Channel switching method and vehicle

Abstract

This application relates to a channel switching method and vehicle, belonging to the field of communication technology. It includes preprocessing multi-source spatiotemporal data acquired during vehicle operation to obtain a spatiotemporal prediction input dataset; inputting the spatiotemporal prediction input dataset into a pre-trained neural network model to obtain channel quality parameters of the current channel and each candidate channel; if the channel quality parameters of the current channel trigger a degradation threshold, recording the trigger time and calculating the comprehensive channel score of the current channel and each candidate channel to determine the primary target channel; determining candidate switching times based on the trigger time; calculating the switching benefit of each candidate switching time to determine the optimal switching timing; combining the primary target channel and the optimal switching timing to generate a switching control command; and issuing the switching control command to the vehicle-mounted wireless communication module to switch the vehicle-mounted wireless communication module to the primary target channel. This achieves proactive planning and execution of channel switching before signal degradation occurs.

Description

Technical Field

[0001] This application relates to the field of communication technology, and in particular to a channel switching method and a vehicle. Background Technology

[0002] With the widespread adoption of in-vehicle smart terminals, in-vehicle wireless communication has become a core communication module supporting navigation, audio-visual entertainment, and vehicle-to-everything (V2X) data interaction. However, during dynamic driving, a vehicle's communication signal is susceptible to interference from surrounding buildings, other wireless devices operating on the same frequency, and complex electromagnetic environments, often resulting in signal attenuation, communication lag, or even disconnection. This severely impacts the user's in-vehicle network experience, making stable and continuous channel switching crucial.

[0003] Currently, existing methods mainly employ real-time responsive channel switching technologies, such as adaptive switching based on dynamic spectrum sensing, which triggers switching to other idle channels when a decline in the current channel quality is detected; and introducing artificial intelligence models to analyze the real-time channel status to improve the accuracy and response speed of switching decisions; some other solutions enhance signal stability through multi-protocol coordination or beamforming techniques to compensate for performance deficiencies in channel switching.

[0004] The aforementioned existing technologies generally suffer from the fundamental problems of passive response mechanisms and lagging handover, making it impossible to predict and avoid signal degradation before it occurs. Furthermore, existing solutions often rely on single real-time channel parameters, lacking the integration of multi-dimensional information such as vehicle status, spatial environment, and historical data, resulting in insufficient forward-looking decision-making. In addition, delays and interruptions during the handover process directly impact user experience, while methods that rely on additional hardware or complex protocols also incur additional cost and power consumption burdens. Summary of the Invention

[0005] This application addresses, to at least some extent, one of the technical problems in the related art.

[0006] Therefore, this application aims to provide a channel switching method and vehicle. By predicting channel quality parameters and determining the primary target channel and optimal switching time based on the comprehensive channel score after triggering the degradation threshold, a switching control command is generated to execute the channel switching. This realizes the proactive planning and execution of channel switching before signal degradation occurs, solving the problems of communication interruption and poor user experience caused by passive response and switching lag in the existing channel switching technology of vehicle wireless communication.

[0007] To achieve the above objectives, in a first aspect, this application provides a channel switching method applied to an in-vehicle controller, comprising: Acquire multi-source spatiotemporal data during vehicle operation, preprocess the multi-source spatiotemporal data, and obtain a spatiotemporal prediction input dataset; The spatiotemporal prediction input dataset is input into a pre-trained neural network model to obtain the channel quality parameters of the current channel and each candidate channel; If the channel quality parameters of the current channel trigger the degradation threshold, the trigger time is recorded, and based on the channel quality parameters, the comprehensive channel score of the current channel and each candidate channel is calculated. The main target channel is determined based on the comprehensive channel score. Candidate handover times are determined based on the trigger time, the handover benefit of each candidate handover time is calculated based on the comprehensive channel score, and the optimal handover timing is determined based on the handover benefit. Combining the primary target channel and the optimal handover timing, a handover control command is generated and sent to the vehicle-mounted wireless communication module to enable the vehicle-mounted wireless communication module to switch to the primary target channel.

[0008] In the above technical solution, multi-source spatiotemporal data is acquired and input into a neural network model to predict channel quality parameters. When the degradation threshold is triggered, the triggering time is recorded and a comprehensive channel score is calculated to determine the primary target channel and the optimal handover timing. Finally, a control command is generated to execute the handover. Because of the adoption of a prediction-based trigger judgment and benefit calculation mechanism, channel handover is no longer a passive remedy after signal degradation, but rather an active planning of handover timing based on predictions of future channel quality. This effectively avoids communication interruptions during the handover process and significantly improves the continuity and stability of vehicle wireless communication.

[0009] In some embodiments of this application, the spatiotemporal prediction input dataset is input into a pre-trained neural network model to obtain the channel quality parameters of the current channel and each candidate channel, including: Through the first branch of the neural network model, spatiotemporal feature extraction and fusion inference are performed on the time series data of the channel parameters of the candidate channels in the spatiotemporal prediction input data to obtain the channel quality parameters of each candidate channel. The channel quality parameters of the current channel are obtained by mapping the time series data of the channel parameters in the spatiotemporal prediction input data through the second branch of the neural network model.

[0010] In the above technical solution, the first branch of the neural network model is used to extract and fuse the temporal and spatial features of the candidate channel time series data to obtain the candidate channel quality parameters, while the second branch is used to directly map the current channel time series data to obtain the current channel quality parameters. Based on the parallel processing of the two branches, the acquisition of the current channel parameters does not require a complex feature extraction and reasoning process. This ensures the real-time nature of the current channel state and avoids unnecessary computational overhead. At the same time, the deep feature extraction of the first branch can fully explore the spatiotemporal variation patterns of the candidate channel.

[0011] In some embodiments of this application, the spatiotemporal feature extraction and fusion inference of the spatiotemporal prediction input data are performed through the first branch of the neural network model to obtain the channel quality parameters of each candidate channel, including: The spatiotemporal prediction input dataset is convolved through the first branch to extract multi-dimensional spatial features, and temporal features are extracted based on the multi-dimensional spatial features. A linear transformation is performed on the time-series characteristics to obtain the auxiliary channel quality parameters of each candidate channel; The time-series features are subjected to feature dimensionality reduction and nonlinear fusion to obtain a fused feature vector. Based on the fused feature vector, the core channel quality parameters of each candidate channel are obtained. By combining the core channel quality parameters and the auxiliary channel quality parameters, the channel quality parameters of each candidate channel are obtained.

[0012] In the above technical solution, the first branch performs convolution processing on the spatiotemporal prediction input dataset to extract multi-dimensional spatial features, and then extracts temporal features based on the spatial features. The temporal features are then subjected to linear transformation and feature dimensionality reduction fusion to obtain auxiliary parameters and core parameters respectively. This fully utilizes the ability of convolution operations to capture spatial correlations and the advantages of temporal modeling to analyze changing trends. At the same time, the branch processing separates key parameters from non-key parameters, which not only ensures the prediction accuracy of core parameters, but also reduces the computational cost of auxiliary parameters through lightweight linear transformation.

[0013] In some embodiments of this application, after obtaining the channel quality parameters of the current channel and each candidate channel, the process includes: Based on the spatial environment data in the multi-source spatiotemporal data, the road environment where the vehicle is located is identified, and the road environment identification result is obtained. Based on the road environment recognition results, the scene type in which the vehicle is located is determined; Based on the scenario type, a preliminary screening threshold is set, and the core channel quality parameters of each candidate channel are compared with the preliminary screening threshold to obtain channels that fail the preliminary screening. The unqualified channels in the initial screening are removed from each of the candidate channels, and channels are supplemented from the preset list of available channels based on the scenario type to obtain the final candidate channels.

[0014] In the above technical solution, the road environment where the vehicle is located is identified based on spatial environment data. Then, differentiated initial screening thresholds are set according to the scene type to screen the core channel quality parameters of the candidate channels. Channels that do not meet the requirements of the current environment are eliminated. Channels are dynamically supplemented from the preset list of available channels based on the same scene type. This ensures that the initial screening of channels closely matches the propagation characteristics of the physical environment in which the vehicle is currently located. This avoids misjudgments that may occur when using a uniform fixed threshold in different scenes. It ensures that the eliminated channels are indeed unsuitable due to environmental factors. At the same time, the scene-related supplementation mechanism maintains the sufficiency and quality of the number of candidate channels, thereby effectively improving the environmental adaptability and reliability of the candidate channel set on which subsequent switching decisions are based.

[0015] In some embodiments of this application, after obtaining the channel quality parameters of the current channel and each candidate channel, the method further includes: Based on the scenario type, set a degradation threshold corresponding to the channel quality parameters of the current channel; Compare the current channel quality parameters with the degradation threshold to determine whether a channel handover needs to be initiated. If the channel quality parameters of the current channel trigger the degradation threshold, a channel handover needs to be initiated, and the comprehensive channel score of the current channel and each candidate channel is calculated based on the channel quality parameters. If the channel quality parameters of the current channel do not trigger the degradation threshold, then there is no need to initiate a channel handover, and the current channel is maintained.

[0016] In the above technical solution, by setting a degradation threshold corresponding to the current channel quality parameters based on the scenario type, the current channel quality parameters are compared with the degradation threshold to determine whether a handover needs to be initiated. The subsequent scoring and handover process is only started when the threshold is triggered. This can make full use of the adaptive adjustment capability of the scenario type to the threshold, so that the timing of the handover trigger is more in line with the actual communication needs in different environments, avoiding false triggering or missed triggering caused by fixed thresholds. At the same time, the condition judgment filters out situations where no handover is needed.

[0017] In some embodiments of this application, based on the channel quality parameters, a comprehensive channel score for the current channel and each candidate channel is calculated, and a primary target channel is determined based on the comprehensive channel score, including: Based on the scenario type, determine the scenario-based benchmark range of the channel quality parameters, and configure scenario-based weights for the channel quality parameters; Based on the aforementioned scenario-based benchmark range, the channel quality parameters are standardized to obtain standardized values ​​of the channel quality parameters. Based on the scenario-based weights, the standardized values ​​are weighted and calculated to obtain the comprehensive channel score; The candidate channel with the highest comprehensive channel score is selected as the primary target channel.

[0018] In the above technical solution, by determining the scenario-based benchmark range of channel quality parameters based on scenario type and configuring scenario-based weights, the comprehensive channel score is obtained by weighted calculation after standardization of the parameters, and the candidate channel with the highest score is selected as the main target channel. This can deeply integrate scenario characteristics into the parameter processing and weight allocation process, so that the importance difference of different parameters in different environments can be reflected. The standardization process eliminates the influence of parameter dimensions and value ranges. Finally, the main target channel selected is the one with the best comprehensive performance in the current scenario.

[0019] In some embodiments of this application, candidate handover times are determined based on the trigger time, handover benefits for each candidate handover time are calculated based on the comprehensive channel score, and the optimal handover timing is determined based on the handover benefits, including: Based on the triggering time, a time window within a preset time range in the future is determined as the candidate switching time; Based on the comprehensive channel score of the primary target channel, the comprehensive channel score of the current channel, and the preset service impact coefficient, the handover benefit for each candidate handover moment is calculated. Candidate switching times where the switching benefit is greater than the benefit threshold are selected to obtain the constraint-satisfied times. Based on the constraint-satisfied times, the optimal switching timing is determined.

[0020] In the above technical solution, candidate handover times within the future time window are determined based on the trigger time. The handover benefits of each candidate time are calculated based on the comprehensive channel score of the primary target channel and the current channel, as well as the preset service impact coefficient. The optimal handover time is determined by screening the time when the benefit is greater than the threshold. This can incorporate the handover lead time and service impact into the quantitative calculation, ensuring that the selected handover time can effectively avoid the predicted degradation points and can be executed during the relatively low service hours. This ensures the success of the handover while minimizing the impact on the user experience.

[0021] In some embodiments of this application, the switching control command is sent to the vehicle-mounted wireless communication module to control the vehicle-mounted wireless communication module to perform channel switching, including: The actual quality parameters of the main target channel are collected in real time, and the actual quality parameters are compared with the degradation threshold to obtain the comparison result; The actual quality parameters are compared with the corresponding channel quality parameters to generate prediction bias data; Based on the comparison results and the prediction deviation data, the pre-trained neural network model is optimized.

[0022] In the technical solution, after the channel handover is executed, the actual quality parameters of the main target channel are collected in real time and compared with the degradation threshold to evaluate the handover effect. At the same time, the actual parameters are compared with the predicted parameters to generate prediction deviation data. Based on the comparison results and prediction deviation, the neural network model is optimized. The handover effect can be verified by actual data and the prediction error can be quantified. This allows the model to continuously learn the channel change patterns in the real environment and continuously adjust and optimize its own parameters, thereby gradually improving the prediction accuracy in long-term operation.

[0023] Secondly, this application provides a vehicle, including: The vehicle controller is used to execute the channel switching method described above to generate switching control commands; The vehicle-mounted wireless communication module is communicatively connected to the vehicle-mounted controller and is used to receive the switching control command and perform channel switching based on the switching control command.

[0024] In the above technical solution, the intelligent cockpit generates switching control commands by executing channel switching methods, and the vehicle wireless communication module with communication connection receives the commands and executes channel switching. This design separates the decision-making core from the execution unit. The intelligent cockpit uses its powerful computing capabilities to complete complex spatiotemporal predictions and decision calculations, while the vehicle wireless communication module focuses on executing specific channel switching operations. The two work together to ensure both the intelligence and real-time nature of decision-making, as well as the reliability and efficiency of execution.

[0025] In a third aspect, this application provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that the processor implements the aforementioned channel switching method when executing the computer program.

[0026] In the above technical solution, by embedding the channel switching method into the electronic device in software form, the device can autonomously complete the processing of multi-source spatiotemporal data, the prediction and evaluation of channel quality, and the generation of switching decisions based on its internal computing resources. It can realize the complete intelligent switching process locally without relying on external servers or cloud computing power, thereby significantly reducing the dependence on network communication and interaction latency. This ensures that the electronic device can maintain efficient channel management capabilities even in unstable network conditions or offline environments, effectively improving the communication stability and real-time decision-making of the device in complex and ever-changing environments.

[0027] As can be seen from the above technical solutions, additional aspects and advantages of this application will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of this application. Attached Figure Description

[0028] Fig. 1A flowchart illustrating a channel switching method provided in an embodiment of this application; Fig. 2 A flowchart illustrating another channel switching method provided in an embodiment of this application; Fig. 3 This is a schematic diagram of the internal structure of a vehicle provided for an embodiment of this application. Detailed Implementation

[0029] In this application, the terms "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., refer to a specific feature, structure, material, or characteristic described in connection with that embodiment or example, which is included in at least one embodiment or example of this application. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples. Moreover, without contradiction, those skilled in the art can combine and integrate the different embodiments or examples described in this specification, as well as the features of different embodiments or examples.

[0030] The present application will now be described in detail through exemplary embodiments. However, it should be understood that, without further description, elements, structures, and features in one embodiment may be advantageously incorporated into other embodiments. With the popularization of in-vehicle intelligent terminals, in-vehicle wireless communication modules have become core communication modules, supporting key functions such as navigation map updates, multimedia entertainment, and vehicle network data interaction. However, during vehicle operation, in-vehicle wireless communication signals are susceptible to the influence of complex environments, such as obstruction by surrounding buildings, interference from other wireless devices operating on the same frequency, and superposition interference from mobile operator signals. These factors can lead to signal attenuation, communication lag, or even disconnection, severely impacting the user experience.

[0031] Currently, the channel switching technology for vehicle wireless communication mainly focuses on real-time response optimization, the core idea of ​​which is to trigger adjustment after detecting signal degradation. Specifically, dynamic spectrum sensing and adaptive switching technology monitors parameters such as the signal-to-noise ratio and bit error rate of the current channel in real time. When the parameters deteriorate to the threshold, it switches to other idle channels. However, this technology is a passive response. There is a delay of 1 to 3 seconds from the detection of signal deterioration to the completion of the switch. Communication interruption is likely to occur during this period, which cannot meet the real-time data transmission requirements of vehicle wireless communication. Artificial intelligence-assisted channel state analysis technology introduces neural network models such as long short-term memory neural networks to improve the accuracy of channel state analysis. However, the model only analyzes real-time channel data and aims to optimize the response speed of handover decisions. It does not involve the prediction of future channel quality and has not yet escaped the nature of passive response. Multi-protocol coordination technology improves communication stability through redundant backup of vehicle wireless communication and mobile communication networks, but it increases hardware costs and power consumption, and does not solve the core problem of channel switching lag in vehicle wireless communication itself. Beamforming technology improves signal reception strength by adjusting the antenna direction, but its suppression effect on multipath interference and co-channel interference is limited, and it cannot cope with signal degradation caused by sudden changes in the path during driving, and still needs to rely on the switching mechanism to ensure communication.

[0032] In summary, existing channel switching technologies for vehicle wireless communication generally suffer from switching lag, which cannot be avoided in advance before signal degradation occurs, making it difficult to guarantee the stability of vehicle wireless communication and user experience. Current solutions for switching channels in vehicle wireless communication generally employ passive logic that triggers switching after signal degradation detection. This results in a significant time lag during the switching process, which can easily lead to data transmission interruptions and stuttering. Meanwhile, channel optimization relies solely on real-time channel parameters, resulting in a relatively limited data dimension. It fails to integrate key information such as vehicle driving status, spatial environment, and historical interference data, leading to insufficient accuracy and foresight in handover decisions. In addition, the relevant solutions lack a spatiotemporal prediction mechanism for changes in channel quality within a short time window in the future. They cannot accurately predict the risk of signal degradation on the driving path and can only passively deal with signal problems that have already occurred. This results in a poor user experience. The delay and disconnection during the switching process cause navigation map loading to be slow, online video to be interrupted, and vehicle network data interaction to fail, which is clearly perceived by users. Some multi-protocol collaboration solutions require the addition of communication modules, which increases hardware costs and vehicle power consumption, and is also more difficult to adapt to different vehicle models.

[0033] Based on this, this application proposes a channel switching method and vehicle. By predicting channel quality parameters and after triggering a degradation threshold, the method determines the primary target channel and the optimal switching timing based on a comprehensive channel score, generates a switching control command to execute the channel switching, and realizes proactive planning and execution of channel switching before signal degradation occurs. This solves the problems of communication interruption and poor user experience caused by passive response and switching lag in existing vehicle wireless communication channel switching technologies.

[0034] In the following, embodiments of this application will be described in detail with reference to the accompanying drawings.

[0035] As attachedFigs. 1-2 As shown in an illustrative embodiment of a channel switching method in this application, the channel switching method is applied to an in-vehicle controller and includes: S1: Acquire multi-source spatiotemporal data during vehicle operation, preprocess the multi-source spatiotemporal data, and obtain the spatiotemporal prediction input dataset; Furthermore, the multi-source spatiotemporal data includes vehicle operating status data, current channel parameter time series data, spatial environment data, and historical wireless communication signal heatmap data; Vehicle operating status data is collected in real time through onboard sensors and controller local area network bus, with a sampling frequency of 10Hz; the vehicle operating status data includes the vehicle's position, speed, direction and steering angle. The vehicle's position is obtained through a dual-mode positioning system of Global Positioning System and Beidou, with positioning accuracy controlled within 1 meter; the vehicle's speed is obtained through a vehicle speed sensor, with a speed measurement error not exceeding 0.5 km / h; the vehicle's direction is obtained through an inertial measurement unit, and the vehicle's steering angle is obtained through a steering angle sensor. The vehicle-mounted wireless communication chip installed in the vehicle's infotainment system collects real-time channel parameter timing data of the current channel at a sampling frequency of 20 Hz. The channel parameter timing data of the current channel includes throughput, delay, jitter, signal-to-noise ratio, and interference intensity, which are used to characterize the real-time communication quality of the current channel. The system interacts with the high-precision map service module via an Ethernet interface to obtain spatial environment data of the vehicle's current and future 3 to 5 seconds driving path. The spatial environment data includes road type, surrounding building density, and distribution information of obstructions such as tunnels or bridges that may cause signal blockage. The road type includes highways, urban roads, and rural roads. By accessing the local historical data storage unit, the historical wireless communication signal heat map data of the current driving segment can be retrieved. The historical wireless communication signal heat map data covers channel quality records under the same time period and weather conditions within the past three months. The data range includes the historical channel parameter time series data of the current channel, each candidate channel, and other available channels in the preset available channel list. At the same time, it is associated with the corresponding road type, vehicle operation status data, and historical switching records to provide empirical statistical basis for subsequent predictions. The current channel refers to the wireless communication channel that the vehicle is currently using and that carries real-time data transmission. Candidate channels refer to a predetermined number of channels selected from a predetermined list of available channels based on preset filtering conditions. These channels are used to calculate a comprehensive channel score to determine the primary target channel. Other available channels in the preset available channel list refer to the channel resource pool that is preset by the device or maintained by the system, excluding the current channel and candidate channels, and is used to supplement the selection based on the scenario type when the number of candidate channels is insufficient after the initial screening, so as to maintain the required number of candidate channels.

[0036] For example, when a vehicle is traveling at 80 km / m on a city road, the real-time location of the vehicle is collected as 116.25° east longitude and 40.03° north latitude, the speed is 72 km / m, the direction of travel is 15° north of east, and the turning angle is 2°. The current channel throughput was obtained from the vehicle-mounted wireless communication chip installed in the vehicle's infotainment system. The latency was 23ms, the jitter was 8ms, the signal-to-noise ratio was 28dBm, and the interference intensity was -65dBm. By interacting with the high-precision map service module via the Ethernet interface, information was obtained that a tunnel would be entered 500m ahead. The tunnel is 350m long, there are no tall buildings around the tunnel, and the road type is a highway. By accessing the local historical data storage unit, historical wireless communication signal heatmap data for the tunnel section during the same period over the past three months were obtained. The data shows that the signal-to-noise ratio drops by an average of about 10 dBm after entering the tunnel, and the interference intensity increases by an average of 5 dBm. Historical records also indicate that the section has triggered three channel switching events.

[0037] Furthermore, the acquired multi-source spatiotemporal data is preprocessed to ensure the quality, consistency, and reliability of the input data; First, since multi-source spatiotemporal data come from different acquisition devices and data interfaces, their original formats are different. For example, vehicle operation status data is usually obtained from the controller local area network bus in binary format, channel parameter time series data is recorded in decimal numerical form, and data from the high-precision map service module is transmitted in JSON format. By standardizing the format, the data in different formats mentioned above are uniformly converted into a standardized data format with timestamps. The timestamp accuracy is controlled within 1ms, which enables subsequent data alignment and fusion to be carried out on a unified data structure, eliminating the processing obstacles caused by differences in data formats. Secondly, for continuous data such as channel parameter time series data and vehicle speed, a sliding window averaging algorithm is used for noise reduction. The window size is set to 5 sampling points. By calculating the mean of the data within the window, random fluctuations caused by instantaneous interference or measurement noise are smoothed out, thereby extracting signal features that reflect the true trend of change. Meanwhile, the 3σ criterion is used to detect and remove outliers from the data. That is, the mean and standard deviation of the data are calculated, and data points that deviate from the mean by more than three times the standard deviation are identified as outliers and removed. Common anomalies include GPS positioning jumps caused by momentary satellite signal interruptions and sudden changes in channel parameters caused by transient communication interference. Through noise reduction and anomaly removal, noise components and erroneous information in the data are effectively filtered out, improving the purity and credibility of the dataset.

[0038] Finally, the vehicle operating status data, channel parameter time series data, spatial environment data, and historical wireless communication signal heat map data, which have undergone format standardization and noise reduction processing, are spatiotemporally correlated and aligned. In the time dimension, precise synchronization is performed based on the timestamp carried by each data to ensure that data from different data sources collected at the same time can be correctly associated; in the spatial dimension, the real-time location of the vehicle is used as the anchor point to match the spatial environment data and historical wireless communication signal heat map data with the current location of the vehicle to ensure that the distribution information of obstructions on the path ahead and historical statistical characteristics can accurately correspond to the location that the vehicle is about to arrive at. By aligning spatiotemporal correlations, a unified spatiotemporal prediction input dataset is generated. The spatiotemporal prediction input dataset is sorted by time series and contains multi-dimensional feature information related to vehicle wireless communication quality at the current moment and in historical time periods, providing a structurally complete and reliable data foundation for subsequent neural network model input.

[0039] S2: Input the spatiotemporal prediction input dataset into the pre-trained neural network model to obtain the channel quality parameters of the current channel and each candidate channel; In some embodiments, the spatiotemporal prediction input dataset is input into a pre-trained neural network model to obtain the channel quality parameters of the current channel and each candidate channel, including: By using the first branch of the neural network model, spatiotemporal features are extracted and fused inference is performed on the time series data of the channel parameters of the candidate channels in the spatiotemporal prediction input data to obtain the channel quality parameters of each candidate channel. By using the second branch of the neural network model, the channel parameter time series data of the current channel in the spatiotemporal prediction input data is mapped to obtain the channel quality parameters of the current channel.

[0040] The first branch of the neural network model extracts and fuses the temporal and spatial features of the candidate channel time series data to obtain the candidate channel quality parameters, while the second branch directly maps the current channel time series data to obtain the current channel quality parameters. Based on the parallel processing of the two branches, the acquisition of the current channel parameters does not require a complex feature extraction and reasoning process. This ensures the real-time nature of the current channel state and avoids unnecessary computational overhead. At the same time, the deep feature extraction of the first branch can fully explore the spatiotemporal variation patterns of the candidate channel.

[0041] In some embodiments, the spatiotemporal feature extraction and fusion inference of the spatiotemporal prediction input data are performed through the first branch of the neural network model to obtain the channel quality parameters of each candidate channel, including: The spatiotemporal prediction input dataset is convolved through the first branch to extract multi-dimensional spatial features, and then temporal features are extracted based on the multi-dimensional spatial features. A linear transformation is performed on the time-series characteristics to obtain the auxiliary channel quality parameters for each candidate channel; The temporal features are subjected to feature dimensionality reduction and nonlinear fusion to obtain a fused feature vector. Based on the fused feature vector, the core channel quality parameters of each candidate channel are obtained. By combining the core channel quality parameters and the auxiliary channel quality parameters, the channel quality parameters of each candidate channel are obtained.

[0042] The first branch performs convolution processing on the spatiotemporal prediction input dataset to extract multi-dimensional spatial features, and then extracts temporal features based on the spatial features. The temporal features are then subjected to linear transformation and feature dimensionality reduction fusion to obtain auxiliary parameters and core parameters respectively. This fully utilizes the ability of convolution operations to capture spatial correlations and the advantages of temporal modeling to analyze changing trends. At the same time, the branch processing separates key parameters from non-key parameters, which not only ensures the prediction accuracy of core parameters, but also reduces the computational cost of auxiliary parameters through lightweight linear transformation.

[0043] Furthermore, the spatiotemporal prediction input dataset is preprocessed into 12-dimensional feature time series data and input into a pre-trained neural network model. Through the dual-branch parallel processing architecture inside the neural network model, the throughput, latency, jitter, signal-to-noise ratio and interference intensity of the current channel and each candidate channel are predicted every 1 second in the next 3 to 5 seconds, so as to obtain the channel quality parameters of the current channel and each candidate channel. Among them, the 12-dimensional feature time series data includes time dimension data, spatiotemporal correlation dimension data, spatial dimension data, and historical spatiotemporal dimension data; The time-dimensional data includes the throughput, latency, jitter, and signal-to-noise ratio of the current channel and each candidate channel within the past 5 seconds; Spatiotemporal correlation data includes the vehicle's position, speed, and direction; Spatial dimension data includes road type, surrounding building density, and information on the distribution of obstructions; Historical spatiotemporal data includes historical wireless communication signal heatmap data; After the 12-dimensional feature time series data is input into the neural network model, it is first processed through a convolutional layer to extract the spatial correlation between features of different dimensions and obtain multi-dimensional spatial features. Then, the extracted multi-dimensional spatial features are input into a long short-term memory network layer containing 64 neurons to capture the dynamic law of channel quality changing over time and obtain time series features. After the Long Short-Term Memory network layer, an attention mechanism layer containing 32 neurons is set up to perform weighted fusion of temporal features, enhance the contribution of key features to the prediction results, and the features output by the attention mechanism layer are further input into the fully connected layer for feature dimensionality reduction and nonlinear fusion. In the output stage, on the one hand, the fused features are mapped to the output layer to generate the core channel quality parameters of each candidate channel, including signal-to-noise ratio, throughput and interference intensity; On the other hand, a lightweight linear transformation is performed on the temporal features extracted from the Long Short Time Memory network layer to obtain the auxiliary channel quality parameters of each candidate channel, including delay and jitter. The total number of model parameters is controlled within 800,000 to adapt to the computing power of the vehicle terminal, the inference latency is no more than 80ms, and the prediction accuracy can reach more than 92%. After processing the above neural network model, we obtain multiple channel quality parameter prediction results in groups of 1 second for the next 3 to 5 seconds. Each group of multiple channel quality parameter prediction results includes the predicted values ​​of five channel quality parameters for the current channel and each candidate channel, namely throughput, latency, jitter, signal-to-noise ratio and interference intensity. At the same time, a confidence level between 0 and 1 is attached to each predicted value for each channel to characterize the reliability of the prediction results. The five channel quality parameters of the current channel are obtained directly through the independent bypass real-time data channel of the input layer, without going through model inference; The core purpose of the model output is to capture the dynamic changing trend of the channel state and provide a data basis for multi-channel horizontal comparison for subsequent handover decisions.

[0044] In some embodiments, after obtaining the channel quality parameters of the current channel and each candidate channel, the process includes: Based on spatial environment data from multi-source spatiotemporal data, the road environment where the vehicle is located is identified, and the road environment identification result is obtained. Based on the road environment recognition results, the scene type of the vehicle is determined.

[0045] By identifying the road environment in which a vehicle is located based on spatial environmental data, and then determining the vehicle's scene type based on the road environment identification results, the rich geographic information contained in the spatial environmental data can be used to transform the physical environment in which the vehicle is located into a calculable and distinguishable scene type label. Since different road environments have different impact patterns on wireless signal propagation, the determination of scene type enables subsequent channel quality prediction and handover decisions to take environmental factors into account.

[0046] Furthermore, based on the characteristics in spatial environment data and vehicle operation status data, the driving scenarios in which the vehicle is located are divided into tunnel scenarios, building cluster scenarios, and congested convoy scenarios. Among them, the tunnel scenario is characterized by a confined space environment, loss of global navigation satellite system signals, and severe wireless signal obstruction; the building cluster scenario is characterized by a dense distribution of high-rise buildings and significant multipath interference effect of wireless signals; the congested convoy scenario is characterized by high density of surrounding vehicles, a large number of wireless devices on the same frequency, and intense channel competition. After identifying the initial scene type, the identification results are verified through multi-source data cross-validation to ensure the accuracy of scene determination. For example, for the determination of a tunnel scene, the following conditions must be met simultaneously: high-precision map data marks the current location as a tunnel area, the GPS signal is lost, and the wireless communication signal strength drops sharply by more than 10 dBm in a short period of time. Only when all of the above conditions are met is it finally confirmed as a tunnel scene. Through this multi-source data cross-validation mechanism, the accuracy of scene type identification can be improved to over 95%. For example, when a vehicle travels to the city center, high-precision map data shows that the surrounding building density level is high, the road type is urban road and there are no tunnel markers. The characteristics of the spatial environment data are consistent with the preliminary judgment of a building cluster scene. At this point, by further combining the vehicle operation status data, the GPS signal is stable and the wireless communication signal-to-noise ratio is 28 dBm, but a large multipath delay spread is detected. After comprehensive verification of multi-source data, it is confirmed that the current scene type is a building cluster scene. For example, when a vehicle is about to enter a tunnel, the high-precision map marks the tunnel entrance 200m ahead. As the vehicle enters, the GPS signal gradually weakens until it is completely lost. At the same time, the wireless communication signal-to-noise ratio drops from 32 dBm to 18 dBm within 3 seconds, a decrease of 14 dBm. When all three conditions are met, the tunnel scenario is finally confirmed.

[0047] In some embodiments, after determining the scene type of the vehicle based on the road environment recognition results, the process includes: Based on the scenario type, a preliminary screening threshold is set, and the core channel quality parameters of each candidate channel are compared with the preliminary screening threshold to obtain channels that fail the preliminary screening. The channels that fail the initial screening are removed from the candidate channels, and channels are selected from the preset list of available channels based on the scenario type to obtain the final candidate channels.

[0048] After obtaining the channel quality parameters, an initial screening threshold is set based on the scenario type. By comparing the core channel quality parameters of the candidate channels with the initial screening threshold, unqualified channels are identified and eliminated. Then, channels are supplemented from the preset list of available channels based on the scenario type. The scenario type can guide the differentiated setting of the threshold, making the initial screening of channels more in line with the characteristics of the current environment, eliminating obviously unsuitable channels, and ensuring a sufficient number of candidate channels through scenario-related supplementation.

[0049] Furthermore, after determining the channel quality parameters of each candidate channel, a differentiated initial screening threshold is first set based on the current scenario type of the vehicle. According to the characteristics of different scenario types, corresponding screening criteria are configured for key quality parameters. For example, in tunnel or building cluster scenarios, due to severe signal obstruction or significant multipath interference, the interference intensity threshold is set to -65 dBm; in congested convoy scenarios, due to dense co-channel devices and intense channel competition, the interference intensity threshold is set to a more stringent -70 dBm. At the same time, a unified switching cost threshold is set for all scenario types. Switching cost refers to the estimated time delay required to switch from the initiating channel to the new channel to resume data transmission, which includes authentication delay, data caching delay and recovery delay. Among them, authentication latency refers to the time required to establish a connection with the target channel and complete identity authentication; data caching latency refers to the time required to temporarily store data that has not yet been transmitted during the handover process; and recovery latency refers to the time required to resend the cached data or resume transmission after the handover is completed. The above latency data are not real-time measurements, but predictions based on historical data records. That is, by statistically analyzing the actual latency performance when performing channel handover in the same or similar scenarios in the past, an estimate of the current candidate channel handover cost is formed. The handover cost threshold is set to 50ms, meaning that channels with a handover cost greater than 50ms will be considered as unqualified channels in the initial screening. After the initial screening threshold is set, the candidate channels are initially screened and removed. The interference intensity and handover cost of each candidate channel can be compared with the above threshold. For candidate channels with interference intensity higher than the threshold of the corresponding scenario or handover cost higher than 50ms, they are judged as unqualified channels in the initial screening and removed from the candidate list. For example, when a vehicle is in a building complex scenario, if the interference intensity of candidate channel A is -60 dBm, which is higher than the threshold of -65 dBm for this scenario, it is determined to be an unqualified channel in the initial screening and is eliminated; if the interference intensity of candidate channel B is -68 dBm and the handover cost is 35 ms, it meets the screening criteria and is retained. After the initial screening and elimination, the number of remaining candidate channels is judged and supplemented. If the number of remaining candidate channels after elimination is less than the preset number, supplementation is carried out from the preset list of available channels associated with the current scenario type. During supplementation, the available channel with the highest historical average signal-to-noise ratio is selected first to ensure the quality of the supplemented channel. The supplementation process continues until the number of candidate channels reaches the preset requirement. For example, after initial screening and elimination, only candidate channel B remains, which is less than the preset number of two channels. At this point, channel C with the highest historical average signal-to-noise ratio is selected from the list of available channels associated with the building cluster scenario for supplementary selection, ultimately resulting in two candidate channels, B and C, which are used for subsequent comprehensive channel scoring and handover decisions. Through the aforementioned initial screening and supplementary selection mechanism, the screening criteria can be adaptively adjusted under different scenarios, eliminating channels that are obviously unsuitable for the current environment to avoid interfering with subsequent decisions. At the same time, scenario-related supplementary selection operations ensure a sufficient number of candidate channels, providing a higher quality and more stable candidate set for subsequent comprehensive channel scoring, thereby improving the accuracy and reliability of the overall handover decision.

[0050] S3: If the channel quality parameters of the current channel trigger the degradation threshold, record the trigger time, and calculate the comprehensive channel score of the current channel and each candidate channel based on the channel quality parameters, and determine the main target channel based on the comprehensive channel score. In some embodiments, after obtaining the channel quality parameters of the current channel and each candidate channel, the method further includes: Set the degradation threshold corresponding to the channel quality parameters of the current channel based on the scenario type; Compare the current channel quality parameters with the degradation threshold to determine whether a channel handover needs to be initiated. If the channel quality parameters of the current channel trigger the degradation threshold, a channel handover needs to be initiated, and the comprehensive channel score of the current channel and each candidate channel is calculated based on the channel quality parameters. If the channel quality parameters of the current channel do not trigger the degradation threshold, then there is no need to initiate a channel handover, and the current channel is maintained.

[0051] By setting a degradation threshold corresponding to the current channel quality parameters based on the scenario type, the system compares the current channel quality parameters with the degradation threshold to determine whether a handover needs to be initiated. The subsequent scoring and handover process is only started when the threshold is triggered. This fully utilizes the adaptive adjustment capability of the scenario type to the threshold, making the timing of the handover trigger more in line with the actual communication needs in different environments. It avoids false triggering or missed triggering caused by fixed thresholds, and filters out situations where no handover is needed through conditional judgment.

[0052] Furthermore, when setting the degradation threshold corresponding to the channel quality parameters of the current channel, the two factors of the vehicle's scenario type and the current service type are taken into account to achieve dual adaptive adjustment of the degradation threshold; On the one hand, based on the type of scenario in which the vehicle is located, the different impacts of different environments on signal propagation are considered, and the degradation threshold is adjusted to adapt to the environment. For example, in tunnel scenarios, due to severe signal obstruction and drastic environmental changes, the signal-to-noise ratio threshold can be appropriately relaxed to avoid frequent switching caused by short-term signal fluctuations. In building cluster scenarios, due to significant multipath interference, the bit error rate threshold can be adjusted accordingly. On the other hand, based on the specific business type of the current vehicle operation and considering the different communication quality requirements of different businesses, the degradation threshold is adjusted to adapt to the business. For example, for navigation map update scenarios with high real-time requirements, a stricter delay threshold is adopted; for audio playback scenarios with high tolerance for communication quality, the various thresholds can be appropriately relaxed. Through the above dual adaptive mechanism, the setting of the degradation threshold is more in line with the current environmental characteristics and business needs, improving the accuracy and rationality of the switching trigger judgment. The preset degradation thresholds can include a signal-to-noise ratio (SNR) threshold and a bit error rate (BER) threshold. For example, the SNR threshold can be set to 20 dBm and the BER threshold can be set to 10 dBm. -5 ; After the degradation threshold is set, the trigger judgment of the degradation threshold is performed. The channel quality parameters of the current channel are compared with the degradation threshold to determine whether there is a channel quality parameter at a certain moment that triggers the degradation threshold within the next 3 to 5 seconds. If, within a future time window, the signal-to-noise ratio of the current channel is lower than or equal to the corresponding degradation threshold, or the bit error rate is higher than or equal to the corresponding degradation threshold, then it is determined that the current channel will experience quality degradation, and a channel handover needs to be initiated, and the trigger time is recorded. If the channel quality parameters of the current channel are consistently better than the degradation threshold within the future time window, it is determined that there is no need to initiate a channel switch, and communication can continue on the current channel.

[0053] In some embodiments, based on channel quality parameters, a comprehensive channel score is calculated for the current channel and each candidate channel, and the primary target channel is determined based on the comprehensive channel score, including: Based on the scenario type, determine the scenario-based benchmark range of channel quality parameters and configure scenario-based weights for the channel quality parameters; Based on the scenario-specific benchmark range, the channel quality parameters are standardized to obtain standardized values ​​of the channel quality parameters. Based on scenario-based weights, the standardized values ​​are weighted and calculated to obtain the comprehensive channel score; The candidate channel with the highest comprehensive channel score is selected as the primary target channel.

[0054] By determining the scenario-based benchmark range of channel quality parameters based on scenario type and configuring scenario-based weights, the comprehensive channel score is obtained by weighted calculation after parameter standardization, and the candidate channel with the highest score is selected as the main target channel. This approach can deeply integrate scenario characteristics into the parameter processing and weight allocation process, so that the differences in importance of different parameters under different environments can be reflected. The standardization process eliminates the influence of parameter dimensions and value ranges, and the final selected main target channel is the one with the best comprehensive performance in the current scenario.

[0055] Furthermore, when determining the scenario-based reference range of channel quality parameters, differentiated reference ranges are set for each channel quality parameter according to different scenario types; For signal-to-noise ratio, the typical range of vehicle wireless communication is uniformly adopted, with a scenario-based benchmark range of -50dBm to -80dBm; For throughput parameters, different scenario-based benchmark ranges are set according to scenario type. In tunnel or building cluster scenarios, the scenario-based benchmark range for throughput is set to 0Mbps to 50Mbps; in congested convoy scenarios, the scenario-based benchmark range for throughput is set to 0Mbps to 30Mbps. For interference intensity, the typical range of vehicle-mounted co-frequency interference is uniformly adopted, and its scenario-based benchmark range is -75dBm to -40dBm; For latency, different scenario-based baseline ranges are set according to scenario type. For example, the scenario-based baseline range for latency in tunnel scenario is 0ms to 100ms, the scenario-based baseline range for building cluster scenario is 0ms to 80ms, and the scenario-based baseline range for congested convoy scenario is 0ms to 60ms. For jitter, the acceptable range for in-vehicle services is uniformly adopted, with a scenario-based benchmark range of 0ms to 50ms.

[0056] After setting the complete scenario-based benchmark range, the various channel quality parameters are standardized. For positively mapped parameters such as signal-to-noise ratio and throughput, the larger the parameter value, the better the channel quality. The normalized value is calculated using the forward normalization formula, which is: Normalized value = (Actual parameter value – Lower limit of the scenario-based reference range) / (Upper limit of the scenario-based reference range – Lower limit of the scenario-based reference range). For inversely mapped parameters such as interference intensity, delay, and jitter, smaller parameter values ​​indicate better channel quality. In this case, the inverse normalization formula is used to calculate the normalized value. The inverse normalization formula is: Normalized value = (Upper limit of the scenario-based reference range - Actual parameter value) / (Upper limit of the scenario-based reference range - Lower limit of the scenario-based reference range). During the standardization process, if the actual parameter value exceeds the range of the scenario-based benchmark, it will be forcibly constrained to the boundary value to ensure that the standardization result strictly falls within the range of 0 to 1. Then, the scenario-based weights are configured and weighted calculations are performed. Based on the differences in the importance of each parameter to different scenario types, corresponding scenario-based weights are set, and the sum of all weight coefficients is 1. For example, in a tunnel scenario, due to severe signal obstruction, the scenario-based weighting of signal-to-noise ratio is set to 0.3, the scenario-based weighting of interference intensity is 0.25, the scenario-based weighting of delay is 0.15, the scenario-based weighting of throughput is 0.15, and the scenario-based weighting of jitter is 0.15. In congested convoy scenarios, due to intense channel competition, the scenario-based weight of throughput is increased to 0.25, the scenario-based weight of signal-to-noise ratio is adjusted to 0.25, the scenario-based weight of interference intensity is 0.2, the scenario-based weight of delay is 0.15, and the scenario-based weight of jitter is 0.15. The standardized values ​​of each parameter are weighted and summed according to the above-mentioned scenario-based weights to obtain the comprehensive channel score of each candidate channel. The comprehensive channel score ranges from 0 to 1, and the higher the score, the better the overall channel quality.

[0057] After obtaining the comprehensive channel score of each candidate channel, the candidate channel with the highest score is selected as the primary target channel, and the candidate channel with the second highest score is marked as an emergency backup channel, so that it can be quickly activated if the primary target channel fails to switch over. Before finally confirming the primary target channel, a secondary verification is required to calculate the difference between the comprehensive channel score of the primary target channel and the current channel. If the difference is less than a preset threshold, such as less than 15%, the handover benefit is deemed insufficient, the handover is abandoned, the current channel communication is maintained, and the channel quality is reassessed after 200ms. If the score difference between the primary target channel and the current channel reaches or exceeds 15%, the channel is confirmed as the final primary target channel, and the subsequent handover timing planning and execution phase begins. The calculation model for the comprehensive channel score is as follows:

[0058] in, For comprehensive channel scoring; This is the standardized value of the signal-to-noise ratio; This is the standardized value of the interference intensity; This is the standardized value for the delay; This is the standardized value for throughput; This is the standardized value for jitter; , , , , These are the scenario-specific weights for signal-to-noise ratio, interference intensity, latency, throughput, and jitter, respectively. For example, when a vehicle is in a building complex scenario, it is necessary to comprehensively evaluate two candidate channels. Candidate channel A has a signal-to-noise ratio of -68dBm, interference intensity of -72dBm, latency of 25ms, throughput of 35Mbps, and jitter of 8ms; candidate channel B has a signal-to-noise ratio of -70dBm, interference intensity of -68dBm, latency of 18ms, throughput of 28Mbps, and jitter of 12ms. The normalized values ​​for signal-to-noise ratio are 0.4 and 0.33; the normalized values ​​for interference intensity are 0.91 and 0.8; the normalized values ​​for latency in the building cluster scenario (0 to 80 ms) are 0.69 and 0.78; the normalized values ​​for throughput in the building cluster scenario (0 to 50 Mbps) are 0.7 and 0.56; and the normalized values ​​for jitter are 0.84 and 0.76. Using the weighted calculation based on the building cluster scenario, the comprehensive channel score of candidate channel A is 0.68, and the comprehensive channel score of candidate channel B is 0.61. Therefore, candidate channel A is selected as the primary target channel. The comprehensive score of the current channel is 0.55. The score difference between the primary target channel and the current channel is 23.6%, which is greater than the 15% threshold. The handover is confirmed to be feasible, and the subsequent handover process begins.

[0059] S4: Determine candidate handover times based on the trigger time, calculate the handover benefit for each candidate handover time based on the comprehensive channel score, and determine the optimal handover timing based on the handover benefit; In some embodiments, candidate handover times are determined based on the trigger time, the handover benefit for each candidate handover time is calculated based on the comprehensive channel score, and the optimal handover timing is determined based on the handover benefit, including: Based on the trigger time, a time window within a preset time range in the future is determined as a candidate switching time. Based on the comprehensive channel score of the primary target channel, the comprehensive channel score of the current channel, and the preset service impact coefficient, the handover benefit at each candidate handover time is calculated. Candidate switching times where the switching benefit is greater than the benefit threshold are selected to obtain the constraint satisfaction time. Based on the constraint satisfaction time, the optimal switching time is determined.

[0060] By determining candidate handover times within a future time window based on the trigger time, calculating the handover benefits of each candidate time based on the comprehensive channel score of the primary target channel and the current channel, as well as the preset service impact coefficient, and selecting times with benefits greater than a threshold to determine the optimal handover timing, the handover lead time and service impact can be incorporated into quantitative calculations. This ensures that the selected handover time can effectively avoid predicted degradation points and can be executed during relatively low service hours, thereby minimizing the impact on user experience while ensuring successful handover.

[0061] Furthermore, when determining the time window for candidate switching moments, the time range within the next 3 to 5 seconds is determined as the baseline based on the triggering time of degradation. Within the candidate window, the candidate switching moments are divided at intervals of 100 ms, resulting in a total of 30 to 50 discrete candidate moments. This ensures the precision of switching timing selection and avoids excessive computational complexity. Then, the handover benefit for each candidate handover time is calculated. The handover benefit value is calculated as follows: the comprehensive channel score of the primary target channel - the comprehensive channel score of the current channel - the service impact coefficient. The business impact coefficient is set according to the specific business type currently being operated by the vehicle. For example, when the vehicle is transmitting 4K video, which is highly sensitive to latency, the business impact coefficient is set to 0.1 to reflect the potential impact of the switching operation on the smoothness of video playback. When the vehicle is performing normal data synchronization, the business impact coefficient is set to 0.05. For audio playback scenarios with low continuity requirements, the business impact coefficient can be further reduced. After the switching benefit calculation is completed, only candidate switching times with switching benefits greater than or equal to the benefit threshold are retained to obtain constraint satisfaction times. Constraint satisfaction times indicate that initiating a switching at these times can obtain sufficient performance gains and is worthwhile to perform the switching operation. Based on the time when the constraints are met, the optimal handover time is further screened. The screening process takes into account both the stability of the channel state and the distribution of the service load. First, the moment with the highest comprehensive channel score of the main target channel is selected from the constrained moments as a candidate. At the same time, the fluctuation of the comprehensive channel score between two consecutive time steps before and after this moment is required not to exceed 5% to ensure that the channel state is relatively stable at this time and to avoid failure when the channel quality fluctuates drastically. Secondly, prioritize time periods with lower vehicle workload, such as times when there are no resource-intensive services running, such as video transmission tasks or remote upgrade package downloads for in-vehicle software. In addition, if there is a sudden interference prediction information issued by the roadside unit, indicating that strong interference may occur within a certain time period, all candidate times within 1 second before the interference occurs should be avoided to prevent failure due to sudden interference during the handover process. After the above multi-dimensional screening, the optimal handover timing was finally determined, accurate to the 10ms level. After determining the optimal handover timing, a 0.1s data buffer time was reserved simultaneously to temporarily store the data to be transmitted before the handover, so as to ensure that no data is lost during the handover process.

[0062] For example, the degradation trigger time is 2.5 seconds after the current time. Based on the trigger time, a time window of 3 to 5 seconds in the future is determined. That is, within the time period from the current time to 2.5 seconds, 25 candidate switching times are divided at 100ms intervals. The current service is 4K video transmission, the service impact coefficient is set to 0.1, the comprehensive channel score of the primary target channel is 0.85, and the comprehensive channel score of the current channel is 0.62. Therefore, the handover benefit is calculated as 0.85 - 0.62 - 0.1 = 0.13, which is lower than the benefit threshold of 0.15. Therefore, it is excluded in the first few moments. As time progresses, the overall channel score of the current channel gradually decreases. At 2.2 seconds, the overall channel score of the current channel drops to 0.58. At this time, the handover benefit is 0.85 - 0.58 - 0.1 = 0.17, reaching the benefit threshold. This moment is the constraint satisfaction moment. Continuing to evaluate subsequent time points, at 2.3s the overall channel score of the current channel is 0.57, and the handover gain is 0.18; at 2.4s the overall channel score of the current channel is 0.55, and the handover gain is 0.20. From the moments when these constraints are met, the score fluctuation within 200ms before and after each moment is examined. Among them, the fluctuation of the comprehensive channel score before and after 2.3s is 4%, and the fluctuation of the comprehensive channel score before and after 2.4s is 6%. Therefore, 2.4s is excluded due to excessive fluctuation. 2.3s meets the fluctuation requirements, and at this time there are no other resource-intensive services running, nor are there any sudden interference warnings. Finally, 2.3s is determined to be the optimal handover time, accurate to 2.30s, and a 0.1s data buffering time is reserved, that is, data is buffered from 2.2s until the handover is officially initiated at 2.3s.

[0063] S5: Combine the primary target channel and the optimal handover timing to generate a handover control command, and send the handover control command to the vehicle wireless communication module to control the vehicle wireless communication module to perform channel handover.

[0064] Furthermore, after generating the switching control command, it interacts with the vehicle-mounted wireless communication module through a designated communication interface to perform the channel switching operation; A communication connection is established with the vehicle's onboard wireless communication chip via a serial peripheral interface. The switching control command is issued in time according to the optimal switching timing to control the onboard wireless communication module to interrupt the current channel connection and connect to the target channel or target access point. During the handover process, a seamless handover mechanism is adopted to ensure the continuity of data transmission. Data to be transmitted is pre-cached before the handover is initiated, and the cached data is quickly resent after the handover is completed, thereby avoiding data loss or communication interruption due to channel interruption. During the handover process, an emergency backup mechanism is set up to deal with possible handover failures. While determining the primary target channel, one alternative channel is reserved in advance as a backup option for emergency handover. When the switching command is issued, if the switching fails due to sudden interference of the primary target channel, authentication failure or other reasons, the system will immediately trigger the emergency switching procedure. There is no need to make predictions and decisions again, and the system will directly switch to the first-ranked alternative channel. The delay of emergency handover is controlled within 50ms to minimize the impact on communication continuity. If the emergency handover still fails, it will automatically degrade to the traditional passive handover mode and push alarm information to the vehicle terminal system to record the failure scenario and cause for subsequent model and decision logic optimization.

[0065] For example, the vehicle is currently connected to channel 36, and the primary target channel is determined to be channel 40, the alternative channel is channel 44, and the optimal switching time is 1.5 seconds after the current time. 0.1 seconds before the optimal switching time arrives, that is, at the 1.4th second, the system begins to buffer the video data to be transmitted. At 1.5 seconds, a switching control command is sent to the vehicle wireless communication chip through the serial peripheral interface. The control chip interrupts the connection with channel 36 and initiates the association process with channel 40. During the switching process, the cached data is temporarily stored in memory. After the connection with channel 40 is successfully established, the cached data is quickly resent to ensure smooth video playback. This switching takes 45ms and is successful. For example, after the switching control command is issued, a sudden interference is detected on the primary target channel 40, with the interference intensity reaching -58dBm, exceeding the preset threshold, causing the association with channel 40 to fail. The system immediately triggers the emergency switching procedure, without needing to re-predict, and directly switches to the backup channel 44. The switching takes 38ms, communication is successfully restored, and the reason for the failure of channel 40 is fed back for subsequent optimization.

[0066] Based on the above channel handover method, multi-source spatiotemporal data is acquired and input into a neural network model to predict channel quality parameters. When the degradation threshold is triggered, the triggering time is recorded and a comprehensive channel score is calculated to determine the primary target channel and the optimal handover timing. Finally, a control command is generated to execute the handover. Because of the adoption of a prediction-based trigger judgment and benefit calculation mechanism, channel handover is no longer a passive remedy after signal degradation, but rather an active planning of handover timing based on predictions of future channel quality. This effectively avoids communication interruptions during the handover process and significantly improves the continuity and stability of vehicle wireless communication.

[0067] In some embodiments, after issuing a switching control command to the vehicle-mounted wireless communication module to control the vehicle-mounted wireless communication module to perform channel switching, the following steps are included: The actual quality parameters of the main target channel are collected in real time, and the actual quality parameters are compared with the degradation threshold to obtain the comparison results; Compare the actual quality parameters with the corresponding channel quality parameters to generate prediction bias data; Based on the comparison results and prediction bias data, the pre-trained neural network model is optimized.

[0068] After channel handover is executed, the actual quality parameters of the primary target channel are collected in real time and compared with the degradation threshold to evaluate the handover effect. At the same time, the actual parameters are compared with the predicted parameters to generate prediction deviation data. Based on the comparison results and prediction deviation, the neural network model is optimized. The handover effect can be verified by actual data and the prediction error can be quantified. This allows the model to continuously learn the channel change patterns in the real environment, continuously adjust and optimize its own parameters, and thus gradually improve the prediction accuracy in long-term operation.

[0069] Furthermore, after the channel handover is completed, the actual quality parameters of the primary target channel are collected in real time and compared with a preset degradation threshold to determine whether the handover is successful. The criteria for successful handover are that the actual quality parameters are better than the degradation threshold and the stable duration reaches more than 2 seconds. If the actual quality parameters fail to meet the above conditions, the handover is determined to have failed, and the reason for failure is recorded for subsequent optimization.

[0070] Subsequently, the comparison results and the deviation data between the actual quality parameters and the channel quality parameters are fed back to the relevant modules for closed-loop optimization. The actual quality parameters are compared with the channel quality parameters, and the prediction deviation of each parameter is calculated to obtain the prediction deviation data. The prediction deviation data, along with the comparison data, is fed back to the neural network model training module for incremental learning and parameter optimization of the model, enabling the model to continuously adapt to the channel change patterns in the real environment. The comparison results and prediction deviation data are also fed back to the historical data storage module to update the signal heat map data of the current driving segment, enriching the accuracy and timeliness of historical statistical information.

[0071] If a brief period of instability occurs during the switching process, the driver will be given minimal visual feedback through the user notification module. For example, if a delay of more than 100ms or a slight stutter is detected during the handover process, a brief "In-vehicle wireless optimization in progress" message is pushed to the vehicle's display screen to inform the user that the system is optimizing the channel. Once the handover is complete and communication is stable, the message will automatically disappear to minimize interference with the user's attention and achieve the design goal of a seamless experience.

[0072] In addition, privacy protection and low power consumption control mechanisms are implemented throughout the entire data collection and processing process; sensitive data involving user privacy, such as vehicle location and driving trajectory, are strictly limited to local processing and storage on the vehicle's terminal and are never uploaded to the cloud. Historical data storage is protected by advanced encryption standards to ensure that only authorized system modules can access it, thus preventing the risk of data leakage. In terms of power consumption control, the data acquisition frequency is dynamically adjusted according to the vehicle's driving status. When the vehicle is stationary, the data acquisition frequency is reduced to 5Hz to reduce unnecessary energy consumption; when the vehicle is traveling at high speed, the normal acquisition frequency of 20Hz is maintained to ensure that the data sampling density meets the prediction requirements. Meanwhile, the neural network model inference adopts an on-demand wake-up mechanism. In non-high-speed mobile scenarios or when the channel quality is stable, the model inference frequency is appropriately reduced, or even the inference task is paused, thereby effectively reducing the power consumption burden of the vehicle's central processing unit.

[0073] like Fig. 3 As shown, this application also provides a vehicle, including: The vehicle controller is used to execute the channel switching method described above to generate switching control commands; The vehicle-mounted wireless communication module is connected to the smart cockpit and is used to receive switching control commands and perform channel switching based on the switching control commands.

[0074] The intelligent cockpit generates switching control commands by executing channel switching methods, and the vehicle wireless communication module with communication connection receives the commands and executes the channel switching. This design separates the decision-making core from the execution unit. The intelligent cockpit uses its powerful computing capabilities to complete complex spatiotemporal predictions and decision calculations, while the vehicle wireless communication module focuses on executing specific channel switching operations. The two work together to ensure both the intelligence and real-time nature of decision-making, as well as the reliability and efficiency of execution.

[0075] Furthermore, the vehicle controller can include, but is not limited to, vehicle-mounted computing units with data processing capabilities such as vehicle-mounted systems and cockpit domain controllers. By deploying the channel switching method on the vehicle-mounted system or cockpit domain controller, the existing vehicle computing resources can be fully utilized without the need for additional dedicated hardware. This reduces system costs while improving the intelligence level and response speed of channel switching decisions.

[0076] This application also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the channel switching method described above.

[0077] By embedding the channel switching method into electronic devices in software form, the devices can autonomously complete the processing of multi-source spatiotemporal data, the prediction and evaluation of channel quality, and the generation of switching decisions based on their internal computing resources. The entire intelligent switching process can be implemented locally without relying on external servers or cloud computing power, thereby significantly reducing the dependence on network communication and interaction latency. This ensures that electronic devices can maintain efficient channel management capabilities even in unstable network conditions or offline environments, effectively improving the communication stability and real-time decision-making of devices in complex and ever-changing environments.

[0078] Although embodiments of this application have been shown and described above, it is understood that the above embodiments are exemplary and should not be construed as limiting this application. Those skilled in the art can make changes, modifications, substitutions and variations to the above embodiments within the scope of this application.

Claims

1. A channel switching method, applied to an on-board controller, characterized in that, include: Acquire multi-source spatiotemporal data during vehicle operation, preprocess the multi-source spatiotemporal data, and obtain a spatiotemporal prediction input dataset; The spatiotemporal prediction input dataset is input into a pre-trained neural network model to obtain the channel quality parameters of the current channel and each candidate channel; If the channel quality parameters of the current channel trigger the degradation threshold, the trigger time is recorded, and based on the channel quality parameters, the comprehensive channel score of the current channel and each candidate channel is calculated. The main target channel is determined based on the comprehensive channel score. Candidate handover times are determined based on the trigger time, the handover benefit of each candidate handover time is calculated based on the comprehensive channel score, and the optimal handover timing is determined based on the handover benefit. Combining the primary target channel and the optimal handover timing, a handover control command is generated and sent to the vehicle-mounted wireless communication module to enable the vehicle-mounted wireless communication module to switch to the primary target channel.

2. The channel switching method according to claim 1, characterized in that, The spatiotemporal prediction input dataset is input into a pre-trained neural network model to obtain the channel quality parameters of the current channel and each candidate channel, including: Through the first branch of the neural network model, spatiotemporal feature extraction and fusion inference are performed on the time series data of the channel parameters of the candidate channels in the spatiotemporal prediction input data to obtain the channel quality parameters of each candidate channel. The channel quality parameters of the current channel are obtained by mapping the time series data of the channel parameters in the spatiotemporal prediction input data through the second branch of the neural network model.

3. The channel switching method according to claim 2, characterized in that, Through the first branch of the neural network model, spatiotemporal feature extraction and fusion inference are performed on the spatiotemporal prediction input data to obtain the channel quality parameters of each candidate channel, including: The spatiotemporal prediction input dataset is convolved through the first branch to extract multi-dimensional spatial features, and temporal features are extracted based on the multi-dimensional spatial features. A linear transformation is performed on the time-series characteristics to obtain the auxiliary channel quality parameters of each candidate channel; The time-series features are subjected to feature dimensionality reduction and nonlinear fusion to obtain a fused feature vector. Based on the fused feature vector, the core channel quality parameters of each candidate channel are obtained. By combining the core channel quality parameters and the auxiliary channel quality parameters, the channel quality parameters of each candidate channel are obtained.

4. The channel switching method according to claim 3, characterized in that, After obtaining the channel quality parameters of the current channel and each candidate channel, the following are included: Based on the spatial environment data in the multi-source spatiotemporal data, the road environment where the vehicle is located is identified, and the road environment identification result is obtained. Based on the road environment recognition results, the scene type in which the vehicle is located is determined; Based on the scenario type, a preliminary screening threshold is set, and the core channel quality parameters of each candidate channel are compared with the preliminary screening threshold to obtain channels that fail the preliminary screening. The unqualified channels in the initial screening are removed from each of the candidate channels, and channels are supplemented from the preset list of available channels based on the scenario type to obtain the final candidate channels.

5. The channel switching method according to claim 4, characterized in that, After obtaining the channel quality parameters of the current channel and each candidate channel, the following is also included: Based on the scenario type, set a degradation threshold corresponding to the channel quality parameters of the current channel; Compare the current channel quality parameters with the degradation threshold to determine whether a channel handover needs to be initiated. If the channel quality parameters of the current channel trigger the degradation threshold, a channel handover needs to be initiated, and the comprehensive channel score of the current channel and each candidate channel is calculated based on the channel quality parameters. If the channel quality parameters of the current channel do not trigger the degradation threshold, then there is no need to initiate a channel handover, and the current channel is maintained.

6. The channel switching method according to claim 5, characterized in that, Based on the channel quality parameters, calculate the comprehensive channel score for the current channel and each candidate channel, and determine the primary target channel based on the comprehensive channel score, including: Based on the scenario type, determine the scenario-based benchmark range of the channel quality parameters, and configure scenario-based weights for the channel quality parameters; Based on the aforementioned scenario-based benchmark range, the channel quality parameters are standardized to obtain standardized values ​​of the channel quality parameters. Based on the scenario-based weights, the standardized values ​​are weighted and calculated to obtain the comprehensive channel score; The candidate channel with the highest comprehensive channel score is selected as the primary target channel.

7. The channel switching method according to claim 1, characterized in that, Based on the trigger time, candidate handover times are determined; based on the comprehensive channel score, the handover benefit for each candidate handover time is calculated; and based on the handover benefit, the optimal handover timing is determined, including: Based on the triggering time, a time window within a preset time range in the future is determined as the candidate switching time; Based on the comprehensive channel score of the primary target channel, the comprehensive channel score of the current channel, and the preset service impact coefficient, the handover benefit for each candidate handover moment is calculated. Candidate switching times where the switching benefit is greater than the benefit threshold are selected to obtain the constraint-satisfied times. Based on the constraint-satisfied times, the optimal switching timing is determined.

8. The channel switching method according to any one of claims 1 to 7, characterized in that, After the switching control command is sent to the vehicle-mounted wireless communication module to control the vehicle-mounted wireless communication module to perform channel switching, the following steps are included: The actual quality parameters of the main target channel are collected in real time, and the actual quality parameters are compared with the degradation threshold to obtain the comparison result; The actual quality parameters are compared with the corresponding channel quality parameters to generate prediction bias data; Based on the comparison results and the prediction deviation data, the pre-trained neural network model is optimized.

9. A vehicle comprising: An onboard controller is configured to execute the channel switching method as described in any one of claims 1 to 8 to generate a switching control command; The vehicle-mounted wireless communication module is communicatively connected to the vehicle-mounted controller and is used to receive the switching control command and perform channel switching based on the switching control command.

10. An electronic device 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 channel switching method as described in any one of claims 1 to 8.

Images