A Method and System for Constructing Vehicle Networking Channels on Urban Expressways Based on Digital Twins

Abstract

This invention proposes a method and system for constructing vehicle-to-everything (V2X) communication channels for urban expressways based on digital twins, belonging to the fields of intelligent transportation and wireless communication technology. It includes: constructing a digital twin space for the urban expressway; collecting multimodal data such as LiDAR point clouds, images, and channel impulse responses; distinguishing between static and dynamic scatterers through point cloud clustering and environment matching; and statistically fitting and generating corresponding scatterer parameter libraries for different traffic flow patterns. Based on this, an intelligent V2X communication channel model is constructed, incorporating line-of-sight, ground reflection, and static / dynamic non-line-of-sight components. A lightweight LSResNet classifier is used to identify traffic flow patterns in real time, dynamically retrieving the parameter library to efficiently solve for the channel impulse response and adapt it to the communication system. This invention achieves refined and adaptive modeling of channel characteristics under different traffic conditions, providing reliable support for high-fidelity simulation verification and optimization design of 6G V2X systems.

Description

Technical Field

[0001] This invention belongs to the field of intelligent transportation and wireless communication technology, and in particular relates to a method and system for constructing vehicle network channels for urban expressways based on digital twins. Background Technology

[0002] The statements in this section are merely background information related to the present invention and do not necessarily constitute prior art.

[0003] With the continuous improvement of autonomous driving levels and the increasing demand for collaborative decision-making by intelligent vehicles, the Internet of Vehicles (IoV) has become a core component of future 6G intelligent transportation systems. The safety of intelligent vehicles highly depends on highly reliable, low-latency real-time information interaction, which places unprecedentedly stringent requirements on a deep understanding and accurate modeling of IoV channel characteristics.

[0004] As a core application scenario for intelligent vehicle-to-everything (V2X) communication, urban expressways experience frequent and dynamic transitions in traffic flow patterns between four typical states: smooth flow, synchronous flow, congested flow, and bottleneck flow. This directly leads to drastic changes in the ratio and spatial distribution of static scatterers such as buildings and guardrails versus dynamic scatterers such as vehicles, resulting in strong non-stationary evolution characteristics of channel multipath components, delay spread, and Doppler spectrum. However, traditional mainstream channel measurement and modeling methods have significant limitations in addressing these challenges. While statistically based stochastic models are computationally efficient, they rely on pre-defined statistical assumptions and struggle to capture and describe the highly time-varying and non-stationary channel characteristics caused by complex traffic flow in urban expressway scenarios. On the other hand, while ray tracing-based deterministic models can achieve high accuracy in static or quasi-static environments, their enormous computational overhead and long modeling cycle prevent them from meeting the need for rapid adaptation to real-time changing traffic environments.

[0005] A more fundamental common problem is that, regardless of whether it's a stochastic or deterministic model, the modeling process is mainly limited to processing and analyzing the radio frequency channel information (such as received signal strength and Doppler shift) collected by the communication equipment itself, severely lacking direct perception and in-depth understanding of the physical environment in which the communication takes place. Based solely on radio frequency information characteristics, it is difficult to accurately distinguish whether the signal reflection source is a dynamic vehicle or a static facility, causing traditional models to fail to fundamentally depict the inherent evolutionary laws of the channel caused by the frequent changes in traffic flow patterns in urban expressway scenarios.

[0006] Meanwhile, existing technologies for identifying traffic flow patterns mostly rely on manual settings, which are inefficient, have significant time lags, and cannot adapt to the real-time changes in traffic flow on expressways. This leads to untimely retrieval of scatterer parameter databases, reducing the real-time performance and accuracy of channel modeling. Furthermore, traditional solutions do not consider the limited computing resources of onboard terminals and lack lightweight traffic flow pattern recognition algorithms, limiting the deployment and application of channel models in practical vehicle-to-everything (V2X) systems. Summary of the Invention

[0007] To overcome the shortcomings of the existing technologies, this invention provides a method and system for constructing vehicle network channels on urban expressways based on digital twins. It achieves centimeter-level spatiotemporal alignment of multimodal sensing data and wireless channel data through digital twin space, accurately classifies and dynamically tracks static and dynamic scatterers, and establishes a four-state vehicle flow scatterer parameter library. A lightweight LS (Laser-Based Array) is employed. The ResNet classifier enables automatic and accurate identification of traffic flow patterns, real-time matching with the corresponding parameter library, and dynamic adaptive construction of the channel model, thereby improving modeling accuracy, real-time performance, and vehicle-mounted device adaptability.

[0008] To achieve the above objectives, one or more embodiments of the present invention provide the following technical solutions: The first aspect of this invention provides a method for constructing vehicle network connectivity channels on urban expressways based on digital twins; A digital twin-based method for constructing vehicle network channels on urban expressways includes: A digital twin space for urban expressways is constructed, and multimodal data is collected based on the digital twin space. The multimodal data includes lidar point cloud data, image data, and channel impulse response data. The lidar point cloud data and channel impulse response data are preprocessed, and the three-dimensional spatial coordinates of the scatterer are extracted. Based on the preprocessed lidar point cloud data, scattering objects in the environment are identified through clustering algorithms; the identification results are spatially matched with the preprocessed scattering object coordinates, and the scattering objects are classified into static scattering objects and dynamic scattering objects according to a preset distance threshold; For different traffic flow patterns, the parameter distributions of the static and dynamic scatterers are statistically analyzed, and the statistically obtained parameter distributions are fitted using a probability distribution model to generate a scatterer parameter library corresponding to different traffic flow patterns. Based on the aforementioned scatterer parameter library, a smart vehicle network channel model for urban expressways is constructed. Image data from the collected multimodal data is input into a trained LS-ResNet traffic flow pattern classifier, which outputs classification results for different traffic flow patterns. Based on the classification results of traffic flow patterns, the corresponding parameter set is retrieved from the scatterer parameter library, and the channel impulse response of the current traffic flow pattern is solved based on the constructed intelligent vehicle-to-everything (V2X) channel model for urban expressways. The V2X communication system is then adapted based on the channel impulse response.

[0009] As a further technical solution, the construction of the digital twin space for urban expressways includes: Use 3D modeling tools to build a geometric model of the urban expressway to scale, and configure the corresponding electromagnetic material parameters for the objects in the model; Based on traffic flow theory, traffic flow patterns are divided into smooth flow, synchronous flow, congested flow, and bottleneck flow. The vehicle density, speed range, and movement rules of each traffic flow pattern are defined, and corresponding vehicle trajectory files are generated. The geometric model and vehicle trajectory file are synchronously imported into the perception simulation platform and electromagnetic simulation platform, and a unified spatiotemporal reference is configured to achieve simulation data alignment.

[0010] As a further technical solution, based on the preprocessed lidar point cloud data, scattering objects in the environment are identified using a clustering algorithm; the identification results are spatially matched with the preprocessed scattering object coordinates, and the scattering objects are classified into static scattering objects and dynamic scattering objects according to a preset distance threshold, including: Density-based spatial clustering is performed on the preprocessed lidar point cloud data to divide the point cloud into multiple independent clusters, and each cluster is classified as a static object or a dynamic vehicle based on its geometric dimensions. For each scatterer coordinate extracted from the channel data, calculate the minimum spatial distance between it and all identified static objects and dynamic vehicles; If the minimum spatial distance is less than the preset matching threshold, the attribute of the scatterer is marked as the attribute of the object closest to it; otherwise, the scatterer is marked as an unidentified scatterer and excluded in subsequent modeling, thereby completing the static and dynamic classification of the scatterer.

[0011] As a further technical solution, for different traffic flow patterns, the parameter distributions of the static and dynamic scatterers are statistically analyzed, and a probability distribution model is used to fit the statistically obtained parameter distributions to generate a scatterer parameter library corresponding to different traffic flow patterns, including: For each traffic flow pattern among smooth flow, synchronous flow, congested flow, and bottleneck flow, the key parameter distribution of static and dynamic scatterers in all communication links is statistically analyzed. The key parameters include at least the number of scatterers, propagation distance, arrival / departure angle, and power delay spectrum. A Gaussian mixture model is used to fit the cumulative distribution function of the number of scatterers and the propagation distance, and a Gaussian distribution is used to fit the statistical distribution of the arrival / departure angle; The parameters of the fitted probability distribution model are stored according to traffic flow patterns, forming a scatterer parameter library containing parameter sets specific to four traffic flow patterns.

[0012] As a further technical solution, based on the aforementioned scatterer parameter library, a smart vehicle network connectivity channel model for urban expressways is constructed, including: Based on the traffic flow pattern of the target simulation scenario, the corresponding parameter set is selected from the parameter library, and the static and dynamic scatterers and scattering cluster parameters of the communication link at the initial moment are randomly generated according to the statistical distribution law of each parameter. Based on the visible region method, the generation, survival and extinction of scattering clusters during the communication duration are simulated to generate a channel coefficient sequence with time non-stationary characteristics. Based on the generated parameters and channel coefficients, a complete channel impulse response is constructed, including line-of-sight components, ground reflection components, static non-line-of-sight components, and dynamic non-line-of-sight components.

[0013] As a further technical solution, the image data from the collected multimodal data is input into the trained LS-ResNet traffic flow pattern classifier, which outputs classification results for different traffic flow patterns, including: The real-time acquired RGB image is input into the LS-ResNet classifier. The RGB image is preprocessed and its size is normalized to a preset resolution before being input into the trained LS-ResNet classifier. The LS-ResNet classifier performs preliminary feature extraction and dimensionality compression on the input image through the Stem layer, and outputs the first feature map; The first feature map is subjected to deep feature extraction through multiple residual blocks of the feature extraction layer. Each residual block extracts the global spatial distribution features and local detail features of the traffic flow through the main path of LS convolution, and performs residual connection through short-circuit path to output the second feature map. The second feature map is converted into a feature vector by the global average pooling layer in the classification head, and then mapped to four output nodes by a fully connected layer. The classification probability of different traffic flow patterns is calculated by the Softmax activation function. The traffic flow pattern corresponding to the maximum classification probability is taken as the classification result of the current frame, and the confidence level corresponding to the classification result is recorded.

[0014] As a further technical solution, adapting the vehicle-to-everything (V2X) communication system based on the aforementioned channel impulse response includes: Based on the channel impulse response reconstructed in real time, channel quality parameters are extracted, and communication bandwidth, transmission power, and time slot resources are dynamically allocated. The beam direction and gain of the vehicle-mounted antenna are adjusted based on the angular parameters of the scatterer in the channel impulse response.

[0015] A second aspect of the present invention provides a system for constructing a vehicle network channel for urban expressways based on digital twins.

[0016] A digital twin-based urban expressway vehicle network channel construction system includes: The digital twin construction and data acquisition module is configured to: construct a digital twin space for urban expressways, and acquire multimodal data based on the digital twin space, including lidar point cloud data, image data, and channel impulse response data; The data processing and feature extraction module is configured to: preprocess the lidar point cloud data and channel impulse response data, and extract the three-dimensional spatial coordinates of the scatterer; The scatterer identification and classification module is configured to: identify scatterers in the environment based on preprocessed lidar point cloud data using a clustering algorithm; spatially match the identification results with the preprocessed scatterer coordinates; and classify scatterers into static scatterers and dynamic scatterers according to a preset distance threshold. The parameter library generation module is configured to: statistically analyze the parameter distributions of the static and dynamic scatterers for different traffic flow patterns, and fit the statistically obtained parameter distributions using a probability distribution model to generate a scatterer parameter library corresponding to different traffic flow patterns. The channel model construction module is configured to: construct a smart vehicle network channel model for urban expressways based on the scatterer parameter library; The traffic flow pattern recognition and channel adaptation module is configured to: input image data from the collected multimodal data into a trained LS-ResNet traffic flow pattern classifier, and output classification results for different traffic flow patterns; retrieve the corresponding parameter set from the scatterer parameter library based on the classification results of the traffic flow pattern, and solve the channel impulse response of the current traffic flow pattern based on the constructed intelligent vehicle-to-everything (V2X) channel model for urban expressways, and adapt the V2X communication system based on the channel impulse response.

[0017] A third aspect of the present invention provides a computer-readable storage medium having a program stored thereon, which, when executed by a processor, implements the steps of a method for constructing a vehicle network channel for an urban expressway based on digital twins as described in the first aspect of the present invention.

[0018] The fourth aspect of the present invention provides an electronic device, including a memory, a processor, and a program stored in the memory and executable on the processor, wherein the processor executes the program to implement the steps in the method for constructing a vehicle network channel for an urban expressway based on digital twins as described in the first aspect of the present invention.

[0019] The above one or more technical solutions have the following beneficial effects: (1) This invention integrates multimodal sensing data from LiDAR, RGB cameras, and other devices natively mounted on intelligent vehicles with wireless channel information, performing centimeter-level spatiotemporal alignment and fusion processing in a unified digital twin space. This breaks through the inherent limitations of traditional channel modeling methods that rely solely on limited radio frequency information, enabling the model to directly perceive and accurately distinguish between static scatterers such as buildings and guardrails and dynamic scatterers such as vehicles in the environment. Through this precise classification of scatterer attributes based on real-world environmental perception, the core problem of inaccurate channel characterization caused by the inability of traditional models to identify the static and dynamic attributes of scatterers is fundamentally solved, laying a solid data foundation for constructing high-fidelity channel models.

[0020] (2) Addressing the challenges of frequent traffic flow patterns on urban expressways, this invention, based on traffic flow theory, clearly defines and constructs a refined scatterer parameter library adapted to four typical traffic flow patterns: smooth flow, synchronous flow, congested flow, and bottleneck flow. The model can dynamically call and match the corresponding parameter set to generate channels based on real-time or preset traffic conditions. This allows the channel model constructed in this invention to accurately characterize the differences in channel characteristics across all scenarios, from low-density high-speed to high-density near-stagnation. Simultaneously, this invention utilizes high-fidelity digital twin simulation to generate massive and accurate labeled datasets offline in the early stages, extracting the distribution patterns of key parameters through statistical analysis to form a lightweight parameter library. During actual channel construction or simulation, time-consuming real-time ray tracing calculations are unnecessary; efficient random generation based on parameter distribution is sufficient, combined with algorithms such as the visible area method to dynamically update the birth and death of scattering clusters. While maintaining near-deterministic model accuracy, this significantly reduces computational complexity and time overhead, providing a practical channel modeling tool for real-time simulation, rapid deployment, and online optimization of 6G vehicle-to-everything (V2X) systems.

[0021] (3) This invention models the survival, extinction, and regeneration of scattering clusters in the time dimension by introducing the visible region method, which can naturally simulate the evolution of channel multipath structure caused by the relative motion of vehicles and changes in occlusion relationships. The channel impulse response model it finally constructs can flexibly configure and synthesize the contributions of each component according to theoretical or measured data, so as to comprehensively and realistically reflect the time-varying, frequency-varying, and space-varying characteristics of the channel in the dynamic scenario of urban expressways, providing an extremely realistic channel environment for the development and testing of advanced communication algorithms.

[0022] Advantages of additional aspects of the invention will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of the invention. Attached Figure Description

[0023] The accompanying drawings, which form part of this invention, are used to provide a further understanding of the invention. The illustrative embodiments of the invention and their descriptions are used to explain the invention and do not constitute an improper limitation of the invention.

[0024] Figure 1 This is a flowchart of the method in the first embodiment.

[0025] Figure 2 This is a system structure diagram of the second embodiment. Detailed Implementation

[0026] It should be noted that the following detailed descriptions are exemplary and intended to provide further illustration of the invention. Unless otherwise specified, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains.

[0027] It should be noted that the terminology used herein is for the purpose of describing particular implementations only and is not intended to limit the exemplary implementations of the present invention.

[0028] Where there is no conflict, the embodiments and features in the embodiments of the present invention can be combined with each other.

[0029] Example 1 This embodiment discloses a method for constructing a vehicle-to-everything (V2X) channel for urban expressways based on digital twins. It constructs a high-fidelity digital twin scenario and simultaneously collects LiDAR, image, and channel data. Point cloud clustering is used to identify environmental objects and match them with channel scatterers, accurately distinguishing between dynamic and static scatterers. Based on this, the distribution of scatterer parameters under different traffic flow patterns is statistically analyzed to establish a parameterized model library. Finally, the visible area method is combined to construct a simulation model that reflects the non-stationary characteristics of the channel, providing high-precision channel simulation support for the design of 6G V2X systems.

[0030] Specifically, such as Figure 1 As shown, the method for constructing vehicle network channels on urban expressways based on digital twins includes: Step S1: Construct a digital twin space for urban expressways and collect multimodal data based on the digital twin space. The multimodal data includes lidar point cloud data, image data, and channel impulse response data.

[0031] Step S101: In constructing the digital twin space of the urban expressway, firstly, a 3D model of the objects in the scene is built. Using the Blender 3D modeling tool, the physical environment of the target urban expressway is replicated at a 1:1 scale, accurately restoring core road structures such as viaducts, piers, and guardrails. At the same time, the geometric dimensions and spatial layout of static objects such as buildings and green belts in the scene are recorded. For the material characteristics of various objects in the scene, the corresponding dielectric constant, surface roughness, and penetration loss parameters are configured. Among them, the guardrails and piers use reinforced concrete material parameters, and the road surface uses asphalt material parameters to ensure the authenticity of subsequent electromagnetic simulation. Finally, the completed 3D model can be exported to a file format compatible with the Wireless InSite and AirSim platforms, laying the foundation for cross-platform simulation.

[0032] Secondly, based on traffic flow theory and relevant regulations, realistic movement trajectories are designed for the four types of traffic flow. Specifically: smooth flow (8 vehicles, 80–100 km / h) travels in uniform lanes with no frequent lane changes; synchronized flow (23 vehicles, 40–80 km / h) travels at the same speed in the same lane with a few smooth lane changes; congested flow (43 vehicles, 20–40 km / h) travels intermittently with reduced spacing; bottleneck flow (57 vehicles, 0–20 km / h) is densely distributed with localized stagnation.

[0033] Finally, the initial position, velocity curve, and lane change rules of the vehicle are defined using Python scripts. The constructed geometric model and vehicle trajectory file are then synchronously imported into the perception simulation platform and the electromagnetic simulation platform, and a unified spatiotemporal reference is configured to achieve simulation data alignment.

[0034] Step S102: In the perception simulation platform, a vehicle equipped with a LiDAR and image acquisition device is simulated to run along a preset trajectory, and multimodal perception data containing object geometry and texture information is collected at a preset sampling period. Specifically, in this embodiment, the 3D model can be imported into the AirSim platform for perception data simulation. The vehicle is equipped with a LiDAR and an RGB camera, and the sampling frequency is set to 0.01 s. The LiDAR meets centimeter-level positioning, and the RGB camera has a resolution of 1920×1080. A Python script is used to start the four-state traffic flow simulation and control the vehicle's movement, collecting 6400 sets of LiDAR point cloud data and 6400 RGB images, which are then stored according to traffic flow pattern and timestamp.

[0035] Image data acquired using image acquisition equipment provides visual texture information, which complements the geometric information of point clouds, enabling precise characterization of environmental features in urban expressway scenarios. In channel modeling, image data is used to assist in the separation and modeling of static scatterers (buildings, guardrails, etc.) and dynamic scatterers (moving vehicles). Specifically, RGB images are used to verify the clustering results of LiDAR point clouds and simultaneously provide visual feature support for the DBSCAN clustering algorithm, achieving effective grouping of point cloud data.

[0036] In the electromagnetic simulation platform, based on the same scene model and the real-time vehicle position, wireless signal propagation is simulated using ray tracing, and channel impulse response data including multipath component parameters are collected with the same sampling period. Specifically, in this embodiment, the 3D model is imported into Wireless InSite, configured with a carrier frequency of 5.9 GHz and a bandwidth of 20 MHz, and the vehicle is equipped with an omnidirectional antenna; the vehicle position is dynamically updated using MATLAB scripts, and the simulation duration is set to 2 s with a time interval of 0.01 s, consistent with the sampling frequency of the sensing data. Based on the above settings, ray tracing simulations are conducted, and parameters such as channel impulse response and Doppler shift of 22,400 communication links are collected.

[0037] Finally, by aligning the timestamps with the spatial coordinates, it is ensured that the multimodal sensing data and the channel impulse response data are strictly synchronized and correspond in time and space, thus forming a fused dataset.

[0038] Step S2: Preprocess the lidar point cloud data and channel impulse response data, and extract the three-dimensional spatial coordinates of the scatterer.

[0039] Since the origin of the LiDAR point cloud coordinates is the current position of the acquisition platform, its coordinate system origin is not uniform and needs to be aligned with the global coordinate system. At each acquisition moment, the real-time spatial pose parameters of the sensor in the carrier coordinate system are obtained, and the relative position information of the point cloud acquired by the LiDAR relative to the sensor is recorded.

[0040] For cases where the detection platform moves along different directions of the coordinate axes, corresponding sign transformations and superposition operations are performed on the point cloud coordinate components based on the motion direction vectors. Specifically: When the platform moves along the positive lateral direction, the x-coordinate component of the relative coordinate is added to the x-coordinate component of the sensor position in the same direction; when moving along the negative lateral direction, the x-coordinate component of the relative coordinate is inverted and then superimposed with the sensor position; when moving along the positive longitudinal direction, the y-coordinate component of the relative coordinate is added to the y-coordinate component of the sensor position in the same direction; when moving along the negative longitudinal direction, the y-coordinate component of the relative coordinate is inverted and then superimposed. At the same time, the height coordinate component is inverted and superimposed with the sensor height position, thus obtaining the absolute coordinates of the point cloud in the global coordinate system.

[0041] Furthermore, since the point cloud data acquired by LiDAR contains a large number of redundant ground points, which offer no advantage in distinguishing between static and dynamic objects, this embodiment also performs a ground redundancy point filtering operation on the point cloud data. By selecting the minimum z-coordinate among all point cloud coordinates and adding a coordinate correlation factor, this minimum z-coordinate is used as a threshold to identify whether a point is a ground point. If the z-coordinate of the point is greater than a certain value, it is retained; otherwise, it is deleted, ultimately yielding the effective point cloud data after ground point filtering.

[0042] For channel impulse response data, Wireless InSite electromagnetic simulation software based on ray tracing method can be used to simulate and collect wireless channel data. For each pair of transmitter and receiver communication links, the spatial coordinates of all scatterers contained in the link are extracted, and non-ground scatterers are filtered using the same threshold as that used to remove ground points from lidar point clouds. Only scatterers at a certain height above the ground are retained, thus obtaining the coordinates of scatterers without the influence of ground scattering, and classifying and storing them according to different communication links.

[0043] Step S3: Based on the preprocessed lidar point cloud data, scatterers in the environment are identified through a clustering algorithm; the identification results are spatially matched with the preprocessed scatterer coordinates, and the scatterers are classified into static scatterers and dynamic scatterers according to a preset distance threshold.

[0044] Step S301: Perform density-based DBSCAN clustering on the lidar point cloud data to identify environmental targets. The density clustering algorithm divides the data point set into connected regions with spatial density higher than a preset standard, and each region has a significant distance between it. Specifically, the DBSCAN method requires configuring a neighborhood radius parameter and a minimum density threshold parameter. If the number of points contained in the neighborhood of a data point exceeds the minimum density threshold, it is marked as a core point; if the number of points in its neighborhood does not reach the minimum density threshold but is within the neighborhood of any core point, it is defined as a boundary point. All other cases are classified as noise points. By iteratively traversing all data points in the point cloud, the DBSCAN method is used to assign each point to a corresponding cluster, and finally outputs several cluster sets. Each cluster contains a set of points with sufficient spatial density and the clusters remain relatively separated.

[0045] Furthermore, based on the cluster set corresponding to the detected target, the three-dimensional coordinate information of the point cloud within each cluster is extracted one by one, and its maximum and minimum values ​​in the X, Y, and Z coordinate axes are calculated. Then, the spatial size parameters of the corresponding object are determined by the axial range. Each axial range is compared with a preset vehicle size determination threshold. If the scale of the current cluster in all three dimensions is smaller than the vehicle size threshold, it is identified as a dynamic vehicle target; otherwise, it is determined as a static obstacle.

[0046] Step S302: Perform a traversal calculation on the spatial distance between each scatterer and the detected vehicles. If the distance between the scatterer and any vehicle is less than a preset distance threshold, the scatterer is determined to be a dynamic scatterer. Using the same determination mechanism, if the distance between the scatterer and a static object is less than the threshold, it is classified as a static scatterer. Remaining scatterers that do not overlap with any vehicle or static object are defined as unknown scatterers, indicating that they exceed the effective detection range of the lidar point cloud. Given that such unknown scatterers typically maintain a relatively large spatial distance from the transceiver ends of the communication link, their path loss and weak scattering contribution in the 5.9GHz band are negligible, and therefore their impact on channel modeling can be excluded.

[0047] Step S4: For different traffic flow patterns, the parameter distributions of the static and dynamic scatterers are statistically analyzed, and the statistically obtained parameter distributions are fitted using a probability distribution model to generate a scatterer parameter library corresponding to different traffic flow patterns.

[0048] Step S401: Determine the quantity parameters of static and dynamic scatterers and static and dynamic scattering clusters. For static and dynamic scatterers, the quantity of static and dynamic scatterers in the communication link from vehicle i to vehicle j is denoted as... The distance between the transmitter and receiver directly affects the estimation of the number of scatterers; therefore, the static / dynamic scatterer number parameter is defined as follows: , represented as:

[0049] in, and These represent the positions of the i-th and j-th vehicles, respectively.

[0050] The ratio of static to dynamic scatterers in the next communication link of the four-state vehicle flow is statistically analyzed, and the cumulative distribution function (CDF) is fitted using the Gaussian Mixture Model (GMM), which is expressed as:

[0051] in, It is expressed as the cumulative distribution function of the number of static / dynamic scatterers fitted using GMM; The number of static / dynamic scatterers; The weights of the k-th Gaussian distribution are... Let represent the cumulative distribution function of the k-th Gaussian distribution, where It is the CDF of the standard normal distribution. and These are the mean and standard deviation of the k-th Gaussian distribution, respectively, which can be determined through statistical fitting of channel data acquired in the digital twin space. Similarly, the statistical distribution of the number of static and dynamic scattering clusters, as well as the GMM fitting parameters, can be obtained.

[0052] Step S402: Determine the distance parameters of the static scatterer and the dynamic scatterer. Set the distance parameters of the m-th static scatterer and the n-th dynamic scatterer between the i-th and j-th vehicles. , Represented as:

[0053]

[0054] in, and Let m and n represent the positions of the m-th static scatterer and the n-th dynamic scatterer between the i-th and j-th vehicles, respectively. This represents the calculation of the Frobenius norm. Similarly, the distance distributions of static and dynamic scatterers can be fitted using the GMM distribution, expressed as:

[0055] in, To represent the cumulative distribution function of static / dynamic scatterers at different distances fitted using GMM; The weights of the k-th Gaussian distribution are... Let represent the CDF of the k-th Gaussian distribution, where It is the CDF of the standard normal distribution. and These are the mean and standard deviation of the k-th Gaussian distribution, respectively, which can be determined by statistical fitting of channel data collected in the digital twin space.

[0056] Step S403: Determine the angular parameters of the static and dynamic scatterers. The angular parameters of the scatterers mainly include the horizontal angle of arrival (AAoA), horizontal angle of departure (AAoD), pitch angle of arrival (EAoA), and pitch angle of departure (EAoD). Taking the horizontal angle of arrival as an example, the ratio of the horizontal angle of arrival of the m-th static scatterer to the n-th dynamic scatterer between vehicles i and j is... , They are represented as follows:

[0057]

[0058] in, and Let represent the horizontal angles of arrival of the m-th static scatterer and the n-th dynamic scatterer at the transceiver end, respectively. Based on this, the horizontal angles of arrival of the static and dynamic scatterers in each link were analyzed using the constructed digital twin intelligent vehicle network and multimodal perception intelligent fusion dataset. Their distribution conforms to a Gaussian distribution, expressed as:

[0059]

[0060] in, Let be the cumulative distribution function of the static scatterer AAoA; Let be the cumulative distribution function of the dynamic scatterer AAoA; Let be the error function. and The mean and standard deviation of the Gaussian distributions representing the horizontal arrival angles of static and dynamic scatterers, respectively, can be determined through statistical fitting of channel data acquired in a digital twin space. Similarly, the horizontal departure angle can be calculated. Angle of elevation reached and pitch departure angle It follows a Gaussian distribution.

[0061] Step S404: Determine the power-time delay parameters of the static and dynamic scatterers. Path power is characterized as an exponential function of time delay. The path power is decomposed into static scatterer path power and dynamic scatterer path power. The path power between vehicle i and vehicle j, passing through the m-th static scatterer and the n-th dynamic scatterer, can be expressed as:

[0062]

[0063] in, In representing the polynomial fitting of static / dynamic scatterers, the th k The delay parameter of the item, k =1,2,3 It represents the path delay of static scatterers and dynamic scatterers m / n. Follows Gaussian distribution To achieve accurate polynomial fitting, and Taking the logarithm of both sides, we get:

[0064]

[0065] In the formula, the parameters can be determined by statistical fitting of channel data collected in the digital twin space.

[0066] Step S405: The fitted probability distribution model parameters are stored according to traffic flow patterns to form a scatterer parameter library containing parameter sets specific to four traffic flow patterns.

[0067] Step S5: Based on the scatterer parameter library, construct the intelligent vehicle-to-everything (V2X) channel model for urban expressways.

[0068] Step S501, Model the LoS component of the intelligent vehicle network channel on the urban expressway: the complex channel gain of the line-of-sight transmission link from vehicle i to vehicle j. It can be represented as:

[0069] Where Q(t) is a rectangular window function, satisfying The value is 1 when the condition is met and 0 when the condition is not met. T0 is the observation time interval. The frequency is the Doppler frequency. For phase shift; Doppler frequency of the line-of-sight component from vehicle i to vehicle j. Phase shift and latency It can be represented as:

[0070]

[0071]

[0072] in, , and These represent the inner product, initial phase difference, and carrier wavelength, respectively. and Let represent the velocity vectors of the i-th and j-th vehicles, respectively. Furthermore, the distance vector DLoS(t) between the i-th and j-th vehicles can be calculated as:

[0073] Step S502, Modeling the ground reflection component of the intelligent vehicle network channel on the urban expressway: The complex channel gain of the ground reflection component of the transmission link from vehicle i to vehicle j can be expressed as:

[0074] in, , , The ground reflection power, Doppler frequency, phase, and time delay from vehicle i to vehicle j, respectively, can be expressed as:

[0075]

[0076] t Real-time latency It can be calculated as follows:

[0077] in, Let be the distance vector between the i / jth vehicle and the ground reflection point, and let their magnitudes be respectively and .

[0078] in, Let be the ground clearance of vehicles i and j; Let be the azimuth distance between the i-th vehicle at the transmitter and the ground reflection point, which can be expressed as:

[0079] Therefore, the distance vector between the i / jth vehicle and the ground reflection point It can be represented as:

[0080] in, and Distance vectors The azimuth and elevation angles. Since the azimuth angle of the ground reflection path is usually consistent with the Loss (LoS), and the total power of the LosS path and the ground reflection path remains constant, only the elevation angle of the ground reflection path needs to be considered. The calculation is as follows:

[0081] in, The elevation angle of the ground reflection path.

[0082] Step S503, Scatterer Parameter Generation: Based on the statistical distribution of the relevant parameters of the obtained static and dynamic scatterers, the quantity, distance, angle, and power parameters of static and dynamic scatterers under four traffic flow patterns—smooth flow, synchronous flow, congested flow, and bottleneck flow—are generated respectively. Specifically, at the initial time t0, according to the statistical distribution law of the scatterer quantity parameter, the number of static and dynamic scatterers between the i-th and j-th vehicles at the initial time t0 is determined, denoted as N. s (t0) and M s (t0); Based on the statistical distribution characteristics of the scatterer distance parameters, the distance is generated according to the GMM distribution. The spatial distance between the i-th static scatterer and the j-th dynamic scatterer at the initial time t0 is calculated and denoted as D. i (t0) and D j(t0); Based on the statistical distribution of the scatterer angle parameters, the AAoD parameter θ is generated following the GMM. i k (t0), AAoA parameter α i k (t0), EAoA parameter β i k (t0), EAOD parameter φ j k (t0). After generating the above parameters, clustering operations are performed on the static scatterer and the dynamic scatterer respectively to obtain static scattering clusters and dynamic scattering clusters, the number of which is denoted as N. Sta (t0) and M Dyn (t0), this quantitative characteristic also follows the statistical distribution law of the scattering cluster quantity parameter obtained in step 5.

[0083] Step S504: Based on the visible area method, considering newly formed static clusters and dynamic clusters, the non-stationarity and consistency of the intelligent vehicle-to-everything (V2X) channel under four traffic flow patterns on urban expressways are characterized. During the time evolution, the model continuously updates the position evolution of dynamic scatterers through the GMM probability distribution and uses the visible area algorithm to simulate the birth and death of clusters. Furthermore, the K-means clustering algorithm is used to cluster scatterers at different speeds. The visible area of ​​each vehicle is modeled as a hemisphere centered on the vehicle, with the visible area radius R for vehicles i and j being... i R j This is the maximum distance between the vehicle and the initially generated velocity change cluster at the initial moment, located within R during time t. i R j Clusters within the range are defined as visible clusters.

[0084] Due to time The distance between the cluster and the transmitter / receiver is still less than the radius of the visible area; therefore, the cluster remains within the visible area and affects the channel. In time... The number of surviving allochthonous clusters between the i-th and j-th vehicles is determined by Given. Besides the surviving clusters, time... Several new allochthonous clusters also exist, targeting time. The specific distance between the i-th and j-th vehicles is a quantitative parameter related to the allochthorium cluster. The number of new clusters will be randomly generated based on the GMM distribution, and the number of new clusters will be determined by the formula. Calculation. When Greater than hour, Clusters with different speeds participate in the channel implementation. If Less than If no new cluster is generated, then no new cluster will be generated.

[0085] Step S505, Modeling the NLoS Components of the Intelligent Vehicle-to-Everything (V2X) Channel on Urban Expressways: Define the clusters whose centroids are closer to the transmitter / receiver as Tx / Rx clusters, and randomly match each Tx cluster with an Rx cluster. The p-th static Tx / Rx cluster and the q-th dynamic Tx / Rx cluster are respectively represented as... and The velocity vector of the q-th dynamic Tx / Rx cluster is represented as In the NLoS component from vehicle i to vehicle j, the nth component passes through the p-th static scattering cluster. p The complex channel gain of the scattering point can be calculated as follows:

[0086] in, This represents the normalized power of the static scatterer. The Doppler frequency of the static scatterer at the transmitter / receiver end. The phase shift can be calculated using the following formulas:

[0087]

[0088] in, It follows an exponential distribution, representing the p-th pair of static scattering clusters. The latency of virtual links between them. This indicates the relationship between the transmitter / receiver and the p-th pair of static scattering clusters. The nth p The distance between each scattering point The calculation is as follows:

[0089] Total delay for this part The calculation is as follows:

[0090] Similarly, the NLoS components from vehicle i to vehicle j pass through the nth dynamic scattering cluster of the qth dynamic scattering cluster. q Complex channel gain of each scatterer It can be calculated as:

[0091] Among them, Doppler frequency Phase shift They can be calculated as follows:

[0092]

[0093] Finally, the time delay through the dynamic scattering cluster link can be expressed as: .

[0094] Step S506, construct a digital twin-based intelligent vehicle network channel model for urban expressways: In the intelligent vehicle network channel, the channel impulse response (CIR) of the communication channel h(t,τ) from vehicle i to vehicle j can be expressed as:

[0095] in, Let represent the Rice factor of the transmission link from vehicle i to vehicle j. , , These are the power ratios of the ground reflection component, the static scattering cluster component, and the dynamic scattering cluster component, respectively, satisfying... .

[0096] Step S6: Input the image data from the collected multimodal data into the trained LS-ResNet traffic flow pattern classifier and output the classification results of different traffic flow patterns; retrieve the corresponding parameter set from the scatterer parameter library according to the classification results of traffic flow patterns, and solve the channel impulse response of the current traffic flow pattern based on the constructed urban expressway intelligent vehicle network channel model, and adapt the vehicle network communication system based on the channel impulse response.

[0097] Step S61 involves using the multimodal data collected from the AirSim simulation platform to associate RGB images with defined four-state traffic flow labels (smooth flow, synchronous flow, congested flow, and bottleneck flow), forming the foundation of the dataset. Additionally, web crawling technology is used to crawl images captured by vehicle-mounted cameras on urban expressways at different times (daytime and nighttime) and under different weather conditions. The crawled image data is then automatically labeled with traffic flow patterns: smooth flow, synchronous flow, congested flow, and bottleneck flow. After labeling, manual verification is performed to ensure an accuracy rate of at least 98%, serving as a supplementary set to the training data. Together, these two sets constitute the traffic flow recognition dataset.

[0098] In step S62, ResNet-18 is selected as the base network, and a large-small convolutional structure is introduced to replace some of the 3×3 convolutions in the traditional ResNet to build a lightweight LS-ResNet classifier. At the same time, relying on the advantages of LS convolution in "large field of view perception and small range aggregation", combined with the residual connection stability of ResNet, the global distribution features of traffic flow pattern and local vehicle motion features are accurately extracted.

[0099] Let the image resize of the vehicle-mounted RGB camera be... This feature is directly used as input to the LS-ResNet classifier. ,Right now:

[0100] The specific structure of the LS-ResNet model is as follows: (1) Stem layer: A 7×7 convolution (stride 2) + batch normalization (BN) + ReLU activation function is used to map the input features into a low-resolution, high-channel feature map (64 channels, 112×112 resolution), achieving initial feature extraction and dimensionality compression; then max pooling (3×3, stride 2) is used to further reduce the feature map resolution, expand the receptive field of the lower convolutional layer while retaining key features, as shown below:

[0101] (2) Feature Extraction Layer: Contains four stages of residual blocks, with two residual blocks in each stage, for a total of eight residual blocks. Each residual block consists of a main path composed of "LS convolution + BN + ReLU" and a shortcut path composed of "1×1 convolution + BN" for dimension matching, forming a residual connection structure to avoid gradient vanishing in deep networks. Specifically, the output features of each residual block... :

[0102] in, The core operation of the main path is LS convolution, which consists of two parts: Large Kernel Perception (LKP) and Small Kernel Aggregation (SKA), targeting the input image features. The global spatial distribution and local detailed features of traffic flow are extracted, and mathematically represented as follows:

[0103] in, For Large Kernel Perception (LKP) operations, a 7×7 deep separable convolution with a large kernel is used. By combining 1×1 convolution, the focus is on extracting the global spatial distribution features of traffic flow (such as overall road vehicle density and congestion range), mathematically represented as:

[0104] In the formula, This indicates a 7×7 depth-separable convolution, used to reduce computational cost; Represents the input feature X Neighborhood enables large-scale feature coverage; PW represents 1×1 point convolution, used to adjust the channel dimension and match the output channels of the residual block.

[0105] For kernel-small aggregation operations, a 3×3 kernel-small dynamic convolution is used. Divide the channel into Groups (C is the number of input channels) share aggregation weights, focusing on extracting local detailed features of traffic flow (such as the outline and position of individual vehicles) to achieve adaptive aggregation of local features. Mathematically, this is expressed as...

[0106] In the formula, The aggregate weights generated for LKP are used to adaptively adjust the aggregate strength of local features; The grouping weights obtained after reshaping ( ); The c-th channel of the i-th pixel is the input feature, corresponding to the local pixel information of the RGB image; express of Neighborhood, enabling the extraction of local details.

[0107] For shortcut path operations, when the main path is inconsistent with the input feature channels, a 1×1 convolution is used to adjust the channel dimensions to ensure the feasibility of residual connections, represented as follows:

[0108] (3) Classification Head: A fully connected layer is used to convert the feature map output by the feature extraction layer into a 512-dimensional feature vector. This vector is then mapped to four output nodes through the fully connected layer, corresponding to the classification probabilities of the four traffic flow patterns. The output layer uses the Softmax activation function to ensure the normalization of the classification probabilities. Specifically, the weights of the last fully connected layer of the model are set as follows: (512 is the dimension of the feature vector after global average pooling), the bias is... The feature vector after global average pooling is Then the classification probabilities of the four traffic flow patterns The mathematical representations of (corresponding to smooth flow, synchronous flow, congested flow, and bottleneck flow respectively) are:

[0109] in, , and , The k-th row of the weight matrix It is the k-th element of the bias vector b.

[0110] (4) Model training: Cross-entropy loss function is used as training loss, AdamW is selected as the optimizer, the learning rate is initially set to 1e-4, and cosine annealing strategy is used to adjust the learning rate. The training rounds are 300 rounds, and L2 regularization (weight decay coefficient is 1e-5) is used to avoid model overfitting. During the training process, the accuracy of the validation set is used as the indicator to save the optimal model parameters. The final LS-ResNet classifier trained has an accuracy of no less than 95.3% on the test set, which is suitable for real-time traffic flow pattern recognition scenarios with only RGB input.

[0111] (5) Confidence determination: Define the classification result at the current time. The traffic flow pattern with the highest probability corresponds to a confidence level. The probability value for this form is mathematically represented as:

[0112]

[0113] Where k=1 corresponds to smooth flow, k=2 corresponds to synchronous flow, k=3 corresponds to congested flow, and k=4 corresponds to bottleneck flow; The higher the confidence level, the more reliable the classification result.

[0114] (6) Sliding window filtering is used to smooth the classification results: In order to avoid the noise interference of a single frame of data causing fluctuations in the classification results, a sliding window of length w (empirical value w=5, corresponding to 0.05 s) is used to smooth the classification results of consecutive w frames. The most frequently occurring traffic pattern within the window is statistically analyzed and used as the final classification result for the current moment. Mathematically, it is represented as:

[0115] Here, mode represents the mode operation. If there are multiple modes, the class with the highest mean confidence score is taken as the final result.

[0116] (7) Low confidence anomaly handling: Set a safety threshold Mean confidence score of all frames It can be calculated as

[0117] like If the current classification result is deemed unreliable, a re-matching mechanism is triggered: RGB images of the current environment (3 consecutive frames) are re-acquired and input into the LS-ResNet classifier for re-inference. Simultaneously, historical traffic flow data from the digital twin space in S1 is combined to assist in matching the current traffic flow pattern until the mean confidence level is satisfied. Output the final classification result.

[0118] Step S63: Based on the established four-state vehicle flow-specific scatterer parameter library, and according to the output classification results... Retrieve the parameter set corresponding to the traffic flow pattern : like (Smooth Flow), retrieve the smooth flow parameter set. ; like (Synchronization Stream), retrieve the synchronization stream parameter set ; like (Congestion Flow), retrieve congestion flow parameter set ; like (Bottleneck flow), retrieve bottleneck flow parameter set . Based on the retrieved parameter stream Set the parameters of the constructed channel as follows: Obtain the channel impulse response Channel impulse response based on real-time reconstruction The extracted channel quality parameters (such as signal-to-noise ratio SNR and channel capacity C) are used to dynamically allocate communication bandwidth, transmission power, and time slot resources, which can be mathematically represented as follows:

[0119] In the formula, B is the communication bandwidth (20 MHz). , For transmission power, For noise power spectral density; according to Dynamically adjusting the time slot allocation ratio ensures increased transmission rate in high-channel-quality scenarios and guaranteed transmission reliability in low-channel-quality scenarios. Based on the angular parameters of the scatterers in the channel impulse response, the beam direction and gain of the vehicle-mounted antenna are adjusted to ensure the beam is precisely aligned with the line-of-sight link or the main scattering link, reducing channel loss. Mathematically, this is expressed as:

[0120] In the formula, The antenna beam direction angle. To determine the optimal beam direction for the channel impulse response under different beam directions, the channel gain is maximized to achieve real-time beam scheduling and adaptation.

[0121] Finally, this invention enables high-fidelity link-level and system-level simulations of the communication protocols, resource scheduling algorithms, or end-to-end performance of 6G vehicle-to-everything (V2X) systems deployed in urban expressway scenarios, in order to verify the effectiveness of the design or optimize system parameters.

[0122] Example 2 This embodiment discloses a system for constructing vehicle network channels on urban expressways based on digital twins; like Figure 2 As shown, the urban expressway vehicle network channel construction system based on digital twins includes: The digital twin construction and data acquisition module is configured to: construct a digital twin space for urban expressways, and acquire multimodal data based on the digital twin space, including lidar point cloud data, image data, and channel impulse response data; The data processing and feature extraction module is configured to: preprocess the lidar point cloud data and channel impulse response data, and extract the three-dimensional spatial coordinates of the scatterer; The scatterer identification and classification module is configured to: identify scatterers in the environment based on preprocessed lidar point cloud data using a clustering algorithm; spatially match the identification results with the preprocessed scatterer coordinates; and classify scatterers into static scatterers and dynamic scatterers according to a preset distance threshold. The parameter library generation module is configured to: statistically analyze the parameter distributions of the static and dynamic scatterers for different traffic flow patterns, and fit the statistically obtained parameter distributions using a probability distribution model to generate a scatterer parameter library corresponding to different traffic flow patterns. The channel model construction module is configured to: construct a smart vehicle network channel model for urban expressways based on the scatterer parameter library; The traffic flow pattern recognition and channel adaptation module is configured to: input image data from the collected multimodal data into a trained LS-ResNet traffic flow pattern classifier, and output classification results for different traffic flow patterns; retrieve the corresponding parameter set from the scatterer parameter library based on the classification results of the traffic flow pattern, and solve the channel impulse response of the current traffic flow pattern based on the constructed intelligent vehicle-to-everything (V2X) channel model for urban expressways, and adapt the V2X communication system based on the channel impulse response.

[0123] Example 3 The purpose of this embodiment is to provide a computer-readable storage medium.

[0124] A computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps in a method for constructing a vehicle network channel for an urban expressway based on digital twins as described in Example 1.

[0125] Example 4 The purpose of this embodiment is to provide an electronic device.

[0126] An electronic device includes a memory, a processor, and a program stored in the memory and executable on the processor. When the processor executes the program, it implements the steps in the method for constructing a vehicle network channel for an urban expressway based on digital twins as described in Embodiment 1.

[0127] The steps and methods involved in the apparatuses of Embodiments 2, 3, and 4 above correspond to those in Embodiment 1. For specific implementation details, please refer to the relevant description section of Embodiment 1. The term "computer-readable storage medium" should be understood as a single medium or multiple media including one or more instruction sets; it should also be understood as including any medium capable of storing, encoding, or carrying an instruction set for execution by a processor and enabling the processor to perform any of the methods in this invention.

[0128] Those skilled in the art will understand that the modules or steps of the present invention described above can be implemented using general-purpose computer devices. Optionally, they can be implemented using computer-executable program code, thereby allowing them to be stored in a storage device for execution by a computer device, or they can be fabricated as separate integrated circuit modules, or multiple modules or steps can be fabricated as a single integrated circuit module. The present invention is not limited to any particular combination of hardware and software.

[0129] While the specific embodiments of the present invention have been described above in conjunction with the accompanying drawings, this is not intended to limit the scope of protection of the present invention. Those skilled in the art should understand that various modifications or variations that can be made by those skilled in the art without creative effort based on the technical solutions of the present invention are still within the scope of protection of the present invention.

Claims

1. A method for constructing vehicle network connectivity channels on urban expressways based on digital twins, characterized in that, include: A digital twin space for urban expressways is constructed, and multimodal data is collected based on the digital twin space. The multimodal data includes lidar point cloud data, image data, and channel impulse response data. The lidar point cloud data and channel impulse response data are preprocessed, and the three-dimensional spatial coordinates of the scatterer are extracted. Based on the preprocessed lidar point cloud data, scattering objects in the environment are identified through clustering algorithms; the identification results are spatially matched with the preprocessed scattering object coordinates, and the scattering objects are classified into static scattering objects and dynamic scattering objects according to a preset distance threshold; For different traffic flow patterns, the parameter distributions of the static and dynamic scatterers are statistically analyzed, and the statistically obtained parameter distributions are fitted using a probability distribution model to generate a scatterer parameter library corresponding to different traffic flow patterns. Based on the aforementioned scatterer parameter library, a smart vehicle network channel model for urban expressways is constructed. The image data from the collected multimodal data is input into the trained LS-ResNet traffic flow pattern classifier, and the classification results of different traffic flow patterns are output. Based on the classification results of traffic flow patterns, the corresponding parameter set is retrieved from the scatterer parameter library, and the channel impulse response of the current traffic flow pattern is solved based on the constructed intelligent vehicle-to-everything (V2X) channel model for urban expressways. The vehicle-to-everything (V2X) communication system is then adapted based on the channel impulse response.

2. The method for constructing urban expressway vehicle network connectivity channels based on digital twins as described in claim 1, characterized in that, The construction of the digital twin space for urban expressways includes: Use 3D modeling tools to build a geometric model of the urban expressway to scale, and configure the corresponding electromagnetic material parameters for the objects in the model; Based on traffic flow theory, traffic flow patterns are divided into smooth flow, synchronous flow, congested flow, and bottleneck flow. The vehicle density, speed range, and movement rules of each traffic flow pattern are defined, and corresponding vehicle trajectory files are generated. The geometric model and vehicle trajectory file are synchronously imported into the perception simulation platform and electromagnetic simulation platform, and a unified spatiotemporal reference is configured to achieve simulation data alignment.

3. The method for constructing urban expressway vehicle network connectivity channels based on digital twins as described in claim 1, characterized in that, Based on the preprocessed lidar point cloud data, scattering objects in the environment are identified through clustering algorithms; The identification results are spatially matched with the preprocessed scatterer coordinates, and the scatterers are classified into static scatterers and dynamic scatterers based on a preset distance threshold, including: Density-based spatial clustering is performed on the preprocessed lidar point cloud data to divide the point cloud into multiple independent clusters, and each cluster is classified as a static object or a dynamic vehicle based on its geometric dimensions. For each scatterer coordinate extracted from the channel data, calculate the minimum spatial distance between it and all identified static objects and dynamic vehicles. If the minimum spatial distance is less than the preset matching threshold, the attribute of the scatterer is marked as the attribute of the object closest to it; otherwise, the scatterer is marked as an unidentified scatterer and excluded in subsequent modeling, thereby completing the static and dynamic classification of the scatterer.

4. The method for constructing urban expressway vehicle network connectivity channels based on digital twins as described in claim 1, characterized in that, For different traffic flow patterns, the parameter distributions of the static and dynamic scatterers are statistically analyzed, and a probability distribution model is used to fit the statistically obtained parameter distributions to generate a scatterer parameter library corresponding to different traffic flow patterns, including: For each traffic flow pattern among smooth flow, synchronous flow, congested flow, and bottleneck flow, the key parameter distribution of static and dynamic scatterers in all communication links is statistically analyzed. The key parameters include at least the number of scatterers, propagation distance, arrival / departure angle, and power delay spectrum. A Gaussian mixture model is used to fit the cumulative distribution function of the number of scatterers and the propagation distance, and a Gaussian distribution is used to fit the statistical distribution of the arrival / departure angle; The parameters of the fitted probability distribution model are stored according to traffic flow patterns, forming a scatterer parameter library containing parameter sets specific to four traffic flow patterns.

5. The method for constructing urban expressway vehicle network connectivity channels based on digital twins as described in claim 1, characterized in that, Based on the aforementioned scatterer parameter library, a smart vehicle network connectivity channel model for urban expressways is constructed, including: Based on the traffic flow pattern of the target simulation scenario, the corresponding parameter set is selected from the parameter library, and the static and dynamic scatterers and scattering cluster parameters of the communication link at the initial moment are randomly generated according to the statistical distribution law of each parameter. Based on the visible region method, the generation, survival and extinction of scattering clusters during the communication duration are simulated to generate a channel coefficient sequence with time non-stationary characteristics. Based on the generated parameters and channel coefficients, a complete channel impulse response is constructed, including line-of-sight components, ground reflection components, static non-line-of-sight components, and dynamic non-line-of-sight components.

6. The method for constructing urban expressway vehicle network connectivity channels based on digital twins as described in claim 1, characterized in that, The image data from the collected multimodal data is input into the trained LS-ResNet traffic flow pattern classifier, which outputs classification results for different traffic flow patterns, including: The real-time acquired RGB image is input into the LS-ResNet classifier. The RGB image is preprocessed and its size is normalized to a preset resolution before being input into the trained LS-ResNet classifier. The LS-ResNet classifier performs preliminary feature extraction and dimensionality compression on the input image through the Stem layer, and outputs the first feature map; The first feature map is subjected to deep feature extraction through multiple residual blocks of the feature extraction layer. Each residual block extracts the global spatial distribution features and local detail features of the traffic flow through the main path of LS convolution, and performs residual connection through short-circuit path to output the second feature map. The second feature map is converted into a feature vector by the global average pooling layer in the classification head, and then mapped to four output nodes by a fully connected layer. The classification probability of different traffic flow patterns is calculated by the Softmax activation function. The traffic flow pattern corresponding to the maximum classification probability is taken as the classification result of the current frame, and the confidence level corresponding to the classification result is recorded.

7. The method for constructing urban expressway vehicle network connectivity channels based on digital twins as described in claim 1, characterized in that, Adapting the vehicle-to-everything (V2X) communication system based on the aforementioned channel impulse response includes: Based on the channel impulse response reconstructed in real time, channel quality parameters are extracted, and communication bandwidth, transmission power, and time slot resources are dynamically allocated. The beam direction and gain of the vehicle-mounted antenna are adjusted based on the angular parameters of the scatterer in the channel impulse response.

8. A system for constructing vehicle network channels for urban expressways based on digital twins, characterized in that: include: The digital twin construction and data acquisition module is configured to: construct a digital twin space for urban expressways, and acquire multimodal data based on the digital twin space, including lidar point cloud data, image data, and channel impulse response data; The data processing and feature extraction module is configured to: preprocess the lidar point cloud data and channel impulse response data, and extract the three-dimensional spatial coordinates of the scatterer; The scatterer identification and classification module is configured to: identify scatterers in the environment based on preprocessed lidar point cloud data using a clustering algorithm; spatially match the identification results with the preprocessed scatterer coordinates; and classify scatterers into static scatterers and dynamic scatterers according to a preset distance threshold. The parameter library generation module is configured to: statistically analyze the parameter distributions of the static and dynamic scatterers for different traffic flow patterns, and fit the statistically obtained parameter distributions using a probability distribution model to generate a scatterer parameter library corresponding to different traffic flow patterns. The channel model construction module is configured to: construct a smart vehicle network channel model for urban expressways based on the scatterer parameter library; The traffic flow pattern recognition and channel adaptation module is configured to: input image data from the collected multimodal data into a trained LS-ResNet traffic flow pattern classifier, and output classification results for different traffic flow patterns; retrieve the corresponding parameter set from the scatterer parameter library based on the classification results of the traffic flow pattern, and solve the channel impulse response of the current traffic flow pattern based on the constructed intelligent vehicle-to-everything (V2X) channel model for urban expressways, and adapt the V2X communication system based on the channel impulse response.

9. A computer-readable storage medium having a program stored thereon, characterized in that, When executed by the processor, the program implements the steps in the method for constructing a vehicle network channel for urban expressways based on digital twins as described in any one of claims 1-7.

10. An electronic device comprising a memory, a processor, and a program stored in the memory and executable on the processor, characterized in that, When the processor executes the program, it implements the steps in the method for constructing urban expressway vehicle network channels based on digital twins as described in any one of claims 1-7.

Images