Internet of vehicles channel prediction method in complex traffic scene and related equipment

Abstract

The invention relates to the technical field of Internet of Vehicles communication, and discloses an Internet of Vehicles channel prediction method in a complex traffic scene and related equipment. The method comprises the following steps: constructing an urban road scene of an Internet of Vehicles channel, simulating in the urban road scene to obtain a V2I communication result, and acquiring scene data through the V2I communication result; carrying out preprocessing and forward noise addition on the scene data, and then carrying out side information embedding to obtain channel data with spatial-temporal characteristics and a key dependency relationship; and inputting the channel data into the trained reverse denoising model, generating a predicted optimal channel index, and predicting the Internet of Vehicles channel in the complex traffic scene by using the predicted optimal channel index. According to the method, the conditional diffusion model is adopted, the channel prediction problem is effectively solved through implicit state evolution, the limitation of depending on lengthy formula derivation or a basic nonlinear model is overcome, and the complex nonlinear mapping relation between the channel state and the communication channel can be accurately captured.

Description

Technical Field

[0001] The present invention relates to the technical field of vehicle - to - everything (V2X) communication, and specifically provides a method and related devices for predicting V2X channels in complex traffic scenarios. Background Art

[0002] As a key driving force for intelligent transportation, vehicle - to - everything (V2X) communication technology significantly improves the overall intelligence level of the transportation system through the efficient interconnection and collaboration among vehicles, roads, and the cloud. Among them, vehicle - to - infrastructure (V2I) communication technology, as a core component of V2X, provides rich sensing information and intelligent decision - making support through seamless cooperation between vehicles and infrastructure, thereby optimizing road resource allocation and ensuring the smooth operation of traffic flow.

[0003] In V2I communication, channel prediction technology plays a crucial role. By predicting the channel state at the next moment in real - time, this technology effectively addresses the challenges brought about by the high mobility of vehicles and meets the stringent requirements of V2X communication for low latency and high reliability. However, despite the fact that channel prediction technology has become a research hotspot and made a series of progress, there are still many deficiencies in existing research, which are specifically manifested in the following aspects: First of all, although the channel prediction method based on formula derivation performs well in static or low - speed scenarios, its prediction accuracy is severely restricted in complex and changeable traffic environments. Such methods rely on complex mathematical modeling and statistical analysis but fail to fully consider real - world factors such as vehicle lane - changing behavior, multi - vehicle occlusion effects, and vegetation shielding interference, resulting in insufficient prediction ability for communication link stability.

[0004] Secondly, although the channel prediction method based on machine learning simplifies the prediction process and improves efficiency, it shows certain limitations in dealing with non - linear mappings in complex scenarios. Such methods learn the rules of channel changes from historical data by training models. However, in the face of the complex changes in channel states in the traffic system, existing models are difficult to fully capture these non - linear features, thus reducing the prediction accuracy.

[0005] Furthermore, the channel prediction method based on deep learning shows relatively high prediction accuracy in some scenarios, especially in dealing with complex non - linear relationships. However, existing technologies usually rely on simplified V2I communication assumptions and fail to effectively integrate various influencing factors in real - world scenarios, such as the impact of different weather conditions and traffic flow density on the communication link. This further limits the applicability of deep - learning methods in different environments.

[0006] In summary, with the increasing dynamic variability and environmental complexity of traffic scenarios, the channel quality fluctuates frequently, severely restricting the reliability and real-time performance of communication. Therefore, there is an urgent need to design a channel prediction method with stronger robustness and environmental adaptability to cope with the complex and changeable traffic environment and improve the stability and reliability of V2I communication. Summary of the Invention

[0007] In order to overcome the defects existing in the above-mentioned prior art, the purpose of the present invention is to provide a vehicle-to-infrastructure (V2I) channel prediction method and related devices in complex traffic scenarios to solve the technical problem of insufficient prediction accuracy in high-dynamic, multipath effect, and occlusion scenarios in the prior art.

[0008] The present invention is implemented through the following technical solutions: In the first aspect, the present invention provides a V2I channel prediction method in complex traffic scenarios, including: Construct a city road scenario for the V2I channel, simulate the V2I communication results under the city road scenario, and collect scenario data through the V2I communication results; Embed side information into the scenario data after preprocessing and forward noise addition to obtain channel data with spatio-temporal features and key dependency relationships; Input the channel data into a trained denoising autoencoder to generate a predicted optimal channel index, and predict the V2I channel in complex traffic scenarios using the predicted optimal channel index.

[0009] Preferably, the city road scenario includes vehicle density, RSU distribution, and weather scenario.

[0010] Preferably, the scenario data includes the optimal channel index value, communication signal time data, vehicle position information, and environmental information, where the environmental information includes road topology, traffic flow, and occlusion distribution.

[0011] Furthermore, the preprocessing of the scenario data includes: Perform data partitioning and data processing on the scenario data; The process of data partitioning includes: Collect the optimal channel index values into a data set and partition it into a truly known part, a pretend unknown part, and a truly unknown part; the truly known part and the pretend unknown part are used for model training, and the truly unknown part is used for model application; The process of data processing includes: Align the communication signal time data, clean and normalize the vehicle position information, and record the corresponding environmental information.

[0012] Further, forward noise addition is to perform forward noise addition processing on the pretended unknown part and the real unknown part to obtain forward noise-added data.

[0013] Furthermore, side information embedding is to perform side information embedding on the real known part and the forward noise-added data, where the side information includes communication signal time data and the position information of the vehicle; The specific process of the side information embedding includes: Input the real known part and the forward noise-added data into the convolutional layer for preliminary feature extraction, and embed the time information into the preliminary features through the fully connected layer and the convolutional layer; Extract the time-dependent features in the data through the time series extraction layer, and capture the law of the channel state changing with time according to the time-dependent features; The feature extraction layer extracts the time-critical features through multi-layer convolution and feature fusion; The position information of the vehicle is embedded into the preliminary features through the extended convolution, and the high-order spatial features are extracted; Fuse the time-critical features and the high-order spatial features and input them into the gated activation unit, and then output the channel data with spatio-temporal features and key dependencies.

[0014] Furthermore, input the channel data into the trained reverse denoising model to generate the predicted optimal channel index, including: Input the channel data after side information embedding into the reverse denoising model, gradually remove the noise, generate the predicted optimal channel index of the pretended unknown part, and optimize the trained reverse denoising model by calculating the error between the denoised data and the real data; Input the pure noise data into the trained reverse denoising model to generate the predicted optimal channel index of the real unknown part, which is used to indicate the channel state at the next moment.

[0015] In a second aspect, the present invention also provides a vehicle-to-infrastructure (V2I) channel prediction system in a complex traffic scenario, including: An urban road scene simulation and acquisition module, which is used to construct an urban road scene of the V2I channel, simulate the V2I communication result under the urban road scene, and acquire the scene data through the V2I communication result; A scene data prediction and processing module, which is used to perform side information embedding on the scene data after preprocessing and forward noise addition to obtain channel data with spatio-temporal features and key dependencies; A channel index prediction module, which is used to input the channel data into the trained reverse denoising model to generate the predicted optimal channel index, and predict the V2I channel in the complex traffic scenario with the predicted optimal channel index.

[0016] In a third aspect, the present invention further provides a mobile terminal, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the steps of the vehicle-to-internet (V2I) channel prediction method in the complex traffic scenario as described above are implemented.

[0017] In a fourth aspect, the present invention further provides a computer-readable storage medium storing a computer program, which when executed by a processor implements the steps of the vehicle-to-internet (V2I) channel prediction method in the complex traffic scenario as described above.

[0018] Compared with the prior art, the present invention has the following beneficial technical effects: The present invention provides a vehicle-to-internet (V2I) channel prediction method in a complex traffic scenario. By constructing an urban road scenario for the V2I channel and simulating the V2I communication results in this scenario, scene data highly relevant to the complex traffic scenario can be effectively obtained. Preprocessing and forward noise addition operations are performed on the collected scene data to eliminate feature scale differences, improve data quality, and enhance model convergence efficiency. The forward noise addition provides a unified and controllable initial distribution condition for the reverse denoising process. Through side information embedding, channel data with spatio-temporal features and key dependencies is extracted. The present invention solves the problem of dynamic modeling of wireless channels at the access network (AN) level in network slicing. By constructing an urban road scenario, simulating V2I communication, and optimizing channel prediction, the best channel index can be predicted to more accurately describe the channel state change, thus significantly improving the prediction accuracy. The present invention is connected before and after with the access network security authentication and jointly serves the integrity of the slice end-to-end architecture, ensuring the stability and security of the network slice in the access link.

[0019] Furthermore, the true known part and the pretended unknown part are used for model training. The pretended unknown part simulates the data loss situation that the model may encounter in actual applications, and the pretended unknown part also retains the true results for evaluating the model performance. During the training process, the model can perform in-depth analysis and learning from the true results of the true known part and the pretended unknown part through self-supervised learning methods, and gradually master the ability to accurately infer unknown information from partial known information. This mechanism not only restores the common data loss problems in the real scenario but also constructs a scientific and reliable basis for model training and evaluation, significantly enhancing the adaptability and generalization ability of the model in complex traffic scenarios.

[0020] Furthermore, the pretended unknown part plays a role in simulating unknown situations during model training. After forward noise addition processing, the data will exhibit more diverse change patterns. During the training process, when the model faces this pretended unknown data with noise, it needs to deeply explore the stable features and potential rules inherent in the data to accurately distinguish the differences between noise interference and real channel change signals. This prompts the model to learn more essential channel features, avoiding simply memorizing the surface patterns in the training data, thereby optimizing the model's learning ability for complex channel characteristics and more accurately coping with the dynamic changes in the channel environment.

[0021] Furthermore, the present invention introduces a side information module, which significantly enhances the model's time series prediction ability by embedding time and location information. Traditional channel prediction methods usually ignore the time variation and the dynamics of driving behavior, resulting in a significant decline in their prediction performance in variable traffic environments. By inputting time and location information as side information into the model, the present invention fully explores the time series correlation of the channel, enabling the model to adaptively capture the dynamic changes in driving behavior. This strategy significantly improves the model's prediction performance in complex traffic scenarios, enhances the comprehensiveness and accuracy of channel prediction, and is particularly prominent in environments with high uncertainty. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] Figure 1 is a flowchart of the vehicle-to-everything (V2X) channel prediction method in a complex traffic scenario according to an embodiment of the present invention; Figure 2 is a schematic diagram of an urban road scenario of a vehicle-to-everything (V2X) channel according to an embodiment of the present invention; Figure 3 is a schematic diagram of a V2I communication prediction algorithm model according to an embodiment of the present invention; Figure 4 is a bar chart of the comparison of CRPS values of different experimental schemes under different weather conditions and traffic flows according to an embodiment of the present invention; Figure 5 is a schematic diagram of the principle of a vehicle-to-everything (V2X) channel prediction system in a complex traffic scenario according to an embodiment of the present invention; In the figure: 1. Urban road scenario simulation and acquisition module; 2. Scenario data prediction and processing module; 3. Channel index prediction module. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0023] In order to enable those skilled in the art to better understand the solution of the present invention, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0024] The object of the present invention is to provide a vehicle-to-everything (V2X) channel prediction method and related devices in a complex traffic scenario, so as to solve the technical problem of insufficient prediction accuracy in the prior art in high-dynamic, multipath effect, and occlusion scenarios.

[0025] The term explanations involved in the present invention are as follows: RSU: Road Side Unit; Nheads: N attention heads; Air-Sim: Air-Sim is an open-source cross-platform simulator based on a game engine, which can be used for physical and visual simulations of robots such as unmanned aerial vehicles and unmanned vehicles; WaveFarer: WaveFarer is a high-fidelity radar simulator that can consider multipath and scattering of structures and vehicles in the environment around the radar system, as well as critical atmospheric and scattering effects at frequencies up to and exceeding 100 GHz; its applications include simulating automotive driving scenarios, indoor sensors, and far-field radar cross-sections; Wireless InSite: Wireless InSite is a prediction tool used to understand the wireless coverage, channel multipath, and data throughput of 5G, 6G, and WiFi networks. Through advanced accelerated three-dimensional ray tracing and fast-ray-based alternative methods, as well as empirical models, the multipath channel characteristics in indoor, urban, and rural environments can be predicted efficiently and accurately. The dynamic scenario modeling of vehicles and pedestrians can capture the attenuation that changes over time, while the frequency scan includes broadband effects. The communication analysis function applies multiple-input multiple-output algorithms to channel prediction to estimate the coverage and throughput of wireless networks.

[0026] The following further describes the present invention in detail with reference to the accompanying drawings: Embodiment 1 Refer to Figure 1 , in an embodiment of the present invention, a vehicle-to-everything (V2X) channel prediction method in a complex traffic scenario is provided, including: Step 1, construct an urban road scenario of the V2X channel, simulate the V2I communication result under the urban road scenario, and collect scenario data through the V2I communication result; Specifically, the urban road scenario includes vehicle density, RSU distribution, and weather scenario.

[0027] Among them, the scenario data includes the best channel index value, communication signal time data, vehicle position information, and environmental information, where the environmental information includes road topology, traffic flow, and occlusion distribution.

[0028] Among them, according to Figure 2As shown, in the medium-density scenario, the scenario contains 10 cars, 3 buses and 11 RSUs; in the high-density scenario, the number of vehicles increases to 15 cars and 6 buses, while keeping the number of RSUs at 11. The weather scenarios include sunny, rainy and snowy days; 1000 time frames are collected for each scenario of the original data, including non-lost time frames (75%) and lost data (25%). Each non-lost time frame contains the signal-to-noise ratio of the top three channel index values of V2I communication, the serial number of the communication signal time frame, and the position coordinates of the vehicle and the RSU. Each lost time frame only contains the serial number of the communication signal time frame, and the position coordinates of the vehicle and the RSU.

[0029] Step 2: After preprocessing and forward noise addition to the scenario data, side information is embedded to obtain channel data with spatio-temporal features and key dependencies; Specifically, according to Figure 3 As shown, the preprocessing of the scenario data includes: Partition and process the scenario data; The process of the data partition includes: The set of the best channel index values collected is used as a data set and divided into a truly known part , a pretended unknown part and a truly unknown part; the truly known part and the pretended unknown part are used for model training, and the truly unknown part is used for model application; The process of the data processing includes: Align the communication signal time data, clean and normalize the position information of the vehicle, and record the corresponding environmental information.

[0030] Among them, the non-lost time frames are randomly divided into a truly known part and a pretended unknown part , the lost time frames are used as the truly unknown part, and the truly known part is used for training (80%) and validation (10%), the pretended unknown part is used for testing (10%), and the lost time frames are used to apply the trained denoising model. Align the communication signal time frames, and align and normalize the vehicle position coordinates according to the position coordinates of the RSU.

[0031] Specifically, according to Figure 3 As shown, forward noise addition is to perform forward noise addition processing on the pretended unknown part and the truly unknown part to obtain forward noise-added data.

[0032] Among them, for the pretended unknown part after preprocessing Perform 50 forward noise addition processes to obtain forward noise-added data The true known part No noise addition process is performed

[0033] Among them, the expression of the forward noise addition process is as follows

[0034] Among them is a function of the noise addition intensity constant

[0035] The noise addition intensity is interpolated from 0.0001 to 0.5 according to the law of a quadratic function at each time step. The current state only depends on the state of the previous moment, forming a Markov chain process in the time series. Through multi-step noise addition, the channel state data gradually changes from the true distribution to the distribution of noise perturbations

[0036] Step 3: Input the channel data into the trained reverse denoising model to generate a predicted optimal channel index, and use the predicted optimal channel index to predict the vehicle network channel in a complex traffic scenario

[0037] Specifically, according to Figure 3 as shown, in the side information embedding, for the true known part and the forward noise-added data perform side information embedding, where the side information includes communication signal time data and vehicle position information; the specific process of the side information embedding is as follows First, input the true known part and the forward noise-added data into the convolutional layer for preliminary feature extraction, and the convolutional kernel size is 1

[0038] Next, embed the time information into the preliminary features in 128 dimensions through the fully connected layer and the convolutional layer

[0039] Then, further extract the time-dependent features in the data through the time series extraction layer to capture the law of the channel state changing with time. The time series extraction layer uses the multi-head attention mechanism (the nheads parameter is 8). The feature extraction layer extracts the time-critical features related to channel prediction through multi-layer convolution and feature fusion. The vehicle position information is embedded into the preliminary features in 16 dimensions through the extended convolution operation to further extract the high-order spatial features in the data. The feature fusion fuses the time features and the spatial features to obtain the fused data

[0040] ​Finally, the gating activation unit is used to control the flow and activation of features, retain the features most important for channel prediction, and suppress irrelevant or redundant information. The fused data contains rich spatio-temporal features and key dependencies, providing high-quality input for subsequent denoising and channel prediction.

[0041] Specifically, the channel data after side information embedding is reversely denoised through a denoising function. The denoising function learns to recover the original signal from the noisy data, and the total number of output channels is 128. The channel data after side information embedding is input into the denoising model, and noise is gradually removed 50 times in total to generate denoised data with 1 output channel, and the denoised data is calculated and the real data The error between them is used to optimize the denoising model.

[0042] Specifically, based on the trained denoising model, the best channel index is generated from pure noise data for the real unknown positions. The shape of the pure noise data is (16, 3, 1000), where 16 is the batch size, 3 is the first three channel indices of the signal-to-noise ratio, and 1000 is the time frame length. The pure noise data is input into the trained denoising model to generate a channel index, which is used to indicate the channel direction at the next moment. The specific meaning of the channel index is that the 180-degree plane of the RSU parallel to the road is evenly divided into 128 parts, and the channel index represents one of the directions. For example, the index values of 0 and 127 represent the leftmost and rightmost directions respectively. The channel index can accurately represent the communication direction and further improve the performance of the communication system.

[0043] To verify the beneficial effects of the proposed conditional diffusion model in the channel prediction task, a series of simulation experiments were carried out based on a public dataset The present invention was experimentally compared with the typical representative algorithm Support Vector Machine (SVM) in machine learning and the typical representative algorithm Long Short-Term Memory (LSTM) in deep learning.

[0044] Data source: The dataset uses Air-Sim and WaveFarer to collect multi-modal perception data, and uses Wireless InSite to collect communication data. The dataset realizes a communication-aware integrated system by deeply integrating and precisely aligning Air-Sim, WaveFarer and WirelessInSite. In addition, The dataset also covers various weather conditions, multiplexing frequency bands and different times of the day.

[0045] Hardware environment and software platform: The hardware environment uses an Intel Xeon W-3235 12-core processor and an RTX 3090 GPU, and the programming language used is Python 3.9.

[0046] To ensure the reliability of the experimental results and the scientific nature of model evaluation, five-fold cross-validation is used to divide the dataset. In each fold of validation, the non-missing data is randomly divided into a training set (80% × 70%), a validation set (80% × 30%), and a test set (20%). Each sample generates 100 probability points and takes the median as the final prediction result to enhance the stability and accuracy of the prediction, thus effectively reducing the influence of outliers and ensuring that the model can provide more reliable predictions when facing uncertainty or extreme data. Considering the possible real missing problems in the actual channel state data, a self-supervised learning mechanism is introduced in the experiment. By setting two cases of missing probabilities of 0.1 and 0.3, the influence of different degrees of data missing on the model performance is simulated, so as to more comprehensively evaluate the robustness of the model.

[0047] To simulate the actual vehicle networking scenario, different vehicle densities and RSU distributions are set to reflect the typical urban road communication environment. In the medium-density scenario, the scenario contains 10 cars, 3 buses, and 11 RSUs; in the high-density scenario, the number of vehicles increases to 15 cars and 6 buses, while keeping the number of RSUs at 11. Through these two different density scenario settings, the adaptability of the model under different traffic conditions can be comprehensively evaluated. In addition, to simulate the real road conditions, two weather scenarios of rainy and snowy days are introduced. In the rainy day scenario, the rainfall is set to 50 mm / hr and the road humidity is 100%; in the snowy day scenario, the snowfall is set to 10 mm / hr and the road humidity is also 100%. At the same time, to more accurately reflect the influence of bad weather on channel propagation, the experiment also models the physical space and electromagnetic space parameters under different weather conditions. For example, in rainy weather, the ambient temperature is set to 22.2 °C and the humidity is 100%; in the snowy day scenario, the temperature is set to -10 °C and the humidity is 20%.

[0048] In the communication parameter settings, the experiment adopted an integrated design of communication and sensing to ensure the accuracy of channel prediction and the efficiency of model training. Specifically, each RSU and vehicle were equipped with 128 and 32 antenna units respectively to support high-precision beamforming and channel prediction. In terms of frequency band selection, the experiment simultaneously simulated the communication performance of the Sub-6 GHz band and the millimeter wave band. In the Sub-6 GHz band, the carrier frequency was set to 5.9 GHz and the communication bandwidth was 20 MHz; in the millimeter wave band, the carrier frequency was set to 28 GHz and the communication bandwidth was 2 GHz. Through the comprehensive comparison of the communication performance of different frequency bands, the experiment further verified the adaptability and performance of the model in a multi-band environment.

[0049] The summary of the simulation experiment parameter settings is shown in Table 1: Table 1 Simulation Experiment Parameter Settings for Channel Prediction Based on Conditional Diffusion Model

[0050] According to Figure 4 As shown, under medium traffic density on sunny days, the Continuous Ranked Probability Score (CRPS) of CSDM was 0.184, significantly lower than 0.675 of SVM and 0.445 of LSTM, with improvements of approximately 72.8% and 58.7% respectively; under high traffic density on sunny days, CSDM still performed outstandingly, with a CRPS value of 0.232, compared with SVM (0.683) and LSTM (0.448), with improvements of approximately 66.2% and 48.5% respectively; under medium traffic density on rainy days, the CRPS of CSDM was 0.258, significantly better than 0.689 of SVM and 0.490 of LSTM, with improvements of approximately 62.5% and 47.4% respectively; while under medium traffic density on snowy days, the CRPS of CSDM was 0.392, still lower than 0.681 of SVM and 0.535 of LSTM, with improvements of approximately 42.5% and 26.7% respectively. These results indicate that CSDM can effectively improve the accuracy of channel prediction in environments of sunny days, rainy days, or snowy days, outperforming the SVM and LSTM methods.

[0051] The time performance of the model is an important indicator to measure its practicality. Especially in the vehicle-to-everything (V2X) scenario, real-time performance is a core requirement for the channel prediction model. Therefore, the experiment separately recorded the single-sample training time and the single-channel prediction time, and quantitatively analyzed the time performance of the model in combination with the size and type of the input data. The results are shown in Table 2. The experimental results show that the input data of the model includes timestamps, historical best channel indices, and location information, and the size of a single piece of data is 25 Bytes. In the training stage, the training time for a single sample is 2.411 ms, indicating that the model can maintain high efficiency when training each piece of data. In the inference stage, the time for a single-channel prediction is only 0.098 ms, and the model can complete the prediction in a short time, fully demonstrating its high computational efficiency and application potential. This result verifies that the proposed channel prediction model based on CSDM can not only ensure efficient training but also quickly complete real-time channel prediction, laying a good foundation for practical deployment.

[0052] Table 2 Time Performance of the Channel Prediction Model Based on the Conditional Diffusion Model

[0053] In summary, a vehicle-to-everything (V2X) channel prediction method in a complex traffic scenario provided by the present invention can more efficiently capture the non-linear relationship between the channel state and spatio-temporal factors by introducing the historical true known channel direction index as the conditional input, overcoming the limitations of traditional methods in high-dynamic scenarios. Specifically, the channel prediction problem can be transformed into a time series prediction problem, that is, by learning the implicit patterns and spatio-temporal variation rules in the historical time series data, the channel direction index at a future time can be inferred.

[0054] The introduction of the side information module in the present invention provides more accurate and rich additional information for the conditional diffusion model, especially showing significant advantages in mining time correlation and dynamic driving behavior. First, the time information embedding integrates information such as timestamps and time intervals into the model through fully connected layers and convolutional layers, enhancing the modeling ability of time dependence and enabling the model to more accurately capture the law of channel state change over time. Second, the spatial information embedding integrates spatial information such as vehicle location and road topology into the model through convolutional layers and expansion layers, improving the adaptability of the model to dynamic driving behavior. Finally, feature fusion integrates time features and spatial features and then passes through a gated activation unit to generate high-quality side information embedding data, providing more comprehensive additional information for channel prediction.

[0055] The present invention exhibits excellent robustness in multiple scenarios, mainly due to the accurate prediction of non-linear relationships by the model and the selection of embedded information in the side module. In high-dynamic scenarios, the conditional diffusion model can capture the rapid changes in the channel state, ensuring the stability of the prediction results. Timestamps can accurately record the temporal variation law of the channel state, providing key information for the model in the time dimension. Compared with other time-related information (such as time intervals or time windows), timestamps can more directly reflect the specific time points of the channel state, thus more effectively capturing the instantaneous change characteristics of the channel state. Vehicle position information can directly reflect the dynamic changes of the vehicle in space, providing key features for the model in the spatial dimension. Environmental factors related to weather may interfere with the performance of some sensors or the raw data collected, while the acquisition of vehicle position information mainly relies on satellite signal transmission, which has high environmental adaptability and can maintain stable working performance under most meteorological conditions (including rainy and snowy days). In addition, by combining vehicle position and time data, the model can further map out other important space-related information, such as vehicle speed, acceleration, and driving direction, etc.

[0056] The present invention adopts a conditional diffusion model (Conditional Score-based Diffusion Model, CSDM), which effectively solves the channel prediction problem through implicit state evolution, overcomes the limitations of relying on long formula derivations or basic non-linear models, and can accurately capture the complex non-linear mapping relationship between the channel state and the communication channel. The conditional diffusion model has significant advantages in dynamic modeling and can more accurately simulate the changes in the channel state according to the input conditions, thus significantly improving the prediction accuracy.

[0057] The present invention conducts systematic experiments on the proposed channel prediction method for the first time under various weather conditions and different traffic flow densities. Traditional channel prediction models are usually tested in idealized environments, ignoring the variable actual application scenarios, resulting in a significant decline in their performance in complex environments. The present invention designs an experimental environment covering different weather conditions (such as sunny, rainy, and snowy days) and traffic densities (such as medium density and high density), comprehensively verifying the robustness and stability of the model under various actual conditions. The experimental results show that the proposed method can maintain excellent prediction performance in multiple scenarios, fully demonstrating the wide adaptability and reliability of the method in complex real-world environments.

[0058] Embodiment 2 According to Figure 5 as shown, the present invention also provides a vehicle-to-everything (V2X) channel prediction system in a complex traffic scenario, including: The urban road scene simulation acquisition module 1 is used to construct the urban road scene of the vehicle-to-infrastructure (V2I) communication channel, simulate the V2I communication results under the urban road scene, and collect the scene data through the V2I communication results; The scene data prediction and processing module 2 is used to perform side information embedding on the scene data after preprocessing and forward noise addition to obtain channel data with spatio-temporal features and key dependency relationships; The channel index prediction module 3 is used to input the channel data into a trained denoising autoencoder in reverse to generate a predicted optimal channel index, and use the predicted optimal channel index to predict the V2I communication channel in a complex traffic scene.

[0059] Embodiment 3 The present invention also provides a mobile terminal, including a memory, a processor, and a computer program stored in the memory and executable on the processor, such as a V2I communication channel prediction program in a complex traffic scene.

[0060] When the processor executes the computer program, the steps of the above-mentioned V2I communication channel prediction method in a complex traffic scene are implemented, for example: Construct the urban road scene of the V2I communication channel, simulate the V2I communication results under the urban road scene, and collect the scene data through the V2I communication results; Perform side information embedding on the scene data after preprocessing and forward noise addition to obtain channel data with spatio-temporal features and key dependency relationships; Input the channel data into a trained denoising autoencoder in reverse to generate a predicted optimal channel index, and use the predicted optimal channel index to predict the V2I communication channel in a complex traffic scene.

[0061] Alternatively, when the processor executes the computer program, the functions of the above-mentioned modules in the system are implemented, for example: The urban road scene simulation acquisition module 1 is used to construct the urban road scene of the V2I communication channel, simulate the V2I communication results under the urban road scene, and collect the scene data through the V2I communication results; The scene data prediction and processing module 2 is used to perform side information embedding on the scene data after preprocessing and forward noise addition to obtain channel data with spatio-temporal features and key dependency relationships; The channel index prediction module 3 is used to input the channel data into a trained denoising autoencoder in reverse to generate a predicted optimal channel index, and use the predicted optimal channel index to predict the V2I communication channel in a complex traffic scene.

[0062] Exemplarily, the computer program may be divided into one or more modules / units, which are stored in the memory and executed by the processor to implement the present invention. The one or more modules / units may be a series of computer program instruction segments capable of performing specific functions, and these instruction segments are used to describe the execution process of the computer program in the mobile terminal.

[0063] For example, the computer program may be divided into a urban road scene simulation and acquisition module 1, a scene data prediction and processing module 2, and a channel index prediction module 3; The specific functions of each module are as follows: The urban road scene simulation and acquisition module 1 is used to construct an urban road scene of a vehicle-to-infrastructure (V2I) channel, simulate V2I communication results in the urban road scene, and acquire scene data through the V2I communication results; The scene data prediction and processing module 2 is used to perform side information embedding on the scene data after preprocessing and forward noise addition to obtain channel data with spatio-temporal features and key dependency relationships; The channel index prediction module 3 is used to input the channel data into a trained reverse denoising model to generate a predicted optimal channel index, so as to predict the V2I channel in a complex traffic scene using the predicted optimal channel index.

[0064] The mobile terminal may be a computing device such as a desktop computer, a notebook, a palm computer, and a cloud server. The mobile terminal may include, but is not limited to, a processor and a memory.

[0065] The processor may be a central processing unit (CPU), or may also be other general-purpose processors, a digital signal processor (DSP), an application specific integrated circuit (ASIC), a field-programmable gate array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor may be a microprocessor or the processor may also be any conventional processor, etc. The processor is the control center of the mobile terminal, and connects various parts of the entire mobile terminal using various interfaces and lines.

[0066] The memory may be used to store the computer program and / or module. The processor realizes various functions of the mobile terminal by running or executing the computer program and / or module stored in the memory, and by calling the data stored in the memory.

[0067] The memory may mainly include a program storage area and a data storage area. Among them, the program storage area may store an operating system, application programs required for at least one function (such as a sound playback function, an image playback function, etc.); the data storage area may store data created according to the use of the mobile phone (such as audio data, phone book, etc.). In addition, the memory may include a high-speed random access memory, and may also include a non-volatile memory, such as a hard disk, a memory, a plug-in hard disk, a smart media card (SMC), a secure digital (SD) card, a flash card, at least one magnetic disk storage device, a flash memory device, or other volatile solid-state storage devices.

[0068] Embodiment 4 The present invention also provides a computer-readable storage medium storing a computer program, and when the computer program is executed by a processor, the steps of the method for predicting a vehicle networking channel in a complex traffic scenario are implemented.

[0069] If the modules / units integrated in the mobile terminal are implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium.

[0070] Based on such an understanding, to implement all or part of the processes in the above method, the present invention can also be completed by instructing relevant hardware through a computer program. The computer program can be stored in a computer-readable storage medium. When the computer program is executed by a processor, the steps of the above aggregation reinforcement learning resource scheduling method can be implemented. Among them, the computer program includes computer program code, and the computer program code can be in the form of source code, object code, an executable file, or some intermediate form, etc.

[0071] The computer-readable medium may include: any entity or device capable of carrying the computer program code, a recording medium, a USB flash drive, a mobile hard disk, a magnetic disk, an optical disc, a computer memory, a read-only memory (ROM), a random access memory (RAM), an electrical carrier signal, a telecommunication signal, and a software distribution medium, etc.

[0072] It should be noted that the content included in the computer-readable medium can be appropriately increased or decreased according to the requirements of legislation and patent practice in the jurisdiction. For example, in some jurisdictions, according to legislation and patent practice, the computer-readable medium does not include electrical carrier signals and telecommunication signals.

[0073] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to the above embodiments, those of ordinary skill in the art should understand that: it is still possible to modify the specific implementation manners of the present invention or make equivalent substitutions, and any modification or equivalent substitution that does not depart from the spirit and scope of the present invention should be covered within the protection scope of the present invention.

Claims

1. A vehicle-to-everything (V2X) channel prediction method in a complex traffic scenario, characterized in that, Including: Construct a urban road scenario for the vehicle-to-internet (V2I) communication channel, simulate the V2I communication results in the urban road scenario, and collect scenario data through the V2I communication results. Perform side information embedding on the preprocessed and forward-noised scenario data to obtain channel data with spatio-temporal features and key dependencies. Input the channel data into a trained denoising autoencoder to generate a predicted optimal channel index, and use the predicted optimal channel index to predict the V2I communication channel in complex traffic scenarios.

2. The vehicle networking channel prediction method in a complex traffic scenario according to claim 1, wherein, The urban road scenario includes vehicle density, roadside unit (RSU) distribution, and weather scenario.

3. The vehicle networking channel prediction method in a complex traffic scenario according to claim 1, wherein The scenario data includes the optimal channel index value, communication signal time data, vehicle location information, and environmental information, where the environmental information includes road topology, traffic flow, and obstacle distribution.

4. A vehicle-to-everything (V2X) channel prediction method in a complex traffic scenario according to claim 3, characterized in that, The preprocessing of the scenario data includes: Perform data partitioning and data processing on the scenario data. The process of data partitioning includes: Collect the optimal channel index values as a dataset and partition it into a truly known part, a pretend unknown part, and a truly unknown part; the truly known part and the pretend unknown part are used for model training, and the truly unknown part is used for model application. The process of data processing includes: Align the communication signal time data, clean and normalize the vehicle location information, and record the corresponding environmental information.

5. A vehicle networking channel prediction method in a complex traffic scenario according to claim 4, characterized in that, The forward noise addition is to perform forward noise addition on the pretend unknown part and the truly unknown part to obtain forward-noised data.

6. The vehicle network channel prediction method in a complex traffic scenario according to claim 5, wherein, The side information embedding is to perform side information embedding on the truly known part and the forward-noised data, where the side information includes communication signal time data and vehicle location information. The specific process of the side information embedding includes: Input the truly known part and the forward-noised data into a convolutional layer for preliminary feature extraction, and embed the time information into the preliminary features through a fully connected layer and a convolutional layer. Extract the time-dependent features in the data through a temporal extraction layer, and capture the law of the channel state changing over time according to the time-dependent features. The feature extraction layer extracts the time-critical features through multiple convolutional layers and feature fusion. The vehicle location information is embedded into the preliminary features through dilated convolution, and the high-order spatial features are extracted. Fuse the time-critical features and the high-order spatial features and input them into a gated activation unit, and then output the channel data with spatio-temporal features and key dependencies.

7. A vehicle networking channel prediction method in a complex traffic scenario according to claim 6, characterized in that The step of inputting the channel data into a trained denoising autoencoder to generate a predicted optimal channel index includes: Input the channel data after side information embedding into the denoising autoencoder, gradually remove the noise, generate the predicted optimal channel index for the pretend unknown part, and optimize the trained denoising autoencoder by calculating the error between the denoised data and the real data. Input pure noise data into the trained denoising autoencoder to generate the predicted optimal channel index for the truly unknown part, which is used to indicate the channel state at the next moment.

8. A vehicle networking channel prediction system in a complex traffic scenario, characterized in that, Including: An urban road scene simulation acquisition module, which is used to construct an urban road scene for a vehicle-to-infrastructure (V2I) communication channel, simulate V2I communication results under the urban road scene, and acquire scene data through the V2I communication results; A scene data prediction and processing module, which is used to perform side information embedding on the scene data after preprocessing and forward noise addition to obtain channel data with spatio-temporal features and key dependency relationships; A channel index prediction module, which is used to input the channel data into a trained denoising autoencoder in reverse to generate a predicted optimal channel index, and predict the V2I communication channel in a complex traffic scene using the predicted optimal channel index.

9. A mobile terminal, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the steps of the method for predicting the V2I communication channel in a complex traffic scene according to any one of claims 1-7.

10. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by the processor, it implements the steps of the method for predicting the V2I communication channel in a complex traffic scene according to any one of claims 1-7.

Images