uses a spatio-temporal alignment algorithm for processing. Temporal alignment uses UTC time as a reference to unify the timestamps of all data to the same time zone and performs interpolation on non-uniformly sampled data to generate equally spaced time series. Spatial alignment uses a geographic grid division method to divide urban areas into regular spatial grids, each grid is assigned a unique identifier. All data with spatial location attributes are mapped to the corresponding grid. Aggregation operations combine multi-dimensional data within the same time interval and spatial grid into a complete record. These records are organized in chronological order and stored in a distributed columnar database, forming a traffic-related data warehouse. The data warehouse uses a columnar storage structure to improve query efficiency on large-scale data and provides a complete data foundation for subsequent analysis.

[0061] In the traffic digital twin model construction module, the historical data feature mining unit performs in-depth analysis on the historical traffic operation data stored in the data warehouse. This unit uses multi-dimensional decomposition technology to decompose complex traffic data into different component parts. Time series feature extraction focuses on the periodicity, trendiness, and randomness of traffic data. By analyzing historical data, it identifies daily and weekly traffic flow variation patterns and extracts time series features including autocorrelation and partial autocorrelation. Spatial distribution feature analysis focuses on the distribution of traffic parameters in geographical space, calculates the spatial correlation of traffic states in each region, and identifies the spatial propagation pattern of traffic flow. These features together constitute the spatio-temporal characteristics of the traffic system.

[0062] The dynamic model construction unit constructs a digital twin traffic dynamic model based on the extracted features. This unit uses a long short-term memory network as the basic architecture, with the input layer receiving historical traffic state data including traffic flow, average speed, occupancy rate, and other parameters at multiple time steps. The network captures the long-term dependencies of traffic data through memory units and outputs traffic state prediction values for future time periods. Considering the influence of environmental factors on traffic state, the model introduces a spatial attention mechanism that can dynamically adjust the weight distribution of each spatial region based on external factors such as weather conditions and special events. For example, in low-visibility weather conditions, the model increases the feature weight of related regions to more accurately reflect the impact of environmental factors on traffic state. The event feature adapter component is responsible for processing real-time traffic data streams and continuously updating model parameters through an online learning mechanism, enabling the model to adapt to the dynamic changes of the traffic system.

[0063] The theoretical value generation unit inputs the real-time collected traffic data into the trained digital twin traffic dynamic model. The input data undergoes the same preprocessing and feature extraction process as the historical data and is converted into an input format acceptable by the model. The model calculates the theoretical traffic state values of each region in the future period of time according to the current traffic state and historical regularity. These theoretical values include predicted traffic flow distribution, average travel speed, road congestion index, and other key indicators, forming a digital mapping of the city's traffic system operation. The comparison between the theoretical values and the actual measured values provides a benchmark reference for subsequent anomaly detection and state evaluation.

[0064] The entire implementation process embodies the complete chain from raw data collection to advanced model construction. Through multi-level data processing and model calculation, digital representation of the city's traffic system is achieved. The standardized data interface connects each link, ensuring the coherence of the data processing process and the coordinated operation of the entire system.

[0065] Example 2: Refer to Figure 3 , which involves the specific operation process of the traffic flow graph construction module and the traffic multi-dimensional optimization analysis module. In the traffic flow graph construction module, the dimension linkage graph generation unit first processes the traffic-related feature matrix obtained after feature extraction. This matrix contains multiple dimensions of traffic feature data, such as traffic flow dimension, average speed dimension, travel time dimension, and environmental meteorological dimension. For a selected dimension, the unit extracts all feature vectors in that dimension. Taking the traffic flow dimension as an example, the feature vectors may include traffic values of different time granularities, traffic change rates, traffic stability indicators, and other specific features. The unit calculates the correlation between these features and analyzes the linear or nonlinear relationship between them using a statistical correlation-based measurement method. According to the calculated correlation values, the linkage graph structure of the dimension is constructed. The nodes in the graph represent the features, and the edges represent the correlation between the features. The weight of the edge reflects the correlation strength. This linkage graph can intuitively show the internal relationship between different features in the same dimension.

[0066] The flow linkage map integration unit is responsible for integrating linkage maps of various dimensions into a unified flow linkage map. The unit first identifies key features in each dimension, which are typically representative or highly important features in that dimension. For example, in the traffic flow dimension, peak hour flow can be identified as a key feature; in the average speed dimension, spatial average speed can be identified as a key feature. The unit then calculates the cross-dimension correlation degree between these key features, analyzing the mutual relationship between different dimension features. For example, analyzing the correlation between peak hour flow and spatial average speed, or the correlation between travel time and visibility. Based on these cross-dimension correlation degree values, the unit connects the originally independent dimension linkage maps to form a comprehensive, multi-dimensional flow linkage map. This map can reflect the complex relationship network between different dimension features in the traffic system, providing a structured representation for understanding the operation mechanism of the traffic system.

[0067] In the traffic multi-dimensional optimization analysis module, the time domain cumulative bias calculation unit handles the time sequence comparison between theoretical traffic state values and measured traffic data. Since the theoretical values and measured values may come from devices with different sampling frequencies, the unit first uses a sequence alignment algorithm to perform time synchronization processing on the two sets of data. This algorithm can handle non-uniformly sampled data sequences and find corresponding time points in the two sets of data. The unit sets a sliding time window, and the window length is determined according to the characteristics of traffic flow. Within each time window, the cumulative bias between the theoretical values and the measured values is calculated. The bias calculation considers multiple traffic parameters, including flow, speed, density, etc. Through the sliding window method, the unit generates a time domain bias vector, which records the differences between the theoretical values and the measured values in each time period.

[0068] The spatial distribution bias detection unit focuses on analyzing the differences in spatial distribution between the theoretical values and the measured values. The unit divides the urban traffic network into multiple spatial regions and processes each region separately. Using spatial domain decomposition technology, the traffic state data is decomposed into components of different spatial scales. The energy distribution ratio of the theoretical values and the measured values in each region is calculated, which reflects the differences in spatial energy distribution between the two. At the same time, the unit extracts the spatial bias index, which quantifies the degree of deviation between the theoretical prediction and the actual measurement in spatial distribution. Through these calculations, the unit constructs a spatial bias vector, which describes the matching situation of the theoretical values and the measured values in each spatial region.

[0069] The event sequence matching degree evaluation unit handles the matching problem of traffic event sequences, which include traffic accidents, traffic controls, special activities, and other events that affect traffic status. The unit uses a pattern matching algorithm to compare the theoretically predicted event sequence distribution with the actually observed event sequence. The algorithm calculates the time synchronization error of event trigger points, which is the difference between the theoretically predicted event occurrence time and the actual event occurrence time. At the same time, the unit quantifies the difference degree of event interval distribution, analyzing the matching situation of the theoretically predicted and actual situation in the event occurrence interval. Through these analyses, the unit generates an event matching degree vector, which reflects the degree of agreement between the theoretically predicted and actual situation in the event sequence.

[0070] The difference coefficient matrix generation unit is responsible for integrating the above three vectors into a comprehensive difference representation. The unit uses tensor fusion technology to combine the time domain bias vector, spatial offset vector, and event matching degree vector into a three-dimensional data structure. The three dimensions of this three-dimensional structure correspond to the region number, timestamp, and difference type, respectively. During the fusion process, the unit considers the importance differences of different features and uses a weighted scheme based on feature weights. The weights of each feature are determined by analyzing their importance in the traffic system. Finally, the unit standardizes the fused data to make data of different scales comparable, and outputs the final three-order difference coefficient matrix. This matrix comprehensively reflects the differences between the theoretically predicted traffic status and the actually observed traffic status in each region, each time point, and each difference type, providing a data basis for subsequent anomaly detection and event positioning.

[0071] The entire implementation process embodies the complete flow from feature relationship mining to state difference analysis, and through multi-angle and multi-level comparison and analysis, it realizes the comprehensive evaluation of the accuracy of the digital twin model. The processing links are connected through data flow, forming a complete processing chain, which provides technical support for the state monitoring and anomaly detection of the traffic system.

[0072] Embodiment 3: The traffic incident correlation positioning module receives the difference coefficient matrix from the prelude module, combines with the road network topology information, and realizes the accurate positioning and propagation analysis of abnormal traffic incidents. The road topology modeling unit first processes the traffic node location information, including road intersections, important road monitoring points, traffic hubs, and other key locations. The unit constructs a network topology graph based on the actual connection relationship between nodes, where nodes represent physical locations and edges represent road connections. Each edge is labeled with path impedance parameters, which comprehensively reflect the traffic characteristics of the road, including travel time, traffic capacity, turning restrictions, and other factors. The calculation of impedance values considers multiple dimensions, such as road grade, lane number, design speed, and other static attributes, as well as real-time traffic state and other dynamic information. The unit superimposes the reverse propagation constraint conditions of traffic incident trigger points in the topology graph, which are based on historical event data analysis and specify the possible direction and intensity limit of abnormal event propagation in the network. By integrating node connection relationships, path impedance parameters, and propagation constraint conditions, the unit generates a complete traffic topology model, which includes impedance matrix and node admittance matrix, two core mathematical representations. The impedance matrix quantitatively describes the ease of traffic between nodes, and the node admittance matrix reflects the traffic capacity characteristics of the node itself.

[0073] The abnormal propagation deduction unit maps the difference coefficient matrix to the corresponding nodes of the traffic topology model. The mapping process is based on spatial location matching, which associates the difference coefficient of each region with the nearest road node. The unit performs abnormal propagation simulation operations based on graph structure networks, and the simulation process includes multiple calculation steps. First, calculate the abnormal event attenuation factor based on the path impedance parameter, which represents the intensity attenuation degree of the abnormal event in the propagation process. The calculation of the attenuation factor uses the following formula:

[0074]

[0075] Where: represents the abnormal attenuation factor from node i to node j, represents the path impedance value from node i to node j, is the global attenuation coefficient of the network, reflecting the overall attenuation characteristics of abnormal propagation. The unit captures cross-regional abnormal correlation features through a multi-head attention mechanism, which can simultaneously focus on the abnormal correlation of multiple spatial locations and identify potential paths of abnormal propagation. The calculation of attention weights considers factors such as spatial distance between nodes, traffic connection strength, and historical propagation patterns. The unit uses a random path simulation algorithm to simulate the diffusion trajectory of abnormal events in the topology network, which is based on the Monte Carlo method to generate a large number of possible propagation paths, each representing a possible propagation way of abnormal events. The simulation process considers the influence of road impedance, with lower propagation probability on high impedance paths, while considering the cumulative effect and interaction of abnormal events.

[0076] The probability distribution generation unit statistically analyzes the frequency of anomalous events on each path during simulated propagation. The unit aggregates and analyzes multiple simulation results to calculate the frequency of anomalous events for each node and path. Combining path impedance parameters, the unit calculates the anomalous event dwell probability, which comprehensively reflects the likelihood of an anomaly persisting at a specific location. The probability calculation considers the attenuation characteristics of the anomaly intensity and the admittance characteristics of the nodes themselves. Nodes with high admittance typically have higher anomaly tolerance, but may also lead to prolonged anomaly dwell time. Based on the probability values ​​of all nodes and paths, the unit generates an anomaly probability heatmap covering the entire network. This heatmap uses color gradients to represent probability magnitude, with dark areas representing high-probability areas and light areas representing low-probability areas. Spatial interpolation techniques are used to generate the heatmap to ensure the continuity and smoothness of the probability distribution while maintaining its correspondence with the actual road network.

[0077] The physical region calibration unit performs in-depth analysis of the anomaly probability heat map to identify anomaly probability clusters. The unit employs clustering analysis to group adjacent nodes with similar probability values ​​into the same category. The clustering process considers spatial continuity and topological connectivity, ensuring that each cluster is physically connected and has a complete connection within the road network. Based on the location of traffic nodes and their connection to the road topology, the unit delineates the physical boundaries of abnormal traffic. Boundary delineation uses a geometric algorithm to generate polygonal boundaries based on the clustering results. These boundaries accurately delineate the core areas affected by abnormal events. Boundary generation considers the actual structure of the road network, ensuring that the boundary lines are consistent with the road orientation and include all high-probability nodes. The final output of the physical region calibration provides a spatial definition of the impact range of abnormal events, offering a spatial reference for subsequent traffic management and control measures.

[0078] The entire implementation process embodies a complete analytical chain from network modeling to regional calibration. Through multi-level simulation calculations and spatial analysis, it achieves precise location and impact range assessment of abnormal traffic events. Each processing stage is interconnected through data flow and model parameters, forming a complete analytical system. The simulation calculations consider the complex characteristics of the road network and the dynamic characteristics of anomaly propagation, ensuring the accuracy and practicality of the analysis results. Spatial analysis techniques guarantee the correspondence between the output results and actual geographic space, providing an intuitive spatial reference for traffic management decisions. Probabilistic calculation methods integrate multiple influencing factors, providing a quantitative assessment of the probability of abnormal events occurring. The final regional calibration results provide traffic management departments with clear operational target areas, supporting precise traffic intervention measures.

[0079] Example 4: See Figure 4, involving the cooperative operation of an adaptive traffic control strategy generation module and a traffic state early warning module. The system dynamically adjusts traffic monitoring and control strategies based on the real-time generated abnormal probability heat distribution map, achieving rapid response and processing of abnormal traffic states.

[0080] The adaptive traffic control strategy generation module continuously monitors the dynamic changes of the abnormal probability heat distribution map. The monitoring mode switching unit scans the abnormal probability values of each region in the distribution map in real time. When it detects that the abnormal probability value of a certain region exceeds the preset abnormal residence probability threshold, it immediately triggers the response mechanism. The system issues a mode switching instruction to the traffic monitoring terminals belonging to this region, including high-definition cameras, microwave detectors, geomagnetic sensors, and other monitoring equipment. After the instruction is executed, the monitoring terminals increase the traffic data sampling frequency to a certain multiple of the original frequency, for example, from 1 sample per minute to 2 samples per minute. At the same time, the event sequence real-time tracking mode is enabled, which uses a sliding window technique to continuously analyze traffic flow data. The deployed transient event detector identifies sudden changes in traffic states based on signal processing algorithms, such as sudden braking, sudden lane changes, and other abnormal driving behaviors.

[0081] The traffic flow disturbance test unit implements test operations on adjacent road nodes with the largest abnormal probability gradient changes. The system calculates the probability change rate of each node in the distribution map through a spatial gradient algorithm and identifies the most significant road node pairs with the most significant gradient changes. Multi-band traffic flow disturbance tests are applied to these nodes. Low-frequency disturbances are achieved by adjusting signal timing schemes, such as extending or shortening the green light time of specific phases. High-frequency disturbances are achieved by issuing guidance information through variable message boards, such as suggesting detour routes or adjusting driving speeds. During the test process, the system records the response of traffic flow, including flow changes, speed adjustments, path selection, and other behavior data.

[0082] The traffic state early warning module performs early warning operations based on the abnormal probability heat distribution map. The location information acquisition unit extracts detailed location information of intelligent traffic devices from the city traffic management database, including the latitude and longitude coordinates of traffic signal controllers, surveillance cameras, radar detectors, and other devices. It also acquires the deployment location information of wireless monitoring elements, including RFID readers, Bluetooth beacons, wireless geomagnetic sensors, and others. All location information is stored and managed using a unified geographic coordinate system.

[0083] The data transmission link establishment unit constructs a communication transmission network based on the obtained position information. The unit first calculates the relative distance between devices and determines intermediate forwarding nodes according to a set distance threshold. The generation of the transmission path uses a path planning algorithm, taking into account factors such as communication distance between devices, signal strength, and obstacle interference. A stable transmission path is finally established from the source node to the target node, ensuring reliable transmission of control instructions and monitoring data. The trigger conditions of the monitoring mode switching unit and the corresponding response measures are shown in Table 1.

[0084] Table 1: Correspondence table of monitoring mode switching trigger conditions and response measures

[0085]

[0086] In actual operation, when the system detects that the abnormal probability value of region R004 reaches 0.91, the monitoring sampling frequency of the region is immediately increased to 4 times the original frequency, and an ultra-high sensitivity event detection mode is enabled. The monitoring device starts collecting traffic flow data at a higher sampling rate, and the transient event detector adjusts the parameter settings to improve the ability to capture weak abnormal signals. At the same time, the system implements multi-band traffic flow disturbance tests on the surrounding road nodes of the region, and observes the response characteristics of the traffic flow by adjusting signal timing and issuing guidance information.

[0087] In the establishment process of the wireless transmission link, the system takes the center node of region R004 as the source node and calculates the optimal transmission path to each target node. According to the actual distance between devices and the signal propagation characteristics, appropriate relay nodes are selected to construct a stable multi-hop wireless transmission network. The dynamic adjustment mechanism of the transmission path can optimize the data transmission efficiency according to the real-time channel conditions, ensuring the timely delivery of warning information and control instructions.

[0088] The entire implementation process embodies a complete closed loop from abnormal detection to response control. The system realizes effective management of abnormal traffic states through real-time monitoring, intelligent analysis, and dynamic adjustment. The flexible switching of monitoring modes ensures the pertinence of data acquisition, the traffic flow disturbance test provides a verification means for system response, and the reliable transmission link ensures the coordinated operation of the entire system. The close cooperation of each link enables the system to adapt to complex and variable traffic environments, providing technical support for urban traffic management.

[0089] Example 5: Involving the coordinated operation of the traffic data transmission module and the traffic state warning module. Based on the established communication transmission link, the system realizes efficient transmission of control instructions and monitoring data, and generates a traffic state warning report through intelligent analysis of abnormal characteristics.

[0090] The traffic data transmission module is responsible for establishing and maintaining a communication transmission network. Based on the location information of each node in the urban traffic network, including the geographic coordinates of fixed nodes such as traffic signal controllers, vehicle detectors, and environmental monitoring equipment, and the dynamic position data of mobile monitoring units, the transmission link is established using a hybrid network architecture combining ZigBee and LoRa wireless communication technologies. The ZigBee network is used for short-range, high-frequency data transmission, covering an area with a radius of about 100 meters, and is mainly used to connect densely deployed traffic monitoring equipment. The LoRa network is used for long-range, low-power data transmission, with a coverage radius of several kilometers, responsible for connecting scattered remote monitoring points. The module dynamically determines the intermediate forwarding nodes based on the actual distance between devices and communication needs. The distance threshold is set to 500 meters, and communication between devices beyond this distance requires forwarding through relay nodes. The generation of transmission paths uses an improved path planning algorithm that considers multiple factors such as signal strength, link quality, and network load to select the optimal transmission path for each data packet. The transmission of control instructions uses a priority scheduling mechanism, with emergency control instructions enjoying the highest transmission priority to ensure timely delivery to the target device. The transmission of monitoring data uses data compression and packet transmission technology to reduce network bandwidth occupancy and improve transmission efficiency.

[0091] The core feature extraction unit of the traffic state warning module extracts key abnormal features from the abnormal probability heat distribution map. This unit uses a multi-scale feature extraction algorithm to analyze the spatial distribution characteristics of the heat distribution map. The extracted features include probability distribution peak, spatial gradient extreme, and distribution range index. The probability distribution peak reflects the core strength of the abnormal event and is obtained by finding the local maximum value point in the distribution map. The spatial gradient extreme represents the spatial variation rate of the abnormal probability and is calculated using an edge detection algorithm. The distribution range index describes the spatial extension degree of the abnormal region and is obtained by calculating the connected area of the high probability region. These features together form the key abnormal feature vector, providing input data for subsequent state detection.

[0092] The warning model execution unit constructs a state detection model based on the support vector machine algorithm, which is trained using historical data and can identify different types of traffic abnormal states. The model input is the key abnormal feature vector, and the output is the traffic state classification result and the warning level. In the warning report generation process, the system calculates the state abnormality index using the following formula:

[0093]

[0094] where: represents the state abnormality index, is the total number of feature dimensions, is the weight coefficient of the kth feature, is the current value of the kth feature, is the historical mean of the kth feature, is the historical standard deviation of the kth feature. This index reflects the deviation of the current state from the historical normal state. The weight coefficient is determined by feature importance analysis, and features with high importance have larger weight values. The system divides the traffic state into multiple warning levels according to the size of the state anomaly index, each level corresponds to different response measures. The warning report contains detailed information such as anomaly type, impact range, duration and recommended measures, providing decision support for traffic management personnel.

[0095] In actual operation, the system takes region R004 as an example to demonstrate implementation details. The anomaly probability heat distribution map of this region shows a probability peak of 0.92, a spatial gradient maximum of 0.15, and a distribution range index of 0.78. The feature extraction unit normalizes these numerical values to generate a feature vector [0.92, 0.15, 0.78]. The state detection model receives this feature vector and calculates a state anomaly index of 2.35. According to the preset threshold range, this index value belongs to the high abnormality level, and the system automatically generates a red warning report. The report indicates that there is a serious traffic anomaly in this region, affecting an area of about 500 meters, and recommends starting emergency traffic control measures.

[0096] The data transmission module also ensures the reliable transmission of warning information and control instructions. The system establishes multiple transmission paths for region R004, with the main path using ZigBee network for direct transmission and the backup path using LoRa network for relay transmission. Data encryption and integrity verification are implemented during transmission to prevent information leakage and tampering. The network state monitoring unit monitors the transmission quality in real time and automatically switches transmission paths when the link quality decreases, ensuring the continuity and reliability of communication.

[0097] The entire implementation process embodies the complete process from data transmission to state warning. The system ensures timely information delivery through reliable data transmission and provides accurate warning information through intelligent feature extraction and state recognition. The coordinated cooperation of each link enables the system to effectively respond to various traffic anomalies and provides technical support for urban traffic management. The multi-scale analysis of the feature extraction algorithm ensures comprehensive capture of abnormal features, the machine learning method of the state detection model provides accurate state recognition ability, and the multi-level data transmission mechanism ensures reliable operation of the system.

[0098] It is to be understood that the terminology used herein is for the purpose of describing particular embodiments only and is not intended to be limiting; it is not intended to exclude myriad other embodiments of the present application that other inventors can develop based on the same general inventive concepts embodied by the described embodiments. That is, although the present application is described in terms of particular embodiments and implementations, it is to be understood that the terminology used is for the purpose of descriptive clarity and that it is intended to be limited only by the words recited in the appended claims. It is to be understood that the terms such as first and second, etc., merely are used to differentiate one from another without necessarily requiring or implying any actual relationship or order between such entities or operations. Moreover, the terms "comprises", "comprising", or any other variations thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but can also include other elements not expressly listed or inherent to such process, method, article, or apparatus.

[0099] While the embodiments of the application have been shown and described herein, it is to be understood that the application is not limited to these embodiments. Rather, many modifications, changes, substitutions, and alterations of the present application can be made by one of ordinary skill in the art without departing from the spirit and scope of the present application, which is defined by the following claims and their equivalents.

Claims

1. A smart city traffic dynamic optimization system based on digital twinning, characterized in that, Comprise: Traffic multidimensional data acquisition module: obtain multidimensional data of urban traffic network from multiple heterogeneous data sources, perform data integration processing on the multidimensional data, and obtain traffic related data warehouse; Traffic digital twin model construction module: perform feature extraction operation on the traffic related data warehouse, generate traffic related feature matrix, construct digital twin traffic dynamic model based on the traffic related feature matrix, and output theoretical traffic state value through the digital twin traffic dynamic model; Traffic flow graph construction module: determine the dimension link graph of each dimension in the traffic related feature matrix, and construct the flow link graph atlas related to urban traffic through the dimension link graph of each dimension; Traffic core feature screening module: screen out traffic core feature sequence based on the flow link graph atlas; Traffic multidimensional optimization analysis module: performs multidimensional difference analysis operation on the theoretical traffic state value and the real-time collected measured traffic data, and generates a regional level difference coefficient matrix in combination with the traffic core feature sequence, specifically including: Time domain cumulative deviation calculation unit: slidingly compares the theoretical traffic state value and the measured traffic data in a preset time window, aligns the non-synchronous sampled traffic data sequences using a sequence alignment algorithm, calculates the cumulative deviation amount in each time window, and generates a time domain deviation vector; Spatial distribution offset detection unit: performs spatial domain decomposition processing on the theoretical traffic state value and the measured traffic data, calculates the district energy distribution ratio, extracts the spatial offset index, and constructs a spatial offset vector; Event sequence matching degree evaluation unit: matches the event sequence distribution of the theoretical traffic state value and the measured traffic data based on a pattern matching algorithm, calculates the time synchronization error of the event trigger point, quantifies the divergence value of the event interval distribution, and generates an event matching degree vector; Difference coefficient matrix generation unit: performs tensor fusion operation on the time domain deviation vector, the spatial offset vector and the event matching degree vector, and outputs a three-order difference coefficient matrix with dimensions of regional number multiplied by timestamp multiplied by difference type through feature weight weighted standardization processing; Traffic event correlation positioning module: inputs the difference coefficient matrix into a spatial topology analysis network, combines road network topology parameters and traffic node location information, generates an abnormal probability thermal distribution map of abnormal traffic event propagation path, and locates an abnormal traffic physical area.

2. The digital-twin-based smart city traffic dynamic optimization system of claim 1, wherein, The traffic multidimensional data acquisition module specifically includes: Data preprocessing unit: performs normalization processing on the multidimensional data to eliminate data dimension difference; Semantic mapping unit: performs semantic mapping conversion on the preprocessed multidimensional data to generate standardized traffic data of each dimension; Data warehouse construction unit: aggregates the standardized traffic data of all dimensions to construct a traffic related data warehouse.

3. The digital-twin-based smart city traffic dynamic optimization system of claim 2, wherein, The traffic digital twin model construction module specifically includes: Historical data feature mining unit: performs multidimensional decomposition processing on historical traffic operation data to extract time series features and spatial distribution features; The dynamic model construction unit inputs the processed historical traffic operation data into a time series prediction network to generate a basic traffic state prediction value, embeds a feature compensation network with a spatial attention mechanism to correct prediction deviations caused by environmental disturbances, and dynamically adjusts prediction weights based on an event feature adapter according to real-time traffic data. The theoretical value generation unit inputs real-time traffic data into the digital twin traffic dynamic model to obtain a theoretical traffic state value.

4. The digital-twin-based intelligent city traffic dynamic optimization system according to claim 3, wherein, The traffic flow graph construction module specifically includes: The dimension linkage graph generation unit selects a dimension in the traffic-related feature matrix, obtains a traffic feature vector of the selected dimension, determines the feature linkage degree between each traffic-related feature in the traffic feature vector, and constructs a dimension linkage graph of the selected dimension according to the feature linkage degree. The flow linkage graph integration unit identifies key traffic-related features of each dimension, determines the cross-dimension feature linkage degree between the key traffic-related features, connects the dimension linkage graphs of each dimension according to the cross-dimension feature linkage degree, and generates a flow linkage graph related to urban traffic.

5. The digital-twin-based smart city traffic dynamic optimization system of claim 1, wherein, The traffic event correlation positioning module specifically includes: The road topology modeling unit constructs a road network node connection relationship topology graph according to traffic node location information, labels path impedance parameters between nodes, superimposes reverse propagation constraint conditions of traffic event trigger points in the topology graph, and generates a traffic topology model including an impedance matrix and a node admittance matrix. The anomaly propagation deduction unit maps the difference coefficient matrix to the corresponding nodes of the traffic topology model, performs an anomaly propagation simulation operation based on a graph structure network, which includes calculating an anomaly event attenuation factor according to path impedance parameters, capturing cross-regional anomaly correlation features through a multi-head attention mechanism, and simulating the diffusion trajectory of the anomaly event in the topology network using a random path simulation algorithm. The probability distribution generation unit calculates anomaly event residence probability values by combining path impedance parameters, and generates an anomaly probability heat distribution map covering the entire network. The physical region demarcation unit performs clustering analysis on the anomaly probability heat distribution map, identifies anomaly probability aggregation areas, and demarcates abnormal traffic physical boundaries according to traffic node locations and road topology connection relationships.

6. The digital-twin-based intelligent city traffic dynamic optimization system according to claim 5, wherein, Further including: An adaptive traffic control strategy generation module configures control parameters according to the anomaly probability heat distribution map. The adaptive traffic control strategy generation module specifically includes a monitoring mode switching unit that issues a mode switching instruction to a traffic monitoring terminal belonging to a region whose anomaly probability value exceeds a preset anomaly residence probability threshold, performs an operation of increasing the traffic data sampling frequency to a multiple of the original frequency, enables an event sequence real-time tracking mode, and deploys a transient event detector to capture traffic state mutation segments. A traffic flow disturbance test unit applies a multi-frequency traffic flow disturbance test operation to adjacent road nodes with the largest anomaly probability gradient change in the anomaly probability heat distribution map.

7. The digital-twin-based intelligent city traffic dynamic optimization system according to claim 6, wherein, Further including a traffic state warning module that performs a traffic state warning operation based on the anomaly probability heat distribution map. The traffic data transmission module specifically comprises: a position information acquisition unit: acquiring position information of intelligent traffic devices in a city traffic network and deployment position information of wireless monitoring elements; A data transmission link establishment unit: establishing a communication transmission link based on the position information, including determining intermediate forwarding nodes according to a distance threshold, and generating a transmission path from a source node to a target node.

8. The digital-twin-based intelligent city traffic dynamic optimization system according to claim 7, wherein, Further comprising: A traffic data transmission module, which establishes a communication transmission link based on traffic node position information, is used for transmitting control instructions and monitoring data; The traffic state early warning module specifically comprises: a core feature extraction unit: extracting a key abnormal feature vector from the abnormal probability heat distribution map; An early warning model execution unit: inputting the key abnormal feature vector into a state detection model to generate a traffic state early warning report. 9.A smart city traffic dynamic optimization method based on digital twinning, characterized in that, All modules and method processes of the intelligent city traffic dynamic optimization system based on digital twinning in any one of claims 1 to 8 are included.

Images