Wireless communication spectrum space digital twinning method

Abstract

The invention discloses a wireless communication spectrum space digital twinning method, which comprises the following steps of: generating basic static channel characteristics under a given environment model and a base station deployment condition; wherein the static channel characteristics comprise a path loss distribution characteristic, a shadow fading characteristic and a multipath propagation characteristic; associating the static channel feature with target vehicle information, and generating a single-frame channel feature when a vehicle exists; and capturing dynamic evolution information of channel characteristics in a vehicle movement process, and expanding the single-frame channel characteristics into sequential dynamic channel characteristics to obtain a digital twinborn model of a dynamic scene spectrum space. According to the method, the influence of a dynamic blocking body can be further considered on the basis of static channel generation, channel feature generation in a dynamic environment is realized, and the reality sense and application applicability of digital twinning are remarkably improved.

Description

Technical Field

[0001] This invention relates to the field of network and wireless communication technology, and in particular to a method for creating a digital twin of the wireless communication spectrum space. Background Technology

[0002] With the rise of 5G / 6G networks, network optimization and application deployment increasingly rely on accurate perception of the channel environment. Traditional measurement-driven methods often require extensive field testing, which is both time-consuming and difficult to cover all scenarios. Spectrum space digital twins, however, aim to virtually generate channel characteristics (such as path loss, shadowing fading, and multipath structure) within a coverage area under given environmental and base station deployment conditions, providing support for network planning, interference management, and intelligent scheduling. This approach can significantly reduce field testing costs while providing prior support for dynamic resource management, intelligent beamforming, and network self-optimization, representing an important direction for the intelligent management of future wireless networks.

[0003] In dynamic environments, the channel characteristics of communication links change significantly over time due to the influence of moving obstacles such as vehicles. Experimental studies show that for mmWave channels, large vehicles (buses, trucks, etc.) can momentarily block the Loss-of-Stake (LoS) link, resulting in an additional loss of 20–40 dB and potentially causing a brief link outage. Failure to model and generate these dynamic effects will lead to significant deviations between wireless network simulation and optimization results and actual conditions.

[0004] Currently, many scholars have proposed digital twin methods for wireless communication spectrum space, which can be divided into the following types: 1) Traditional deterministic methods: Statistical channel models (such as WINNER II and QuaDRiGa) can describe path loss and shadow fading in static environments, but they are difficult to characterize the time-varying effects of dynamic obstacles such as vehicles.

[0005] 2) Ray tracing based methods: Ray-tracing (RT) can accurately calculate radio wave propagation in complex environments, including reflection, diffraction, and scattering, and has been widely used in constructing mmWave and V2X channel models. However, its computational cost is enormous, making it unsuitable for real-time dynamic channel feature generation.

[0006] 3) Deep learning-based methods: To reduce computational complexity, researchers have proposed using deep learning to approximate RT results, thereby achieving efficient spectral spatial mapping. Examples include using deep learning to predict millimeter-wave coverage maps and radio map generation methods based on generative adversarial networks (GANs). These methods can quickly generate channel feature distributions, but most only consider static environments and fail to capture effects such as occlusion and multipath evolution caused by vehicle motion.

[0007] Therefore, existing technologies are insufficient to meet the requirements of wireless communication spectrum space digital twins in dynamic environments. Summary of the Invention

[0008] The technical problem to be solved by the present invention is that, in view of the defects of the prior art, the present invention provides a method for digital twin of wireless communication spectrum space, so as to solve the problem that the prior art is difficult to meet the requirements of digital twin of wireless communication spectrum space in dynamic environment.

[0009] The technical solution adopted by this invention to solve the technical problem is as follows: In a first aspect, the present invention provides a method for digital twinning of wireless communication spectrum space, comprising: Given an environment model and base station deployment conditions, basic static channel characteristics are generated; wherein, the static channel characteristics include: path loss distribution characteristics, shadowing fading characteristics, and multipath propagation characteristics; The static channel features are associated with the target vehicle information to generate a single-frame channel feature when the vehicle is present. By capturing the dynamic evolution information of channel characteristics during vehicle movement, the single-frame channel characteristics are extended into time-series dynamic channel characteristics, resulting in a digital twin model of the dynamic scene's spectral space.

[0010] In one implementation, generating basic static channel characteristics under given environmental model and base station deployment conditions includes: The reflection and diffraction propagation process of electromagnetic waves is simulated using an open-source ray tracing simulation tool. The path gain in a static scene is calculated to obtain the static channel characteristics.

[0011] In one implementation, the electromagnetic wave reflection and diffraction propagation process is simulated using an open-source ray tracing simulation tool to calculate the path gain in a static scenario, thereby obtaining the static channel characteristics, including: Geographic data of the target area is extracted using an open street map; wherein, the geographic data includes: building location, building outline, building height, and terrain type; The geographic data is converted into a scene model in 3D software, and the dielectric constant and conductivity of the building materials are configured according to preset standards. The location of each base station is determined based on the coverage of the target area and the location of obstructions, and the key parameters of each base station are configured. Based on the dielectric constant, the conductivity, and the key parameters of each base station, the path gain in the static scenario is calculated to obtain the static channel characteristics.

[0012] In one implementation, associating the static channel features with target vehicle information to generate single-frame channel features when the vehicle is present includes: The static channel features, building height map, and target vehicle height map are input into the channel feature generation network. Global features are extracted through layer-by-layer convolution and pooling operations to obtain the macroscopic laws of environmental structure and signal attenuation. Spatial resolution is gradually restored through upsampling and feature concatenation operations. The multi-channel features are mapped to a single-channel channel feature map using the output layer to obtain the single-frame channel features when the vehicle is present.

[0013] In one implementation, capturing the dynamic evolution information of channel features during vehicle movement and extending the single-frame channel features into temporally sequenced dynamic channel features to obtain a digital twin model of the dynamic scene's spectral space includes: Construct a spatiotemporal joint modeling network; A multi-channel feature map sequence of T consecutive frames during vehicle movement is acquired, and the multi-channel feature map sequence is input into the spatiotemporal joint modeling network; wherein, the multi-channel feature map sequence includes: channel features of the vehicle scene, building height map, and target vehicle height map; Based on the spatiotemporal joint modeling network, spatial features are extracted, and information in the temporal dimension is preserved during the convolutional feature extraction process. This allows for the acquisition of a collaborative representation of channel features in both space and time, resulting in a digital twin model of the dynamic scene's spectral space.

[0014] In one implementation, constructing the spatiotemporal joint modeling network includes: The encoder of the spatiotemporal joint modeling network is obtained by integrating the LSTM structure onto the U-Net structure; A symmetrical upsampling and skip connection mechanism is set on the U-Net structure to obtain the decoder of the spatiotemporal joint modeling network.

[0015] In one implementation, the extraction of spatial features based on the spatiotemporal joint modeling network, while retaining temporal information during convolutional feature extraction, and obtaining a joint representation of channel features in space and time, includes: In the encoder part, when extracting spatial features, the memory state across time steps is retained through a gating mechanism, and the hidden state of each time step is passed from the previous frame to the current frame to realize the temporal memory of the vehicle occlusion effect. In the decoder section, a symmetric upsampling and skip connection mechanism is used to maintain the spatial sharpness of the predicted channel features in the vehicle boundary and shadow regions. By mapping the convolutional layer to the single-channel channel feature output layer, the channel feature value of each pixel is regressed to obtain the joint representation of the channel feature in space and time.

[0016] In a second aspect, the present invention provides a wireless communication spectrum space digital twin system, comprising: The static channel feature generation module is used to generate basic static channel features under given environmental model and base station deployment conditions; wherein, the static channel features include: path loss distribution features, shadowing fading features, and multipath propagation features; A single-frame channel feature generation module is used to associate the static channel features with the target vehicle information to generate a single-frame channel feature when the vehicle is present. The spatiotemporal joint modeling module is used to capture the dynamic evolution information of channel features during vehicle movement, and to extend the single-frame channel features into temporally sequenced dynamic channel features, thereby obtaining a digital twin model of the dynamic scene spectrum space.

[0017] Thirdly, the present invention provides a terminal, comprising: a processor and a memory, wherein the memory stores a wireless communication spectrum space digital twin program, and the wireless communication spectrum space digital twin program, when executed by the processor, is used to implement the operation of the wireless communication spectrum space digital twin method as described in the first aspect.

[0018] Fourthly, the present invention also provides a computer-readable storage medium storing a wireless communication spectrum space digital twin program, which, when executed by a processor, is used to implement the operation of the wireless communication spectrum space digital twin method as described in the first aspect.

[0019] The present invention, by employing the above technical solution, has the following effects: This invention, building upon static channel generation, further considers the impact of dynamic obstructions, enabling channel feature generation in dynamic environments and significantly improving the realism and applicability of digital twins. Furthermore, it drastically reduces computational overhead while maintaining high-precision prediction capabilities. By using deep learning to approximate RT output, real-time simulation in large-scale and complex dynamic scenarios becomes possible, facilitating network optimization, resource scheduling, and intelligent wireless management. Attached Figure Description

[0020] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on the structures shown in these drawings without creative effort.

[0021] Figure 1 This is a flowchart of the wireless communication spectrum space digital twin method in this invention.

[0022] Figure 2 This is a flowchart of the wireless communication spectrum space digital twin in the dynamic environment of this invention.

[0023] Figure 3 This is a schematic diagram illustrating the conversion of map data into a scene model in this invention.

[0024] Figure 4 This is a schematic diagram of the static channel characteristics of the scenario in this invention.

[0025] Figure 5 This is a schematic diagram illustrating the process of modeling the impact of vehicles on the channel in this invention.

[0026] Figure 6 This is a schematic diagram of the dynamic channel feature generation process in this invention.

[0027] Figure 7 This is a schematic diagram of the experimental results in this invention.

[0028] Figure 8 This is a functional schematic diagram of the terminal in one implementation of the present invention.

[0029] The objectives, features, and advantages of this invention will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0030] To make the objectives, technical solutions, and advantages of this invention clearer and more explicit, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.

[0031] Exemplary methods In dynamic environments, the channel characteristics of communication links change significantly over time due to the influence of moving obstacles such as vehicles. Experimental studies show that for mmWave channels, large vehicles (buses, trucks, etc.) can momentarily block the Loss-of-Stake (LoS) link, resulting in an additional loss of 20–40 dB and potentially causing a brief link outage. Failure to model and generate these dynamic effects will lead to significant deviations between wireless network simulation and optimization results and actual conditions.

[0032] Currently, many scholars have proposed digital twin methods for wireless communication spectrum space, which can be divided into the following types: 1) Traditional deterministic methods: Statistical channel models (such as WINNER II and QuaDRiGa) can describe path loss and shadow fading in static environments, but they are difficult to characterize the time-varying effects of dynamic obstacles such as vehicles.

[0033] 2) Ray tracing based methods: Ray-tracing (RT) can accurately calculate radio wave propagation in complex environments, including reflection, diffraction, and scattering, and has been widely used in constructing mmWave and V2X channel models. However, its computational cost is enormous, making it unsuitable for real-time dynamic channel feature generation.

[0034] 3) Deep learning-based methods: To reduce computational complexity, researchers have proposed using deep learning to approximate RT results, thereby achieving efficient spectral spatial mapping. Examples include using deep learning to predict millimeter-wave coverage maps and radio map generation methods based on generative adversarial networks (GANs). These methods can quickly generate channel feature distributions, but most only consider static environments and fail to capture effects such as occlusion and multipath evolution caused by vehicle motion.

[0035] Therefore, existing technologies are insufficient to meet the requirements of wireless communication spectrum space digital twins in dynamic environments.

[0036] To address the above-mentioned technical problems, this invention provides a method for digital twinning of wireless communication spectrum space, comprising: generating basic static channel features under given environmental models and base station deployment conditions; wherein the static channel features include: path loss distribution features, shadowing fading features, and multipath propagation features; associating the static channel features with target vehicle information to generate single-frame channel features when the vehicle is present; capturing the dynamic evolution information of the channel features during vehicle movement, and extending the single-frame channel features into temporally sequenced dynamic channel features to obtain a digital twin model of the spectrum space in a dynamic scene. This invention can further consider the influence of dynamic obstructions on the basis of static channel generation, realizing the generation of channel features in dynamic environments, significantly improving the realism and applicability of digital twins.

[0037] like Figure 1 As shown, this embodiment of the invention provides a method for creating a digital twin of the wireless communication spectrum space, comprising the following steps: Step S100: Under the given environment model and base station deployment conditions, generate basic static channel characteristics; wherein, the static channel characteristics include: path loss distribution characteristics, shadowing fading characteristics, and multipath propagation characteristics.

[0038] In this embodiment, a method for creating a digital twin of the wireless communication spectrum space in a dynamic environment is proposed. This method has the following two main characteristics: 1) Introduce vehicle motion modeling to generate its dynamic impact on path loss, multipath structure, and time-varying channel characteristics; 2) By approximating the RT results through deep learning models, dynamic channel feature generation can be achieved with both high accuracy and high efficiency.

[0039] like Figure 2 As shown, the method provided in this embodiment is mainly divided into three parts: static channel feature generation, vehicle-to-channel impact modeling, and dynamic channel feature generation. This embodiment integrates static spectrum features and vehicle motion information through three stages: static channel feature generation, vehicle-to-channel impact modeling, and dynamic channel feature generation, and finally outputs channel features that change with vehicle motion, thereby realizing a digital twin of the spectrum space in dynamic scenarios.

[0040] Static channel feature generation refers to the process of generating basic static channel features, including path loss distribution, shadowing fading, and multipath propagation, given an environmental model and base station deployment. This part essentially constructs the spectrum space foundation for the entire scenario, providing a digital twin of the spectrum space under static conditions. A high-fidelity foundation can be generated using RT (Real-Time Analysis), laying the groundwork for subsequent dynamic modeling.

[0041] Specifically, in one implementation of this embodiment, step S100 includes the following steps: Step S101: Simulate the reflection and diffraction propagation process of electromagnetic waves using an open-source ray tracing simulation tool, calculate the path gain in a static scenario, and obtain the static channel characteristics.

[0042] In one implementation of this embodiment, step S101 includes the following steps: Step S101a: Extract geographic data of the target area using an open street map; wherein, the geographic data includes: building location, building outline, building height, and terrain type; Step S101b: Convert the geographic data into a scene model in 3D software, and configure the dielectric constant and conductivity of building materials according to preset standards; Step S101c: Determine the location of each base station based on the target area coverage and the location of obstructions, and configure the key parameters of each base station; Step S101d: Calculate the path gain in the static scenario based on the dielectric constant, the conductivity, and the key parameters of each base station to obtain the static channel characteristics.

[0043] In this embodiment, firstly, static channel features free from vehicle interference are generated to provide a foundation for digital twin generation. Specifically, the method is as follows: Static channel characteristics are the foundation of the entire spectrum space digital twin, and their accuracy directly determines the precision of subsequent dynamic modeling. Therefore, this embodiment relies on the open-source ray tracing simulation tool Sionna to generate static channel characteristics of the scene. Sionna can accurately calculate channel characteristics such as path gain (PG) in static scenes by simulating the propagation processes of electromagnetic waves, such as reflection and diffraction, providing a reliable static reference for subsequent vehicle dynamic impact modeling and avoiding the problem of insufficient accuracy of traditional statistical models.

[0044] Generating static channel features using Sionna requires modeling the scenario and setting base station parameters.

[0045] In the scene modeling stage, the geographic data of the target area is first extracted using OpenStreetMap (OSM), including the location, outline, and height of buildings, as well as terrain types (such as roads and green spaces). Then, Blosm is used to convert the OSM data into a scene model in Blender. Simultaneously, the dielectric constant (ε) and conductivity (σ) of the building materials are set according to ITU standards, and the scene is exported in Mitsuba XML format as required by Sionna. These material parameters are directly used for electromagnetic propagation calculations in Sionna ray tracing. The effect of converting map data into a scene model in this embodiment is as follows: Figure 3 As shown.

[0046] For base station setup, the location of the base station in each scenario does not need to be fixed in the center of the scenario. It can be flexibly selected according to actual deployment needs (such as covering the target area, avoiding obstructions, etc.). Other key base station parameters, such as antenna type, can be selected as needed: omnidirectional antenna (0 dBi gain) or directional antenna (such as 6.3 dBi gain, 65° / 8° horizontal / vertical half-power beamwidth).

[0047] Finally, the scene file (Mitsuba format) generated by Blender is input into Sionna, and base station attributes are set to calculate the static channel characteristics of the scene. In this embodiment, the static channel characteristics of the scene are as follows: Figure 4 As shown.

[0048] Considering that the actual target area size may exceed the processing scale of the model for a single scene (subsequent network inputs need to be of a fixed scale), large scenes need to be segmented: if the target area (e.g., 2048m×2048m) exceeds 256m×256m, it is uniformly divided into N×N 256m×256m sub-scenes. At the same time, to reduce the difficulty of merging sub-scenes containing dynamic channel features, an overlap area of ​​25-50m is set between adjacent sub-scenes.

[0049] In this embodiment, a spectrum space basis is constructed for the entire scene using the above method, providing a spectrum space digital twin result under static conditions. A high-fidelity basis can be generated using RT, laying the foundation for subsequent dynamic modeling.

[0050] like Figure 1 As shown, this embodiment of the invention provides a method for creating a digital twin of the wireless communication spectrum space, comprising the following steps: Step S200: Associate the static channel features with the target vehicle information to generate a single-frame channel feature when the vehicle is present.

[0051] In this embodiment, modeling the impact of vehicles on the channel refers to focusing on the core interference of vehicles as the main dynamic obstruction (such as abrupt changes in path loss caused by obstruction and new multipath propagation introduced by reflections from the metal body). Static channel characteristics are correlated with vehicle-related information (such as location and attributes) to construct a mapping relationship of "static characteristics - vehicle impact - channel characteristics with vehicles present." By modeling the dynamic interference of vehicles on the channel, channel characteristics when vehicles are present are generated, providing a foundation for further capturing the continuous impact of vehicle movement.

[0052] Specifically, in one implementation of this embodiment, step S200 includes the following steps: Step S201: Input the static channel features, building height map, and target vehicle height map into the channel feature generation network. Extract global features through layer-by-layer convolution and pooling operations to obtain the macroscopic laws of environmental structure and signal attenuation. Gradually restore spatial resolution through upsampling and feature stitching operations. Use the output layer to map multi-channel features into a single-channel channel feature map to obtain the single-frame channel features when the vehicle is present.

[0053] In this embodiment, the vehicle-to-channel impact modeling module is a crucial link connecting the static channel basis with the dynamic scenario channel generation. The model learns the mapping relationship between static channel features and vehicle spatial distribution to generate channel features that include the vehicle presence effect, providing the basic input for subsequent dynamic sequence modeling.

[0054] The model's input and output dimensions are both 256×256 pixels, with each pixel corresponding to a 1m×1m actual ground resolution. Pixel values ​​represent the path gain or physical properties at the corresponding location. The input includes three types of information: 1) Static channel characteristics (1 channel): Channel characteristics of a car-free scenario obtained by Sionna simulation, used as a baseline; 2) Building height diagram (channel 1), used to describe the shading structure of a fixed environment; 3) Vehicle height map (1 channel): This map depicts the presence and attributes of vehicles in the scene. The vehicle height map uses the vehicle's top height (in meters) as the pixel value, with 0 for areas without vehicles. Different vehicle models exhibit variations in height distribution, and the height variation between the front and rear of the vehicle reflects its orientation and geometry. For example, a continuous gradient with a slightly lower front and a slightly higher body can be considered a characteristic of sedans, while areas with a higher and more uniform height distribution correspond to SUVs or trucks. In this way, the vehicle height map not only expresses the spatial occupancy of the vehicle but also implicitly contains vehicle type and orientation information, enabling the network to learn the differentiated impact of different vehicle models and postures on the channel.

[0055] The model output is channel features (1 channel, dB) including vehicle occlusion and reflection effects, consistent with the input resolution, and can be directly used for subsequent time series modeling. The modeling process for the vehicle's impact on the channel is as follows: Figure 5 As shown.

[0056] To achieve efficient pixel-level modeling, this embodiment employs the U-Net architecture as the core generative network. U-Net, with its multi-scale feature fusion and skip connection mechanism, can capture global propagation trends while preserving local details such as vehicles and buildings. Its contraction path (encoder) extracts global features through layer-by-layer convolution and pooling, revealing the macroscopic laws governing environmental structure and signal attenuation; the expansion path (decoder) gradually restores spatial resolution through upsampling and feature concatenation, ensuring clear power changes at vehicle boundaries in the output.

[0057] Specifically, in the shrinking path (encoder), each downsampling block consists of "3×3 convolution + BN (batch normalization) + ReLU activation" × 2 + "2×2 max pooling (stride 2)"; the number of input channels is 3 (static channel features, building height map, vehicle height map), the first convolutional layer maps the number of channels from 3 to 64, and the number of channels doubles with each subsequent downsampling (64→128→256→512→1024).

[0058] In the expansion path (decoder), each upsampling block consists of "2×2 transposed convolution (stride 2, number of channels halved) + feature concatenation with the corresponding layer of the contraction path (skip connection) + 3×3 convolution + BN + ReLU"×2; the skip connection reuses the high-resolution spatial features of the contraction path (such as vehicle position details) to avoid information loss caused by upsampling.

[0059] In the output layer, 1×1 convolutions are used to map the 64-channel features obtained from the decoder into single-channel channel feature maps, without using an activation function to maintain the continuity of the predicted values. The network uses mean squared error (MSE) as the main loss function to regress the predicted channel feature values.

[0060] In this embodiment, the "multi-scale feature fusion" capability of U-Net is utilized to transform the static interference of vehicles on the channel (such as the occlusion loss of vehicles in fixed positions and the material reflection effect) into features that can be input into U-Net (such as vehicle position and attribute feature maps). These features are linked with the static channel features. By extracting multi-scale environmental and vehicle features through the shrinking path of U-Net and combining the expanding path with jump connections to restore fine spatial information, a mapping relationship of "static channel - vehicle features - vehicle-occupied channel" is constructed to achieve accurate modeling of vehicle static interference.

[0061] like Figure 1 As shown, this embodiment of the invention provides a method for creating a digital twin of the wireless communication spectrum space, comprising the following steps: Step S300: Capture the dynamic evolution information of channel features during vehicle movement, extend the single-frame channel features into time-series dynamic channel features, and obtain a digital twin model of the dynamic scene spectrum space.

[0062] In this embodiment, dynamic channel feature generation refers to the real-time generation of channel features under vehicle motion scenarios, based on static channel twinning and vehicle impact modeling. This means generating time-varying channel characteristics in real time as the vehicle's trajectory and the scene evolve. This part introduces a temporal dimension to the vehicle's influence on the channel, capturing the dynamic evolution of channel features during vehicle motion (e.g., continuous position changes), including continuous dynamic effects such as interference position migration and real-time changes in multipath components caused by vehicle motion.

[0063] Specifically, in one implementation of this embodiment, step S300 includes the following steps: Step S301: Construct a spatiotemporal joint modeling network.

[0064] In one implementation of this embodiment, step S301 includes the following steps: Step S301a: Integrate the LSTM structure onto the U-Net structure to obtain the encoder of the spatiotemporal joint modeling network; Step S301b: Set a symmetrical upsampling and skip connection mechanism on the U-Net structure to obtain the decoder of the spatiotemporal joint modeling network.

[0065] Step S302: Obtain a multi-channel feature map sequence of T consecutive frames during vehicle movement, and input the multi-channel feature map sequence into the spatiotemporal joint modeling network; wherein, the multi-channel feature map sequence includes: channel features of the vehicle scene, building height map, and target vehicle height map; Step S303: Based on the spatiotemporal joint modeling network, spatial features are extracted. During the convolutional feature extraction process, information in the temporal dimension is retained to obtain the collaborative representation of channel features in space and time, thereby obtaining a digital twin model of the dynamic scene spectrum space.

[0066] In one implementation of this embodiment, step S303 includes the following steps: In step S303a, in the encoder part, when extracting spatial features, the memory state across time steps is retained through a gating mechanism, and the hidden state of each time step is passed from the previous frame to the current frame to realize the temporal memory of the vehicle occlusion effect. Step S303b: In the decoder section, the spatial sharpness of the predicted channel features in the vehicle boundary and shadow region is maintained based on the symmetric upsampling and skip connection mechanism. Step S303c: Map the convolutional layer to the single-channel channel feature output layer, regress the channel feature value of each pixel, and obtain the joint representation of the channel features in space and time.

[0067] In this embodiment, based on static channel twinning and vehicle impact modeling, the dynamic channel feature generation stage further realizes the continuous generation of channel features under vehicle motion scenarios, that is, given the vehicle's temporal trajectory, it predicts the evolution of the channel over time in real time. Unlike the static U-Net in the previous stage, this stage introduces a temporal modeling unit into the network structure, enabling the model to capture continuous channel changes caused by vehicle motion.

[0068] Specifically, the model embeds a Long Short-Term Memory (LSTM) network at the encoder end on top of the second-stage U-Net, forming a spatiotemporal joint modeling structure (LSTM-U-Net). This structure achieves coordinated spatial and temporal representation of channel features by preserving temporal information during convolutional feature extraction. The network input is a sequence of multi-channel feature maps for T consecutive frames (each frame with a 0.2s interval), each frame containing three channels of information, including channel features of the vehicle scene in frame T-1, building height, and vehicle height maps; the output is the corresponding channel feature sequence for the vehicle scene in frame T, with each frame having a resolution of 256×256 pixels, and the pixel value representing the channel feature parameter (in dB) at that location at time t. This structure can generate channel feature sequences in a temporally continuous manner during vehicle movement, providing a foundation for constructing a spectral space digital twin of dynamic wireless communication scenarios. The dynamic channel feature generation in this embodiment is as follows: Figure 6 As shown.

[0069] In the encoder section, an LSTM structure is integrated to enhance temporal awareness, building upon the "convolution + normalization + activation" module of the second-stage U-Net. Specifically, the original "3×3 convolution + BN + ReLU" module is replaced with a "3×3 convolution + LSTM + BN + ReLU" module, enabling each layer to retain the memory state across time steps while extracting spatial features through a gating mechanism. The hidden state of each time step is passed from the previous frame to the current frame, achieving temporal memory of vehicle occlusion effects. For example, when the vehicle moves from position A to position B, the occlusion information formed at point A in the previous frame is passed through the memory state, affecting the channel feature estimation at point B in the current frame, thus ensuring the continuity of the prediction results in the temporal dimension.

[0070] The decoder retains the structural design of the second-stage U-Net, continuing to employ a symmetrical upsampling and skip connection mechanism, and no longer embeds an LSTM module to avoid over-smoothing of features during spatial reconstruction, thus maintaining the spatial sharpness of the predicted channel features in vehicle boundaries and shadow regions. The network ultimately maps to a single-channel channel feature output layer via 1×1 convolutions, directly regressing the channel feature value for each pixel. During training, a multi-frame mean squared error (MSE) loss function is used to supervise and constrain the output and ground truth values ​​at consecutive time steps, simultaneously optimizing spatial accuracy and temporal consistency.

[0071] This embodiment, based on the single-frame channel characteristics in the presence of a vehicle, introduces an LSTM structure to extend the single-frame dynamic correction to time-series dynamic generation. By capturing the continuous influence of vehicle movement, it outputs a time-series channel characteristic sequence that updates in real time with vehicle movement. This method can output time-varying channel characteristics such as path loss in real time according to environmental changes and vehicle trajectory, enabling the spectrum space digital twin to not only have static environment accuracy but also reflect the instantaneous fluctuations of the channel in dynamic environments.

[0072] To verify the effectiveness of the method provided in this embodiment, the process of wireless communication spectrum space digital twin is illustrated below through experimental simulation: (1) Dataset: The Guangdong-Hong Kong-Macao Greater Bay Area was selected as the representative test area for the experiment. The area has a dense urban structure and diverse road types, which can reflect the typical characteristics of urban vehicle-to-everything (V2X) environments. 1000 400m×400m scenes were randomly selected from the data. Road, building and terrain information in the scenes were extracted based on OpenStreetMap (OSM) data. Blender was used to perform parametric modeling of building height and material to generate a high-fidelity geometric model that conforms to the characteristics of urban wireless propagation. The base station deployment referenced the vehicle-to-everything (V2X) communication scenario, with the working frequency band set to 5.9 GHz, antenna height of 6-8m, and coverage radius of approximately 150-300m.

[0073] Vehicle motion data is generated using SUMO (Simulation of Urban Mobility), simulating both morning and evening rush hours and low-flow traffic conditions based on urban road networks and traffic flow statistics. Vehicle density in each scenario ranges from 30 to 80 veh / km, and vehicle types include sedans, SUVs, buses, and trucks. Vehicle height information is mapped to corresponding pixels to generate a vehicle height map, and frame-level data is output in a time series format for the network's temporal input. Each scenario generates a 10-frame temporal sample sequence.

[0074] The ground truth of the channel features is obtained by calculating the path gain using Sionna to obtain the static basis of the "carless scenario". Then, the real ground truth of the PG of the "car-in scenario" is generated by adding vehicles, which is used for network training and evaluation.

[0075] (2) Evaluation index: In this embodiment, RMSE is used to measure the generation error. The smaller the generation error, the better the result.

[0076] ; in, and They represent True PG features and predicted PG features of location.

[0077] (3) Experimental Results and Analysis: The experimental results show (e.g.) Figure 7 As shown in the figure, the average RMSE of the test set is 3.31 dB, which proves that the proposed model can accurately reproduce the temporal changes of channel characteristics in dynamic scenarios. Figure 7The figure shows the RMSE variation curves of the model during continuous frame prediction. It can be observed that the prediction error is small in the initial stage (frame 1). As time progresses and the vehicle moves, the RMSE increases slightly, reflecting the increased channel dynamic complexity caused by the vehicle's occlusion position shifting. However, because the LSTM's temporal memory mechanism can continuously capture the feature evolution of the previous frame, the error growth trend flattens out after frame 4, indicating that the model has strong stability and temporal consistency during temporal prediction.

[0078] This embodiment achieves the following technical effects through the above technical solution: This embodiment can further consider the impact of dynamic obstructions on top of static channel generation, realizing channel feature generation in dynamic environments, significantly improving the realism and applicability of digital twins; moreover, this embodiment greatly reduces computational overhead while maintaining high-precision prediction capabilities. By using deep learning to approximate RT output, real-time simulation in large-scale and complex dynamic scenarios becomes possible, facilitating network optimization, resource scheduling, and intelligent wireless management.

[0079] Exemplary device Based on the above embodiments, the present invention also provides a wireless communication spectrum space digital twin system, comprising: The static channel feature generation module is used to generate basic static channel features under given environmental model and base station deployment conditions; wherein, the static channel features include: path loss distribution features, shadowing fading features, and multipath propagation features; A single-frame channel feature generation module is used to associate the static channel features with the target vehicle information to generate a single-frame channel feature when the vehicle is present. The spatiotemporal joint modeling module is used to capture the dynamic evolution information of channel features during vehicle movement, and to extend the single-frame channel features into temporally sequenced dynamic channel features, thereby obtaining a digital twin model of the dynamic scene spectrum space.

[0080] This embodiment achieves the following technical effects through the above technical solution: This embodiment can further consider the impact of dynamic obstructions on top of static channel generation, realizing channel feature generation in dynamic environments, significantly improving the realism and applicability of digital twins; moreover, this embodiment greatly reduces computational overhead while maintaining high-precision prediction capabilities. By using deep learning to approximate RT output, real-time simulation in large-scale and complex dynamic scenarios becomes possible, facilitating network optimization, resource scheduling, and intelligent wireless management.

[0081] Based on the above embodiments, the present invention also provides a terminal, the principle block diagram of which can be as follows: Figure 8 As shown.

[0082] The terminal includes: a processor, a memory, an interface, a display screen, and a communication module connected via a system bus; wherein, the processor of the terminal provides computing and control capabilities; the memory of the terminal includes a computer-readable storage medium and internal memory; the computer-readable storage medium stores an operating system and computer programs; the internal memory provides an environment for the operation of the operating system and computer programs in the computer-readable storage medium; the interface is used to connect to external devices; the display screen is used to display relevant information; and the communication module is used to communicate with a cloud server or other devices.

[0083] When executed by a processor, this computer program is used to implement the operation of the wireless communication spectrum space digital twin method.

[0084] It will be understood by those skilled in the art that Figure 8 The schematic diagram shown is merely a partial structural diagram related to the present invention and does not constitute a limitation on the terminal to which the present invention is applied. A specific terminal may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0085] In one embodiment, a terminal is provided, comprising: a processor and a memory, the memory storing a wireless communication spectrum space digital twin program, which, when executed by the processor, is used to implement the operation of the wireless communication spectrum space digital twin method as described above.

[0086] In one embodiment, a computer-readable storage medium is provided, wherein the computer-readable storage medium stores a wireless communication spectrum space digital twin program, which, when executed by a processor, is used to implement the operation of the wireless communication spectrum space digital twin method described above.

[0087] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile storage medium, and when executed, it can include the processes of the embodiments of the methods described above. Any references to memory, storage, database, or other media used in the embodiments provided by this invention can include both non-volatile and volatile memory.

[0088] In summary, this invention provides a digital twin method for wireless communication spectrum space, comprising: generating basic static channel features under given environmental models and base station deployment conditions; wherein the static channel features include: path loss distribution features, shadowing fading features, and multipath propagation features; associating the static channel features with target vehicle information to generate single-frame channel features when the vehicle is present; capturing the dynamic evolution information of the channel features during vehicle movement, and extending the single-frame channel features into temporally sequenced dynamic channel features to obtain a digital twin model of the spectrum space in a dynamic scene. This invention can further consider the influence of dynamic obstructions on the basis of static channel generation, realizing the generation of channel features in dynamic environments, significantly improving the realism and applicability of digital twins.

[0089] It should be understood that the application of the present invention is not limited to the examples above. Those skilled in the art can make improvements or modifications based on the above description, and all such improvements and modifications should fall within the protection scope of the appended claims.

Claims

1. A method for wireless communication spectrum space digital twinning, comprising: The application relates to a method for generating a digital twin model of a dynamic scene spectrum space. The method comprises the following steps: generating basic static channel features under a given environment model and base station deployment condition, wherein the static channel features comprise path loss distribution features, shadow fading features and multipath propagation features; associating the static channel features with target vehicle information to generate single-frame channel features in the presence of a vehicle; 2. The wireless communication spectrum spatial digital twin method of claim 1, wherein, capturing dynamic evolution information of channel features in the process of vehicle movement, extending the single-frame channel features into time-series dynamic channel features, and obtaining the digital twin model of the dynamic scene spectrum space. The method for generating basic static channel features under a given environment model and base station deployment condition comprises the following steps:

3. The wireless communication spectrum spatial digital twin method of claim 2, wherein, simulating the reflection and diffraction propagation process of electromagnetic waves based on an open-source ray tracing simulation tool, calculating the path gain in a static scene, and obtaining the static channel features. The method for simulating the reflection and diffraction propagation process of electromagnetic waves based on an open-source ray tracing simulation tool, calculating the path gain in a static scene, and obtaining the static channel features comprises the following steps: extracting geographical data of a target area from an open street map, wherein the geographical data comprises building positions, building outlines, building heights and terrain categories; converting the geographical data into a scene model in a three-dimensional software, and configuring the dielectric constant and electrical conductivity of building materials according to a preset standard; determining the position of each base station according to the coverage rate and the position of an obstacle of the target area, and configuring the key parameters of each base station; 4. The wireless communication spectrum spatial digital twin method of claim 1, wherein, calculating the path gain in the static scene according to the dielectric constant, the electrical conductivity and the key parameters of each base station, and obtaining the static channel features. The method for associating the static channel features with target vehicle information to generate single-frame channel features in the presence of a vehicle comprises the following steps:

5. The wireless communication spectrum spatial digital twin method of claim 1, wherein, inputting the static channel features, a building height map and a target vehicle height map into a channel feature generation network, extracting global features through layer-by-layer convolution and pooling operations, obtaining the macroscopic law of environmental structure and signal attenuation, gradually restoring the spatial resolution through upsampling and feature splicing operations, mapping the multi-channel features into a single-channel channel feature map through an output layer, and obtaining the single-frame channel features in the presence of a vehicle. The method for capturing dynamic evolution information of channel features in the process of vehicle movement, extending the single-frame channel features into time-series dynamic channel features, and obtaining the digital twin model of the dynamic scene spectrum space comprises the following steps: constructing a space-time joint modeling network; obtaining a multi-channel feature map sequence of T continuous frames in the process of vehicle movement, and inputting the multi-channel feature map sequence into the space-time joint modeling network; wherein the multi-channel feature map sequence comprises channel features of a vehicle scene, a building height map and a target vehicle height map; 6. The wireless communication spectrum spatial digital twin method of claim 5, wherein, extracting spatial features based on the space-time joint modeling network, retaining the information of the time dimension in the convolution feature extraction process, obtaining the collaborative representation of channel features in space and time, and obtaining the digital twin model of the dynamic scene spectrum space. The method for constructing the space-time joint modeling network comprises the following steps: integrating an LSTM structure on a U-Net structure to obtain an encoder of the space-time joint modeling network. The symmetrical upsampling and skip connection mechanism is arranged on the U-Net structure to obtain the decoder of the spatio-temporal joint modeling network.

7. The wireless communication spectrum spatial digital twin method of claim 5, wherein, The spatial feature is extracted based on the spatio-temporal joint modeling network, the information of the time dimension is reserved in the convolution feature extraction process, and the collaborative representation of the channel feature in space and time is obtained, including: In the encoder part, when the spatial feature is extracted, the memory state across time steps is reserved through the gating mechanism, the hidden state of each time step is transmitted from the previous frame to the current frame, and the time sequence memory of the vehicle occlusion effect is realized; In the decoder part, the symmetrical upsampling and skip connection mechanism is used to maintain the spatial sharpness of the predicted channel feature in the vehicle boundary and the shadow area; The single-channel channel feature output layer is mapped through the convolution layer, the channel feature value of each pixel is regressed, and the collaborative representation of the channel feature in space and time is obtained.

8. A wireless communication spectrum space digital twin system, comprising: It includes: A static channel feature generation module is configured to generate a basic static channel feature under a given environment model and base station deployment condition; wherein the static channel feature includes path loss distribution features, shadow fading features, and multipath propagation features; A single-frame channel feature generation module is configured to associate the static channel feature with target vehicle information to generate a single-frame channel feature when a vehicle exists; A spatio-temporal joint modeling module is configured to capture dynamic evolution information of the channel feature in the vehicle motion process, expand the single-frame channel feature into a time-series dynamic channel feature, and obtain a digital twin model of a dynamic scene frequency spectrum space.

9. A terminal, characterized by comprising: It includes: A processor and a memory, the memory stores a wireless communication frequency spectrum space digital twin program, and the wireless communication frequency spectrum space digital twin program is executed by the processor to implement the operations of the wireless communication frequency spectrum space digital twin method in any one of claims 1-7.

10. A computer-readable storage medium, characterized in that, The computer readable storage medium stores a wireless communication frequency spectrum space digital twin program, and the wireless communication frequency spectrum space digital twin program is executed by the processor to implement the operations of the wireless communication frequency spectrum space digital twin method in any one of claims 1-7.

Images