Traffic event hierarchical transmission strategy determination method, device and equipment

Abstract

The application discloses a traffic event hierarchical transmission strategy determination method, device and equipment, which is applied to the intelligent traffic technical field and the artificial intelligence technical field, can predict the service quality of a communication link based on a communication link quality prediction model, obtain the service quality of the communication link, and determine at least one target communication link for issuing a traffic event based on the service quality when performing communication link switching; based on a mapping relationship among a scene type, a positioning scheme and a hierarchical strategy, when a scene switching is detected based on scene environment data, a target communication link, a target positioning scheme and a target hierarchical strategy corresponding to a target scene are determined; an event score is determined according to the target hierarchical strategy, the traffic event type, the correlation between a vehicle and the traffic event and the scene environment data; and transmission strategy cases corresponding to different events are determined according to the event score, so that the transmission efficiency and safety of the traffic event can be met simultaneously under a complex traffic environment.

Description

Technical Field

[0001] This application belongs to the fields of intelligent transportation technology and artificial intelligence technology, and in particular relates to a method, apparatus and equipment for determining traffic incident classification and transmission strategies. Background Technology

[0002] In complex traffic scenarios such as highways, tunnels, urban expressways, and in severe weather, traffic incidents need to be sent to surrounding vehicles as soon as possible to remind drivers or autonomous driving systems to take timely evasive or deceleration measures to avoid secondary accidents.

[0003] In the prior art, in order to avoid data transmission interruption or loss due to deterioration of communication link quality, when the current link quality is detected to deteriorate, the communication is switched from the current communication link to other available communication links to maintain the continuity of communication.

[0004] However, existing technologies struggle to simultaneously meet the requirements of transmission efficiency and safety in complex traffic environments. Summary of the Invention

[0005] This application provides a method, apparatus, and device for determining a graded transmission strategy for traffic events, which can simultaneously meet the requirements of transmission efficiency and security for traffic events in complex traffic environments.

[0006] In a first aspect, embodiments of this application provide a method for determining a traffic event classification transmission strategy, the method comprising: Acquire traffic incident data, scene environment data, and link quality data within the target area; Based on the communication link quality prediction model, the service quality of the link quality data is predicted to obtain the service quality of the communication link, and at least one target communication link is determined based on the service quality when the communication link is switched. Based on the mapping relationship between scene type, positioning scheme and hierarchical strategy, when scene switching is detected based on scene environment data, the target communication link, target positioning scheme and target hierarchical strategy corresponding to the target scene are determined. According to the target classification strategy, the event score is determined based on the traffic event type, the correlation between the vehicle and the traffic event, and the scene environment data; the traffic event type is determined based on the traffic event data; the correlation between the vehicle and the traffic event is determined based on the traffic event data and the target vehicle location, and the target vehicle location is determined based on the target positioning scheme. The transmission strategy for different events is determined based on the event score.

[0007] Secondly, embodiments of this application provide a traffic event classification transmission strategy determination device, the device comprising: The acquisition module is used to acquire traffic event data, scene environment data, and link quality data within the target area; The prediction module is used to predict the service quality of the link quality data based on the communication link quality prediction model, obtain the service quality of the communication link, and determine at least one target communication link for issuing traffic events based on the service quality when performing communication link switching. The first determining module is used to determine the target communication link, target positioning scheme, and target hierarchical strategy corresponding to the target scene based on the mapping relationship between scene type, positioning scheme, and hierarchical strategy, when scene switching is detected based on scene environment data. The second determination module is used to determine the event score according to the target classification strategy, based on the traffic event type, the correlation between the vehicle and the traffic event, and the scene environment data; the traffic event type is determined based on the traffic event data; the correlation between the vehicle and the traffic event is determined based on the traffic event data and the target vehicle location, and the target vehicle location is determined based on the target positioning scheme; The third determination module is used to determine the transmission strategy corresponding to different events based on the event score.

[0008] Thirdly, embodiments of this application provide an electronic device, the device comprising: A processor and a memory storing computer program instructions; a method for determining a traffic event classification transmission strategy that implements any one of the above when the processor executes the computer program instructions.

[0009] Fourthly, embodiments of this application provide a computer-readable storage medium storing computer program instructions, which, when executed by a processor, implement a traffic event classification transmission strategy determination method as described above.

[0010] Fifthly, embodiments of this application provide a computer program product in which instructions, when executed by the processor of an electronic device, enable the electronic device to perform any of the above-mentioned methods for determining a traffic event classification transmission strategy.

[0011] This application provides a method, apparatus, and device for determining a traffic event classification transmission strategy. It predicts link quality using a communication link quality prediction model, enabling proactive link scheduling to eliminate data loss during switching windows and improve transmission efficiency. Then, based on the mapping relationship between scene type, positioning scheme, and classification strategy, it achieves three-dimensional synchronous adaptation of link configuration, positioning mode, and classification weights during scene switching. Finally, it dynamically calculates event scores by integrating three dimensions: event type, vehicle correlation, and scene environment. Based on the event scores, it determines the corresponding transmission strategy and encryption scheme for different events. This enables priority, high-reliability, and secure transmission of high-risk events, and lightweight resource scheduling for low-risk events, improving transmission efficiency while ensuring the security of early warning transmission in complex traffic scenarios. Attached Figure Description

[0012] To more clearly illustrate the technical solutions of the embodiments of this application, the accompanying drawings used in the embodiments of this application will be briefly introduced below. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0013] Figure 1 A flowchart illustrating a method for determining a traffic incident classification transmission strategy according to an embodiment of this application is shown. Figure 2 This illustration shows a schematic diagram of a traffic incident classification transmission strategy determination device provided in an embodiment of this application; Figure 3 A schematic diagram of the hardware structure of the electronic device provided in an embodiment of this application is shown. Detailed Implementation

[0014] The features and exemplary embodiments of various aspects of this application will be described in detail below. To make the objectives, technical solutions, and advantages of this application clearer, the application will be further described in detail below with reference to the accompanying drawings and specific embodiments. It should be understood that the specific embodiments described herein are only intended to explain this application and not to limit it. For those skilled in the art, this application can be implemented without some of these specific details. The following description of the embodiments is merely to provide a better understanding of this application by illustrating examples.

[0015] It should be noted that, in this document, relational terms such as "first" and "second" are used merely to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising..." does not exclude the presence of additional identical elements in the process, method, article, or apparatus that includes the element.

[0016] It should be noted that the acquisition, storage, use, and processing of data in this application embodiment all comply with the relevant provisions of national laws and regulations.

[0017] It should be noted that in the embodiments of this application, certain software, components, models and other existing solutions in the industry may be mentioned. These should be regarded as exemplary and are only intended to illustrate the feasibility of implementing the technical solution of this application. However, they do not mean that the applicant has used or necessarily used the solution.

[0018] In existing technologies, different traffic events have significantly different requirements for real-time and reliability of data transmission. For example, major traffic accidents or hazardous material leaks require rapid transmission to target vehicles, while ordinary congestion warnings have relatively lower timeliness requirements. If communication links are switched only when the quality of the current link deteriorates, the switching latency and packet loss risks faced by different traffic events (such as emergency accidents and ordinary congestion) are no different. Emergency events cannot obtain priority transmission guarantees, while low-priority events may occupy the same link switching resources as emergency events. This results in high-priority events failing to reach target vehicles in time due to interruptions or delays during link switching, while low-priority events exacerbate channel congestion by occupying too many communication resources. Ultimately, the efficiency and security of traffic event transmission cannot be effectively guaranteed.

[0019] To address the problems of existing technologies, embodiments of this application provide a method and electronic device for determining a graded transmission strategy for traffic events. First, a communication link quality prediction model is introduced to proactively assess link quality, transforming link scheduling from reactive switching to proactive prediction, providing a stable and reliable communication foundation for subsequent event delivery. Then, based on the mapping relationship between scene type, positioning scheme, and graded strategy, three-dimensional synchronous adaptation of link configuration, positioning mode, and graded weights is achieved during scene switching, ensuring globally optimal adaptation under different traffic environments. Next, event scores are dynamically calculated by integrating three dimensions: event type, vehicle correlation, and scene environment, accurately quantifying the actual threat level of events to driving safety. Finally, transmission strategies and encryption schemes are determined based on the differentiated event scores, enabling redundant transmission and high-security encryption for emergency events, while lightweight transmission is used for ordinary events to save resources, thus simultaneously satisfying both transmission efficiency and security for traffic events.

[0020] The following section first introduces a method for determining a traffic event classification transmission strategy provided in the embodiments of this application.

[0021] Figure 1 A flowchart illustrating a method for determining a traffic event classification transmission strategy according to an embodiment of this application is shown. Figure 1 As shown, the method may include the following steps: S101: Acquire traffic event data, scene environment data, and link quality data within the target area; S102: Based on the communication link quality prediction model, perform service quality prediction on the link quality data to obtain the service quality of the communication link, and determine at least one target communication link for issuing traffic events based on the service quality when performing communication link switching; S103: Based on the mapping relationship between scene type, positioning scheme and hierarchical strategy, determine the target communication link, target positioning scheme and target hierarchical strategy when scene switching is detected based on scene environment data; S104: According to the target classification strategy, the event score is determined based on the traffic event type, the correlation between the vehicle and the traffic event, and the scene environment data; the traffic event type is determined based on the traffic event data; the correlation between the vehicle and the traffic event is determined based on the traffic event data and the target vehicle location; the target vehicle location is determined based on the target positioning scheme. S105: Determine the transmission strategy corresponding to different events based on the event score.

[0022] In this embodiment, a communication link quality prediction model is used to predict the service quality of link quality data, proactively predicting link quality and determining the target communication link. This transforms link scheduling from reactive post-switching to proactive pre-switching, eliminating the data loss window caused by switching lag and providing a fundamental guarantee for early warning security. Simultaneously, it reduces invalid retransmissions due to link interruptions, improving transmission efficiency. Through scene switching steps, based on the mapping relationship between scene type, positioning scheme, and hierarchical strategy, the target communication link, target positioning scheme, and target hierarchical strategy are determined in parallel. This allows for synchronous adaptation of differentiated positioning accuracy and risk assessment weights when scene changes, improving the overall environment across all scenarios. The accuracy of vehicle location calculation and traffic incident risk assessment is improved. Finally, according to the target classification strategy, the event score is determined based on the traffic incident type, the correlation between the vehicle and the traffic incident, and the multi-dimensional fusion of scene environment data. This can achieve refined quantitative risk rating and effectively avoid the defects of underestimating high-risk events and overestimating low-risk events. Therefore, the transmission strategy corresponding to different events can be determined according to the event score, which can achieve precise matching between risk level and transmission redundancy mechanism. High-reliability and high-security transmission strategies are adopted for high-priority emergency events, while lightweight transmission strategies are adopted for low-priority routine events to save channel and computing resources, thus ensuring transmission efficiency while comprehensively ensuring the transmission security of traffic incidents.

[0023] In step S101 above, by acquiring traffic event data, scene environment data, and link quality data within the target area, a comprehensive perception of the traffic conditions and communication environment of the target area can be achieved. The target area refers to a pre-defined road network coverage area centered on the location of the traffic event, and its coverage range can be adaptively adjusted according to road type, traffic density, and the impact of the traffic event.

[0024] In one example, traffic event data is used to characterize abnormal traffic conditions occurring within a target area. This traffic event data can include: event type, impact range, and spread rate. Event type can include, but is not limited to: traffic accidents, road construction, traffic congestion, debris spills, severe weather warnings, etc. Impact range can include: the number of lanes occupied by the event and the length of road affected by the event. Spread rate can include: the spread rate of congested traffic flow, the movement speed of fog patches, and other dynamic evolution parameters. Scene environment data is used to characterize the current road type and weather conditions in the target area. This scene environment data can include: road network type and weather conditions. Road network type can include: highways, tunnels, urban expressways, suburban roads, etc. Weather conditions can include: visibility, rainfall, wind speed, and weather level (sunny, cloudy, rainy, foggy, snowy). Link quality data is used to characterize the real-time transmission performance of each available communication link within the target area. This link quality data can include: latency, packet loss rate, bandwidth, and signal strength.

[0025] In one example, the method can also acquire dynamic status data of vehicles within the target area. This vehicle dynamic status data may include: vehicle speed, vehicle type (small car, large car, hazardous materials transport vehicle), location information (latitude and longitude, lane number), following distance, and driving mode (manual driving, automatic driving, braking status). This vehicle dynamic status data is used in subsequent steps to calculate the correlation score between vehicles and traffic events.

[0026] In one example, a multi-dimensional data collection network can be built using roadside multi-source sensing devices and vehicle-mounted terminals to collect the aforementioned traffic event data, vehicle dynamic status data, scene environment data, and link quality data.

[0027] For example, roadside multi-source sensing devices may include one or more of the following: millimeter-wave radar: detection range 0 to 300 meters, detection accuracy ±0.5 meters, used to detect vehicle position, speed, and obstacles within the target area; lidar: point cloud density 128 lines, ranging accuracy ±0.1 meters, used to construct high-precision three-dimensional environmental point clouds and identify the impact range of traffic events; high-definition camera: 4K resolution, frame rate 30fps, used to acquire road images and identify event types and lane occupancy; roadside weather station: monitoring parameters including visibility, rainfall, wind speed, and temperature, used to collect weather conditions from scene environmental data; roadside unit: supporting dual-mode communication of 5G and PC5 (Phase-Direct-to-Communication Interface), used to aggregate various roadside sensing data and communicate with vehicle-mounted terminals; geomagnetic detector: used to detect vehicle presence and assist in determining lane occupancy status. For example, the vehicle terminal may include one or more of the following: vehicle sensors: used to collect vehicle dynamic status data such as vehicle speed, acceleration, heading angle, and braking status; Global Navigation Satellite System (GNSS) positioning module: with positioning accuracy of ±1 meter, used to obtain the latitude and longitude location information of the target vehicle; vehicle radar: used to detect the following distance between the vehicle and the vehicle in front; vehicle unit: used to report vehicle identity and status information such as vehicle type, license plate number, and battery status.

[0028] In step S102 above, the existing link scheduling mechanism relies entirely on the current measured link values ​​for handover decisions. When channel quality fluctuates drastically (e.g., a vehicle enters a tunnel at high speed or a foggy area), the sudden change in measured values ​​can lead to handover lag. During the window period before handover completion, a large number of data packets may be lost, resulting in traffic event data loss and consequently, the inability to transmit traffic events to the target vehicle in a timely manner. Therefore, this application proposes to predict the quality of service (QoS) of the link based on a communication link quality prediction model. This model allows for the early detection of the quality change trend of the communication link over a future period, enabling the determination of at least one target communication link for transmitting traffic events based on QoS when performing communication link handover. Compared to traditional solutions that only trigger handover after link quality has deteriorated, this application can complete handover preparation before the actual link quality declines, eliminating the data loss window caused by handover lag and ensuring the complete transmission of traffic event data.

[0029] In one example, service quality (QoS) prediction can be performed at preset intervals, and communication link switching can be triggered when the predicted QoS of a communication link is determined to be lower than a preset QoS threshold. Specifically, at least one target communication link is selected from the QoS predictions of all available links. For example, the QoS score of each link is re-predicted every 100ms, and when the predicted score of the main link is continuously lower than 6 points, it is determined that the link is about to deteriorate, triggering the link switching process. The link prediction model can output the QoS prediction value for the next 3 seconds, providing sufficient lead time for link switching. Alternatively, when a scenario switch is detected in step S103, a communication link can be selected as the target communication link based on QoS. The link reselection triggered by scenario switch and the link switching triggered periodically complement each other, together forming a complete active link scheduling mechanism: periodic triggering ensures continuous monitoring and timely switching of link quality, while scenario switch triggering ensures synchronous adaptation of link configuration when the scenario changes.

[0030] In one example, the communication link quality prediction model can be a Long Short-Term Memory (LSTM) network. The LSTM network structure can be: input layer (3-dimensional features: latency, packet loss rate, bandwidth) → hidden layer (2 LSTM layers, 64 neurons per layer) → fully connected layer (32 neurons) → output layer (1-dimensional QoS score). The time-series sliding window length of the input layer can be 10, meaning that link quality data from the past 10 time points (corresponding to 1000ms of historical data, calculated at a 100ms sampling period) are used as an input sample to predict the QoS score for the next 3 seconds.

[0031] One training process for a communication link quality prediction model may include the following steps: preprocessing historical link data. Specifically, preprocessing includes: using the Min-Max Normalization Method to map latency, packet loss rate, and bandwidth data to the [0,1] interval to eliminate the influence of different units on model training; and using 3D modeling... The principle (Laida criterion) is to remove outlier data points exceeding the mean ± 3 standard deviations to prevent extreme values ​​from distorting the model's learning direction. Linear interpolation is used to complete missing data points, ensuring the continuity of the time series. Secondly, the preprocessed dataset is divided into training, validation, and test sets with proportions of 70%, 20%, and 10%, respectively. Then, an LSTM network model is constructed, and training parameters are configured. Specifically, the optimizer uses Adam (Adaptive Moment Estimation), the learning rate is set to 0.001, the loss function is mean squared error, and training is iterated for 500 epochs. An early stopping mechanism is introduced, terminating training early when the validation set loss no longer decreases for several consecutive epochs to prevent overfitting. During training, historical link data needs to be labeled with QoS scores as training labels. The QoS score labeling rules are as follows: for latency metrics, the weight is 0.4. For latency less than or equal to 50ms, 4 points are awarded; for latency greater than 50ms but less than or equal to 100ms, 3 points are awarded; for latency greater than 100ms but less than or equal to 200ms, 1 point is awarded; for latency greater than 200ms, no points are awarded. The packet loss rate metric has a weight of 0.3. For packet loss rate less than or equal to 0.5%, 3 points are awarded; for packet loss rate greater than 0.5% but less than or equal to 1%, 2 points are awarded; for packet loss rate greater than 1% but less than or equal to 3%, 1 point is awarded; for packet loss rate greater than 3%, no points are awarded. The bandwidth metric has a weight of 0.3. 3 points are awarded when the bandwidth is greater than or equal to 100Mbps; 2 points are awarded when the bandwidth is greater than or equal to 50Mbps and less than 100Mbps; 1 point is awarded when the bandwidth is greater than or equal to 20Mbps and less than 50Mbps; no points are awarded when the bandwidth is less than 20Mbps. The latency score, packet loss rate score, and bandwidth score are multiplied by their respective weights and then summed to obtain the QoS score of the link at the corresponding time. The total QoS score is 10 points.

[0032] In one example, step S102, which involves predicting the service quality of the link based on the communication link quality prediction model, is implemented in the following way: The mapping relationship between scenario type, link quality data, and communication link quality prediction model is pre-configured.

[0033] In the embodiments of this application, the types of available communication links differ in different scenarios, and the channel propagation characteristics, latency jitter patterns, and packet loss distribution characteristics of different links are fundamentally different. If only one communication link quality prediction model is used to predict the link service quality in different scenarios, a serious drop in prediction accuracy due to model mismatch will occur. For example, in high-speed scenarios, 5G link signal strength is usually excellent, and the service quality prediction model mainly learns the quality change pattern under conditions of small signal fluctuations and sufficient bandwidth. However, when the vehicle enters a tunnel, the 5G signal is completely lost, and the system switches to Power Line Communication (PLC) or Long Range Radio (LoRa) links. The latency characteristics of PLC links (usually 10-20ms, stable) and the packet loss characteristics of LoRa links (significantly affected by tunnel length) are fundamentally different from those of 5G links. If the LSTM model weights trained in high-speed scenarios are continued to be used for QoS prediction of PLC / LoRa links, the model will misjudge the stable low latency of the PLC link as "signal anomaly" and the normal packet loss fluctuations of the LoRa link as "link degradation," ultimately leading to increased QoS score deviation and incorrect allocation of the target communication link. Therefore, by pre-configuring the mapping relationship between scenario type, link quality data, and communication link quality prediction model, it can be ensured that each scenario type uses a prediction model that matches its link characteristics, thereby avoiding prediction distortion caused by model mismatch.

[0034] It should be noted that for each scenario type, long-term historical quality data of each link in that scenario can be collected separately to form a scenario-specific link historical time-series database. When training the communication link quality prediction model, it can be trained independently based on the link historical time-series databases for different scenarios to obtain the communication link quality prediction model weight file corresponding to each scenario.

[0035] Upon detecting a scene change, the link quality data is switched to the target link quality data corresponding to the target scene type, and the communication link quality prediction model is switched to the target communication link quality prediction model corresponding to the target scene type.

[0036] In this embodiment, upon detecting a scene switch, the LSTM prediction module switches the currently used historical link quality database to the link-specific historical time-series database corresponding to the target scene. After the switch, the model's input data source changes from the link quality data stream of the old scene to the link quality data stream of the target scene; the LSTM prediction module switches the currently loaded LSTM model weight file to the scene neuron LSTM weight file corresponding to the target scene, ensuring that the distribution characteristics of the model input data remain consistent with the learning objective of the model parameters, thereby guaranteeing the accuracy and reliability of link quality prediction after the scene switch.

[0037] The target link quality data is input into the target communication link quality prediction model to predict the service quality of the communication link.

[0038] In this embodiment, after switching between link quality data and the communication link quality prediction model, the newly loaded target scenario LSTM model weights are used to perform forward inference calculations on the target link quality data (i.e., time series of latency, packet loss rate, and bandwidth) from the target scenario link type, outputting a QoS prediction score. This QoS score serves as the quantified service quality value for each link in the current scenario, used for subsequent link allocation decisions.

[0039] In one example, to avoid QoS score jumps caused by model abrupt changes, a smooth transition mechanism can be set after calculating the service quality of the communication link. Specifically, during a preset transition period after the scene switch, such as 3 seconds, the final QoS score is output by weighted fusion of the predictions from the old and new models. At the beginning of the transition period, the old model prediction has a higher weight (e.g., an initial weight of 0.7), while the new model prediction has a lower weight (e.g., an initial weight of 0.3). As time progresses, the weight coefficient of the old model prediction gradually decreases, while the weight coefficient of the new model prediction increases synchronously. At the end of the transition period, the weight of the old model prediction decreases to 0, and the weight of the new model prediction increases to 1, using the output value of the new scene LSTM model. The changes in the weight coefficients can be linearly decayed, i.e., uniformly changed during the transition period. This smooth transition mechanism effectively avoids QoS score jumps caused by sudden changes in model weights during scene switching, thereby preventing score jumps from accidentally triggering link switching conditions and avoiding frequent and repeated link switching.

[0040] In this embodiment, due to the fundamental differences in channel propagation characteristics, latency jitter patterns, and packet loss distribution characteristics of different links under different scenarios, uniformly predicting the link service quality under multiple scenarios based solely on a single general prediction model would lead to a significant increase in QoS prediction bias, resulting in incorrect target link allocation and ultimately affecting the reliable transmission of traffic event data. Therefore, this application pre-configures the mapping relationship between scenario types, link quality data, and communication link quality prediction models to ensure that each scenario type is bound to its own specific link quality database and communication link quality prediction model weight parameters. Furthermore, when a scenario switch is detected, the link quality data collection object and communication link quality prediction model are synchronously switched to the target link quality data and target communication link quality prediction model corresponding to the target scenario type, ensuring that the distribution characteristics of the model input data remain consistent with the learning objective of the model parameters. Since the weights of the target scenario communication link quality prediction model are trained based on the scenario-specific historical database, the model has fully learned the unique variation patterns of link quality under that scenario. Therefore, by using a scenario-specific model to predict the quality of similar links collected at the moment, the resulting QoS score can accurately reflect the true service quality of the link in that scenario, thereby reducing the prediction error of cross-scenario links and improving the reliability of cross-scenario traffic event transmission.

[0041] In one example, to use the link quality prediction results to reverse verify the accuracy of scenario identification, after performing service quality prediction on the link quality data based on the communication link quality prediction model to obtain the service quality of the communication link, the following steps are also included: If the service quality of the communication link remains below a preset threshold for an extended period, the scenario type detection is performed again.

[0042] In this application embodiment, the intrinsic characteristics of link quality differ fundamentally under different scenario types. Therefore, the service quality of the predicted communication link under each scenario should fall within the normal fluctuation range corresponding to that scenario. If, in the current scenario, the service quality of the predicted communication link is consistently lower than a preset threshold, such as 6 points, and the duration of this low score reaches a preset time period threshold, such as five consecutive predictions with a 100ms interval between each prediction being lower than 6 points, corresponding to a continuous low score duration of 500ms, then it indicates that the current link anomaly is not caused by instantaneous jitter. In this case, it is highly likely that the current scenario type detection is incorrect, that is, the currently identified scenario type is inconsistent with the actual scenario in which the vehicle is located, causing the currently used LSTM prediction model to be mismatched with the link characteristics in the real scenario, thereby outputting a continuously distorted QoS score. Based on this, this application triggers the reacquisition of scenario environment data (including rereading GNSS signal strength, querying road network type labels in high-precision maps, reading visibility data from roadside weather stations, etc.) and performs scenario type detection to determine whether a scenario misjudgment has occurred.

[0043] In one example, to more accurately distinguish between persistently low scores caused by scene misjudgment and persistently low scores caused by objective channel fluctuations, a gradient descent detection strategy is introduced. Specifically, when the QoS score shows a continuous monotonically decreasing trend within a preset time threshold, it indicates that the link characteristics are undergoing a fundamental change, which is determined to be caused by scene switching; conversely, if the QoS score shows a pattern of fluctuating around a certain average within the preset time threshold, it is determined to be objective channel fluctuation (such as temporary signal interference or sudden load increase), rather than scene identification error. Through this gradient descent detection strategy, the two different types of degradation causes can be more accurately distinguished, avoiding the accidental triggering of scene re-examination procedures in objective channel fluctuation scenarios.

[0044] Upon detecting a scene change, determine the target communication link, target positioning scheme, and target classification strategy corresponding to the changed scene.

[0045] In this embodiment, after detecting an anomaly based on the predicted link service quality and re-detecting the scene type, if it is confirmed that the current real scene is inconsistent with the identified scene type, a scene switch is determined to have occurred. At this time, the target communication link, target positioning scheme, and target classification strategy corresponding to the target scene are redefined, and the above-mentioned three-dimensional parameters are synchronously switched. This enables proactive tracing back to the scene identification level for self-verification and self-correction after an anomaly is detected in the link prediction, fundamentally solving the configuration mismatch problem caused by scene misjudgment and improving the adaptive capability in complex traffic environments.

[0046] In this embodiment, when the service quality of the communication link is lower than the preset threshold for a certain period of time, the scene type detection is re-performed. The link quality prediction result can be used to reverse verify the accuracy of scene identification. Thus, when an error is detected in the scene type detection, the scene type detection can be re-performed, and the target communication link, target positioning scheme, and target classification strategy corresponding to the switched scene can be determined. This improves the scene identification accuracy and avoids abnormal traffic event warnings caused by the mismatch between the scene and the communication link.

[0047] In step S103, the hierarchical strategy refers to the combination of weighted coefficients in the three-dimensional dynamic hierarchical model ( , , ),in The weighting coefficients for traffic incident types. This is a weighted coefficient representing the correlation between vehicles and traffic incidents; These are weighting coefficients for scene environment data. Different scene types have fundamentally different road network characteristics, weather conditions, and driving environments, thus requiring different positioning schemes and classification strategies. For example, in a tunnel scenario where GNSS signals are completely lost, continuing to use a GNSS-dependent positioning scheme will cause vehicle position errors to rapidly diverge to several meters or even tens of meters, leading to distortion in the calculation of the distance correlation between vehicles and events. Similarly, in severe weather scenarios, scene adaptability plays a more significant role in determining event priority; if the weighting coefficients used for sunny scenarios are still applied, the priority of fog events may be underestimated. Therefore, by configuring appropriate positioning schemes and classification strategies for different scene types, it is possible to avoid misjudging the correlation between vehicles and traffic events due to insufficient positioning accuracy, and to avoid deviations in event priority calculation due to unreasonable classification weights, enabling optimal perception accuracy and classification accuracy in any scenario. Based on this, when scene switching is detected based on scene environment data, the target communication link, target positioning scheme, and target classification strategy corresponding to the target scene can be determined according to a pre-set mapping relationship. It should be noted that the target communication link is not a fixed link that is pre-configured statically based on the scenario type, but is dynamically determined by the LSTM prediction model based on the current link quality data to predict the QoS score of each available link in real time, and then based on the predicted score.

[0048] In one example, the scenario type can include: highway, tunnel, urban expressway, suburban road, and severe weather scenario. The criteria for determining each scenario type are as follows: a highway scenario is defined as follows: when the road network type is identified as highway, or the vehicle speed is greater than or equal to 80 km / h for 5 seconds; a tunnel scenario is defined as follows: when the GNSS signal strength is less than or equal to -105 dBm for 3 seconds, or the road network type is tunnel; an urban expressway scenario is defined as follows: when the road network type is urban expressway, or the vehicle speed is between 50 km / h and 80 km / h for 5 seconds; a suburban road scenario is defined as follows: when the road network type is suburban road, or the vehicle speed is less than or equal to 50 km / h for 5 seconds; and a severe weather scenario is defined as follows: when visibility is less than 200m, rainfall is greater than or equal to 20 mm / h, or wind speed is greater than or equal to 15 m / s.

[0049] In one example, the mapping relationship between scene type, positioning scheme, and hierarchical strategy can be as follows: In the highway scene, the positioning scheme adopts GNSS and Inertial Measurement Unit (IMU) fusion positioning (expected accuracy ±0.5m), and the hierarchical weighting coefficient is... =0.3、 =0.4、 =0.3; In tunnel scenarios, the positioning scheme adopts a fusion positioning method using IMU, LiDAR (Light Detection and Ranging), and Simultaneous Localization and Mapping (SLAM) (expected accuracy 3cm), with a hierarchical weighting coefficient of 0.3. =0.2、 =0.4、 =0.4; In the urban expressway scenario, the positioning scheme adopts GNSS and visual SLAM fusion positioning (expected accuracy ±1m), and the hierarchical weighting coefficient is 0.4; =0.3、 =0.5、 =0.2; In suburban road scenarios, the positioning scheme adopts GNSS and IMU fusion positioning (expected accuracy ±2m), and the hierarchical weighting coefficient is 0.2; =0.4、 =0.3、 =0.3; In severe weather scenarios, the positioning scheme adopts GNSS, IMU and radar fusion positioning (expected accuracy ±0.8m), with a graded weighting coefficient of 0.3. =0.3、 =0.3、 =0.4.

[0050] In one example, it should be noted that the target communication link is not a fixed link that is pre-configured statically based on the scenario type, but is dynamically determined by the LSTM prediction model based on the current link quality data to predict the QoS score of each available link in real time, and then based on the predicted score. In different scenarios, the predicted and optimized link configuration combinations can be as follows: In highway scenarios, the system predicts that the 5G link has the highest QoS score, followed by the PC5 link. Therefore, a configuration combination of 5G as the primary link and PC5 as the backup link is adopted. In tunnel scenarios, due to severe 5G signal attenuation, the system predicts that the QoS scores of both the PLC link and the LoRa link are better than those of the wireless cellular link. Therefore, a configuration combination of PLC and LoRa as dual primary links in parallel is adopted. In urban expressway scenarios, the system predicts that both the 5G link and the WiFi 6 (sixth generation wireless network technology) link have better quality. A configuration combination of 5G as the primary link and WiFi 6 as the backup link is adopted. In suburban road scenarios, the system predicts that the overall performance of the 4G link and the PC5 link is the best. A configuration combination of 4G as the primary link and PC5 as the backup link is adopted. In severe weather scenarios, in order to ensure transmission reliability, both the system predicts that the 5G link and the PC5 link must be activated. A configuration combination of 5G and PC5 as dual primary links in parallel is adopted.

[0051] In step S104, according to the target classification strategy, the event score is determined based on the type of traffic event, the correlation between the vehicle and the traffic event, and the scene environment data. This can simultaneously take into account the degree of danger of the event itself, the degree of impact on the vehicle, and the current road environment risk, thereby improving the accuracy of event risk classification, reducing invalid warning pushes, and avoiding the problem of low-risk events being overestimated and high-risk events being underestimated in special high-risk scenarios.

[0052] Scores can be pre-configured for different traffic event types, the correlation between vehicles and traffic events, and scene environment data. The scores for the above three dimensions are then assigned according to a target grading strategy (i.e., the weighting coefficients corresponding to the current scene). , , A weighted calculation is performed to obtain a comprehensive event score, allowing the event score to dynamically change with vehicle location and scene environment. The calculation formula is as follows: (1) in, Indicates the event rating; Indicates the type of traffic incident; Indicates the degree of correlation between the vehicle and the traffic incident; Represents scene environment data; This represents the weighting coefficient corresponding to the type of traffic incident; This represents the weighting coefficient corresponding to the correlation between vehicles and traffic incidents; This represents the weighting coefficient corresponding to the scene environment data.

[0053] In one example, the semantic score E corresponding to a traffic event type can be determined according to the following rules: For major traffic accidents (e.g., serious accidents involving multiple vehicles and injuries), the semantic score is 5 points; for hazardous material vehicle malfunctions (e.g., emergency situations with leakage or fire risks), the semantic score is 4 points; for road construction (e.g., road closures affecting traffic efficiency), the semantic score is 3 points; for traffic congestion (e.g., severe congestion at speeds less than or equal to 20 km / h), the semantic score is 2 points; and for minor road debris (e.g., spilled material that does not affect driving safety), the semantic score is 1 point. Furthermore, for severe weather warnings, the semantic score can be based on the road construction level and is recorded as 3 points.

[0054] The correlation score V between a vehicle and a traffic event can be determined based on traffic event data and the location of the target vehicle. The target vehicle location is obtained based on the target positioning scheme determined in step S103. Specifically, the distance between each target vehicle location and the boundary of each traffic event's impact area is calculated, and the V value is determined according to the following rules: when the distance between the vehicle and the event is less than or equal to 1km, V=5 points, indicating high risk and requiring emergency response; when the distance is greater than 1km and less than or equal to 3km, V=3 points, indicating medium risk and requiring deceleration warning; when the distance is greater than 3km and less than or equal to 5km, V=2 points, indicating low risk and requiring close observation; when the distance is greater than 5km, V=1 point, indicating no direct impact and only notification. It should be noted that since the calculation of the V value depends on the target vehicle location, which is determined by the scene-adaptive positioning scheme in step S103, the correlation score between the vehicle and the traffic event can be accurately determined.

[0055] The score C corresponding to the scene environment data can be determined according to the following rules: When the scene type is a tunnel with severe weather, C=5 points, indicating the most complex environment and extremely low system fault tolerance; when the scene type is a highway with fog, C=4 points, indicating low visibility and significantly shortened driver reaction time; when the scene type is an urban expressway with rain, C=3 points, indicating slippery road surface and significantly increased braking distance; when the scene type is a suburban road with normal weather, C=2 points, indicating a relatively simple environment and high system fault tolerance; when the scene type is a parking scene, C=1 point, indicating low-speed driving and extremely low risk.

[0056] It should be noted that the scene adaptation score C is determined by the current scene type (including road network type and weather conditions) identified by the scene judgment module in step S103. Any combination of road network type and weather conditions corresponds to a preset C value. In one example, the road network type and weather conditions in the scene environment data may be at different risk levels. For example, the road network type is an urban expressway (score 3 points alone), and the weather condition is normal weather (score 1 point alone). The scene adaptation score C after the combination of the two should not be simply the average (2 points), but should be determined according to the safety principle of taking the maximum value. Specifically, by pre-configuring a mapping matrix of road network type risk level and weather condition risk level, when the two risks are inconsistent, the score corresponding to the higher risk level is taken as the C value. In the above scenario of urban expressway with normal weather, the road network type risk level is high (3 points), and the weather condition risk is low (1 point). The system takes the higher score of 3 points as the C value. This value selection principle reflects the conservative strategy of prioritizing driving safety in the hierarchical decision-making process of this application. It is better to slightly overestimate the priority of events and increase the limited resource consumption than to underestimate the priority and miss high-risk warnings.

[0057] In one example, to avoid misjudging the correlation score between vehicles and traffic events at key distance boundaries due to positioning errors, given the existence of errors in the positioning information, a specific implementation of step S104 is as follows: Obtain the distance between the vehicle's location and the location where the traffic incident occurred. Construct a distance confidence interval based on the distance value and the positioning error, which is determined based on the switched positioning scheme.

[0058] In this embodiment, different positioning schemes have different positioning accuracies in real-world scenarios, especially in environments with limited GNSS signals such as tunnels and urban canyons, where positioning errors can reach several meters or even greater. When a vehicle is located near the boundary of the correlation score (e.g., exactly 1km from the event, at the dividing line between V=5 and V=3), the presence of positioning errors may cause a deviation between the measured distance and the actual distance. For example, the actual distance is 0.98km (which should be considered high risk, V=5), but the positioning module outputs a measured distance of 1.02km (which is judged as medium risk, V=3). In this case, misjudging the V value from 5 to 3 leads to an underestimation of the event score S, which may cause the emergency event to be downgraded from Level 1 to Level 2, and from dual-link parallel transmission to primary / backup hot backup, thus reducing the level of security. Therefore, by constructing a distance confidence interval based on the distance value and positioning error, the fluctuation range of the actual distance between the vehicle and the event can be fully covered, avoiding the one-sided judgment of risk based solely on a single measured distance, and preventing misjudgment of correlation caused by positioning deviations from the source.

[0059] Specifically, the corresponding positioning error can be determined based on the switched positioning scheme. Specifically, the expected positioning accuracy parameters for the current scene can be read from the switched positioning scheme in step S103. Then, a distance confidence interval is constructed based on the distance value and the positioning error. The lower limit of the confidence interval is the measured distance value minus the positioning error, and the upper limit is the measured distance value plus the positioning error. The physical meaning of this confidence interval is that the true distance between the vehicle and the event has a probability of falling within this interval, thus covering all possible values ​​of the true distance.

[0060] The first correlation degree is determined based on the lower limit of the distance confidence interval, and the second correlation degree is determined based on the upper limit of the distance confidence interval; wherein, the target correlation degree is the correlation degree corresponding to the high-risk level between the first correlation degree and the second correlation degree; In this embodiment, a first correlation degree is determined based on the lower limit of the distance confidence interval, and a second correlation degree is determined based on the upper limit of the distance confidence interval, thereby obtaining risk level information corresponding to the two ends of the confidence interval. Specifically, the lower and upper limits of the distance confidence interval are substituted into the vehicle-traffic event correlation scoring standard to obtain the first and second correlation degrees.

[0061] In this embodiment, if the first correlation degree and the second correlation degree are the same, it means that the confidence interval falls within the same scoring interval and there is no boundary ambiguity problem. If the first correlation degree and the second correlation degree are different, it means that the confidence interval crosses the scoring boundary and the true distance is uncertain. In this case, the value with the higher score is selected as the target correlation degree for subsequent event score S calculation, thereby avoiding the omission of high-risk warnings due to underestimating priority.

[0062] In one example, to further optimize the accuracy of the conservative compensation strategy, the positioning error can be adaptively adjusted based on the real-time status of the current positioning environment. Specifically, while acquiring location information using a target positioning scheme, the confidence level of the positioning solution (e.g., the horizontal accuracy factor HDOP value in GNSS positioning, the feature point matching rate in visual SLAM, etc.) is calculated in real time. When the positioning confidence level is high, the positioning error is appropriately reduced (e.g., reduced to 60% of the standard value) to avoid resource waste caused by excessive conservatism; when the positioning confidence level is low, the actual positioning error may be greater than the standard value, and the system appropriately amplifies the positioning error (e.g., amplifies to 150% of the standard value) to ensure the sufficiency of conservative compensation. In addition, a resource constraint mechanism is set: if the difference in V value levels before and after correction reaches 2 or more (e.g., V is corrected from 1 point to 5 points), dual-link parallel transmission is not directly enabled, but only link hot backup is enabled to save bandwidth resources and avoid the excessive use of dual-link resources due to the conservative strategy, thus achieving a reasonable balance between driving safety and communication resource efficiency.

[0063] Based on the switched grading strategy, the traffic incident type, target relevance, and scene environment data are calculated to obtain the incident score.

[0064] In this embodiment, the event score is obtained by calculating the score corresponding to the traffic event type, the target correlation degree, and the score corresponding to the scene environment data according to the switched classification strategy (i.e., the weighting coefficient corresponding to the target classification strategy determined in step S103). By using the corrected target correlation degree in the event score calculation, the impact of misjudgment of correlation degree value caused by positioning error on the final event score can be corrected. This allows for a safety-biased priority decision to be made even when the positioning information is not completely accurate, thereby effectively preventing high-risk events caused by positioning errors from being downgraded and ensuring that key early warning information can obtain transmission resources that match its true risk level.

[0065] In this embodiment, the distance between the vehicle location and the location where the traffic event occurred is obtained. A distance confidence interval is constructed based on the distance value and the positioning error. The positioning error originates from the positioning scheme switched in step S103. The positioning accuracy varies in different scenarios, and the width of the confidence interval adapts to the positioning capabilities of different scenarios. Then, a first correlation degree is determined based on the lower limit of the distance confidence interval, and a second correlation degree is determined based on the upper limit of the distance confidence interval. The correlation degree corresponding to the high-risk level in the first and second correlation degrees is selected as the target correlation degree. This ensures that high-risk traffic events are not incorrectly downgraded when the actual distance may fall into different scoring intervals. Thus, according to the switched classification strategy, the traffic event type, target correlation degree, and scenario environment data are calculated to obtain the event score. This avoids the problem of key events being incorrectly downgraded at critical distances due to positioning errors, and improves the accuracy of scenario event classification.

[0066] In one example, to eliminate the impact of positioning errors on vehicle correlation scores in scenarios where positioning errors frequently affect them, after determining the first correlation degree based on the lower limit of the distance confidence interval and the second correlation degree based on the upper limit of the distance confidence interval, the following steps are also included: Within a preset statistical period, count the number of times that the first correlation degree and the second correlation degree are different.

[0067] In this embodiment, if the first correlation degree and the second correlation degree are different, it indicates that the distance confidence interval between the vehicle and the event crosses the correlation degree scoring boundary, meaning that the positioning error has substantially affected the accurate determination of the correlation degree value. If the first correlation degree and the second correlation degree are different too many times within a preset statistical period, it indicates that the interference of the positioning error on the correlation degree value determination in the current scenario is not an occasional event, but a persistent positioning performance bottleneck. The positioning scheme accuracy in this scenario is insufficient to support the accuracy requirements based on distance grading. Each correction of the correlation degree value involves a trade-off between safety and resource efficiency, and frequent corrections, while ensuring safety, also bring unnecessary resource overhead.

[0068] If the number of occurrences reaches a preset threshold, the current scenario type is recorded, and if subsequent occurrences of the same scenario type are detected, the sampling frequency of traffic event data and vehicle location is increased.

[0069] In this embodiment, if the number of occurrences reaches a preset threshold, it indicates that the positioning error in the current scenario is severely interfering with the accuracy of the classification. Active intervention measures are needed to reduce the positioning error at its source, rather than simply compensating after the error occurs. Therefore, in the same scenario, it is necessary to increase the sampling frequency of traffic event data and vehicle positions (e.g., from 50Hz to 100Hz). By acquiring higher-frequency location observation data, time-series filtering can be used to more effectively suppress random noise in positioning, making the positioning results closer to the true value.

[0070] In one example, to prevent the list of location-sensitive scenarios from accumulating indefinitely and remaining valid indefinitely, even if the location conditions of a scenario subsequently improve (e.g., an enhanced location base station is added to the road segment, or satellite signal coverage is optimized), an automatic time-limited removal mechanism is introduced for location-sensitive scenarios. Specifically, each scenario is given a validity period (e.g., 7 days) when it enters the location-sensitive list. During the validity period, the improved sampling frequency is automatically applied each time the scenario enters the list. After the validity period ends, the scenario is automatically removed from the list, and the sampling frequency reverts to the default value. If, after the validity period ends, the scenario triggers an event where the first and second correlation degrees are different a number of times, reaching a preset threshold, the scenario is added back to the list, and the validity period is reset. This time-limited mechanism allows for the periodic reassessment of the location performance requirements of each scenario, avoiding permanent oversampling due to historical records. It also enables timely response to resource release needs after improvements in location conditions, ensuring that the system always operates at the minimum necessary sampling frequency, thus guaranteeing optimal resource utilization efficiency.

[0071] In this embodiment, by counting the number of times the first correlation degree and the second correlation degree are different within a preset statistical period, recording the current scenario type when the number of occurrences reaches a preset threshold, and increasing the sampling frequency of traffic event data and vehicle location when the same scenario type is detected later, it is possible to automatically identify weak positioning scenarios with large positioning errors and prone to misjudgment of risk levels. By increasing the sampling frequency to obtain continuous high-precision data for multiple frames, the frequency of misjudgment of vehicle correlation scores across levels is reduced from the source, thereby reducing the probability of underestimation of event scores and downgrading of high-risk warnings, and improving the hierarchical reliability and driving safety assurance capabilities in weak positioning scenarios.

[0072] In one example, to address the issue that a fixed sampling period cannot adapt to traffic fluctuations during emergency events on a road segment, the method following step S104 includes: Based on the event score, determine the proportion of high-priority traffic events among all traffic events.

[0073] In this embodiment, high-priority traffic events refer to events with a high risk level that require priority in transmission and monitoring. For example, traffic events with an event score S of 9 or higher include Level 1 emergency events (S=12 to 15) and Level 2 high-priority events (S=9 to 11). These two types of events pose a direct or significant threat to driving safety, requiring the system to perform link quality monitoring and data transmission assurance with higher real-time performance. Based on the event score statistics, the proportion of high-priority traffic events in all traffic events is statistically analyzed to generate event load weights. This quantifies the urgency of the current business load. The higher the proportion of high-priority events, the greater the event load weight, indicating a higher density of emergency events in the current target area and a more urgent need for real-time monitoring of link quality changes. This provides a quantitative basis for the dynamic adjustment of subsequent sampling cycles.

[0074] In one example, "all traffic events" refers to all traffic events detected and whose event scoring has been completed in the target area within the current statistical period (e.g., within the past second), including events of all priority levels (levels 1 to 4). The statistical period can be configured according to the actual application scenario, such as the data sampling period.

[0075] Based on the stated proportion, an event load weight is generated according to a preset mapping relationship; In this embodiment, since high-priority traffic events pose a direct or significant threat to driving safety, a higher proportion indicates a more frequent occurrence of emergency events, a more urgent need for real-time monitoring of link quality changes, and a greater consumption of communication and computing resources. Conversely, a lower proportion indicates that the current road segment is dominated by routine events, allowing for a lower monitoring frequency to conserve resources. Therefore, the event load weight is positively correlated with the proportion. For example, when the proportion of high-priority events is 0, the event load weight is 0; when the proportion is greater than 0 and less than or equal to 10%, the event load weight is 0.3; when the proportion is greater than 10% and less than or equal to 30%, the event load weight is 0.6; and when the proportion is greater than 30%, the event load weight is 1.0.

[0076] The sampling period for traffic event data, scene environment data, and link quality data within the target area is adjusted based on the event load weight.

[0077] In this embodiment, a higher event load weight indicates a higher density of high-priority events within the current target area, requiring a higher sensitivity of the system to changes in link quality. Therefore, by dynamically adjusting the sampling period based on the event load weight, the sampling period is shortened (e.g., from 100ms to 50ms) when the event load weight is high, increasing the data acquisition frequency. This allows the communication link quality prediction model to acquire link quality change data more promptly, thereby detecting link degradation trends earlier and triggering predictive switching, ensuring the safe and reliable delivery of high-priority events. Conversely, the sampling period is extended (e.g., from 100ms to 200ms) when the event load weight is low, reducing the data acquisition frequency and minimizing sensor energy consumption, communication bandwidth, and computing resource consumption while maintaining basic monitoring capabilities. This achieves a dynamic match between data acquisition frequency and the urgency of the service load, ensuring real-time link monitoring with the highest sensitivity in high-urgency event-intensive scenarios and maintaining basic monitoring capabilities with the lowest energy consumption in low-load scenarios.

[0078] In one example, to further enhance the forward-looking nature of sampling period adjustments, historical event load weight records for a past period (e.g., the past 10 minutes) can be maintained. By analyzing the changing trends of this historical sequence, the direction of load changes in the future can be predicted, thereby achieving proactive forward-looking adjustments to the sampling period. For example, a lightweight trend prediction model (e.g., a linear regression model or a sliding window averaging model) can be configured, taking the historical event load weight sequence as input and outputting predicted values ​​of load weights for the future period. If the predicted value shows a continuous upward trend, the sampling period is shortened in advance before the load fully increases, preparing for the upcoming high-load state with link quality monitoring. If the predicted value shows a continuous downward trend, the sampling period is gradually extended to avoid wasting resources by maintaining high-frequency sampling after the load has decreased. This proactive prediction and adjustment mechanism further improves adaptability and resource utilization efficiency.

[0079] In this embodiment, the proportion of high-priority traffic events in all traffic events is determined based on event scoring, generating event load weights. The larger the event load weight, the more concentrated the emergency events are in the current target area, and the more urgent the need for real-time monitoring of link quality changes. Then, the sampling period of traffic event data, scene environment data, and link quality data in the target area is adjusted according to the event load weight. This allows the sampling period to change dynamically with the business load, achieving on-demand matching between data collection frequency and the urgency of business load. In high-emergency event scenarios, a high data collection frequency ensures the real-time monitoring of links and the reliable delivery of high-priority events, while a low data collection frequency maintains basic monitoring capabilities in low-load scenarios. This improves resource utilization efficiency and the timeliness of event delivery under different event load conditions.

[0080] In one example, to enable the communication link quality prediction model to anticipate the potential impact of traffic load on link quality and thus predict link degradation trends earlier when traffic load increases, another training method for the communication link quality prediction model is as follows: Obtain the training dataset, which includes: historical service quality sequences, historical link quality data sequences, and event load weight sequences for each communication link.

[0081] In this embodiment, in real-world traffic scenarios, changes in link quality are typically coupled with service load. When a large number of Level 1 and Level 2 high-priority events occur intensively within a target area, a large amount of critical early warning information needs to be sent at a higher frequency, leading to a sharp increase in data transmission on the communication link. This surge in service load directly causes link congestion, manifested as increased measured latency, increased packet loss rate, and decreased effective bandwidth. In other words, changes in link quality are affected not only by channel physical conditions but also by the current service load; the higher the service load, the greater the risk of link congestion and the higher the probability of link quality degradation. Therefore, by introducing an event load weight sequence into the training dataset, the communication link quality prediction model can establish a causal relationship during the learning process: increased event load weight → increased link congestion risk → decreased QoS score. When the model is predicting during the deployment phase, if the input event load weight continues to increase, the model can predict in advance that the link quality may decline accordingly, thus triggering link switching prediction earlier than models that rely solely on link physical indicators.

[0082] The communication link quality prediction model is trained based on the training dataset to obtain the trained communication link quality prediction model.

[0083] In this embodiment, since the training data includes an event load weight sequence, the model can learn the correlation between the event load weight change pattern and the QoS score change during training. For example, the model will learn that when the event load weight in the input sequence shows a continuous upward trend (indicating that high-priority events are increasing intensively), even if the latency and packet loss rate have not changed significantly, the future QoS score tends to decrease because the upcoming large number of high-priority events will occupy link bandwidth and cause congestion.

[0084] In one example, the model structure of the communication link quality prediction model can be: Input layer (4-dimensional features: latency, packet loss rate, bandwidth, and quality of service) → Hidden layer (2 layers of LSTM, 64 neurons per layer) → Fully connected layer (32 neurons) → Output layer (1-dimensional QoS score). Here, quality of service can also be represented by event load weights.

[0085] In another example, to further enhance the impact of event load weights on the prediction of the communication link quality prediction model, the event load weight features are not directly used as input features of the model, but rather as adjustment signals for the model's gating mechanism. Specifically, the event load weights are input to a separate gating network, which outputs an adjustment factor between 0 and 1. This factor dynamically adjusts the activation thresholds of the forget gate and input gate of the communication link quality prediction model. When the load weight is high, the adjustment factor increases, the forget gate threshold decreases, and the communication link quality prediction model tends to retain historical link state information (because link changes are more drastic under high load scenarios, requiring more historical context). When the load weight is low, the adjustment factor decreases, and the model tends to update the state to respond to the latest link changes. This extension scheme makes the impact of event load weights on the model's predictive behavior more refined.

[0086] In this embodiment of the application, a training dataset is obtained, which includes: historical service quality sequences, historical link quality data sequences, and event load weight sequences for each communication link. A communication link quality prediction model is trained based on the training dataset to obtain the trained communication link quality prediction model. This model enables the model to learn the temporal change patterns of link quality while simultaneously establishing the correlation between service load and link quality. This allows the model to have a stronger ability to anticipate changes in traffic scenarios with dynamic changes in service load, enabling it to detect link degradation trends earlier and trigger predictive switching, thus providing more sufficient link preparation time for the issuance of high-priority events.

[0087] In step S105, the transmission strategy corresponding to different events is determined according to the event score, which can achieve a differentiated balance between the secure and reliable transmission of high-priority events and the resource-efficient transmission of low-priority events. The transmission strategy refers to the overall transmission scheme adopted when event data is sent, which may include link configuration methods, transmission reliability mechanisms, and encryption schemes. This application pre-configures the transmission strategies corresponding to different event score ranges to form a mapping relationship between priority and transmission resources. After calculating the event score S in step S104, the system queries and selects the corresponding transmission strategy from the mapping relationship based on the priority level of the S value, and then sends different events to the target vehicles through the target communication link determined in step S102 or step S103. Furthermore, to further improve the timeliness and effectiveness of drivers' perception of emergency events, this application also sets differentiated vehicle-side alert methods for different priority levels.

[0088] For example, the priority level and transmission strategy mapping relationship can be configured as follows: For Level 1 (emergency) events, the event score S is 12 to 15 points. Level 1 events correspond to the highest transmission priority, and a transmission strategy combining dual-link parallel transmission and retransmission mechanism is adopted. Event data is transmitted simultaneously through the main link and the backup link. Successful reception on either link is considered valid, ensuring that data is not lost under extreme network conditions. The encryption scheme adopts quantum encryption technology, using quantum key distribution to achieve one-time pad, with a key length of 2048 bits, which has the ability to resist quantum computer attacks and can provide the highest level of security for the highest priority events. The distribution method adopts a pop-up reminder, voice broadcast and light warning triggered simultaneously to ensure that the driver can perceive the emergency event simultaneously through multiple channels of vision and hearing. For Level 2 (high) events, the event score S is 9 to 11 points. Level 2 events correspond to the second highest transmission priority, employing a primary / backup hot backup transmission strategy. The backup link synchronizes data in real time, achieving zero switching latency in the event of a primary link failure. The encryption scheme uses the AES-256 encryption algorithm (256-bit key length). The notification method combines voice broadcasting and text alerts. For Level 3 (medium) events, the event score S is 6 to 8. Level 3 events correspond to medium transmission priority, employing a transmission strategy combining primary link single transmission and periodic verification. Only the primary link is used for data transmission, with data integrity verification performed every 10 seconds. The encryption scheme uses the AES-128 encryption algorithm (128-bit key length). The notification method is solely text alerts. For Level 4 (low) events, the event score S is 3 to 5. Level 4 events correspond to the lowest transmission priority. The system employs a primary link single transmission strategy, using only the primary link for transmission, without backup mechanisms or periodic verification. The encryption scheme uses the DES encryption algorithm (56-bit key length). The notification method is solely message push notifications.

[0089] In one example, to avoid low-priority events crowding out the bandwidth resources of high-priority events when the link is congested, one specific implementation of step S105 is as follows: Based on the communication link quality prediction model, the link quality data is predicted to obtain the predicted load status information and predicted bandwidth utilization of the target communication link.

[0090] In this embodiment, the event score S calculated in step S104 is based on three dimensions: event semantics E, vehicle relevance V, and scene adaptation C. These three dimensions reflect "how urgent the event itself is" or "how much of a threat it poses to driving safety." However, the transmission priority of an event depends not only on its own urgency but also on the current communication link resource status. When the link is under high load, if a large number of events are sent normally according to the original score, it will lead to excessive competition for bandwidth resources, and the transmission latency and packet loss risk of high-priority events will increase accordingly. Therefore, by predicting the link quality data through a communication link quality prediction model, the load status information and predicted bandwidth occupancy rate corresponding to the target communication link are obtained to quantify the scarcity of current communication resources and provide a data basis for the dynamic correction of subsequent event scores.

[0091] In one example, a communication link quality prediction model can, based on historical sequences of input latency, packet loss rate, and bandwidth, simultaneously output predicted latency, predicted packet loss rate, and predicted bandwidth sequences while outputting a service quality prediction. The output service quality is used to determine the target communication link in step S102, while the output predicted latency, predicted packet loss rate, and predicted bandwidth sequences serve as intermediate results for calculating predicted load status information and predicted bandwidth occupancy. The communication link quality prediction model can employ a multi-task learning architecture, with its output layer containing multiple neurons corresponding to four output tasks: service quality, predicted latency, predicted packet loss rate, and predicted bandwidth. These four outputs share the time-series features extracted from the hidden layer of the preceding communication link quality prediction model. During training, a joint loss function is used for multi-task training, thereby improving the generalization accuracy of each output task.

[0092] The event score is revised based on at least one of the predicted load status information and the predicted bandwidth utilization.

[0093] In this embodiment, event scores are corrected based on at least one of predicted load status information and predicted bandwidth utilization. When the predicted load status is high or the predicted bandwidth utilization continues to rise, the original event score is revised upwards, allowing events originally near the priority level boundary to enter a higher priority level, thus obtaining a higher level of transmission resource guarantee when link resources are scarce. Conversely, when the predicted load status is low and bandwidth is ample, the original score remains unchanged, avoiding unnecessary over-allocation of resources. This achieves dynamic adaptation between event distribution priority and the actual link carrying capacity, ensuring that under link congestion conditions, scarce bandwidth resources are preferentially allocated to events with the greatest impact on traffic safety.

[0094] In one example, a mapping relationship can be preset between predicted load status information, predicted bandwidth utilization, and event score correction. Based on the current predicted load status information and predicted bandwidth utilization, the corresponding correction coefficient can be queried from this mapping relationship. For example, this mapping relationship can be: when the predicted load status is low and the predicted bandwidth utilization is below 70%, the link resources are considered sufficient, the correction is 0 points, and the event score remains unchanged; when the predicted load status is medium or the predicted bandwidth utilization is between 70% and 85%, the link is considered to be congested, the correction is 0.5 points, and the event score is adjusted upwards by 0.5 points; when the predicted load status is high or the predicted bandwidth utilization is above 85%, the link is considered to be in a severely congested state, the correction is 1.0 point, and the event score is adjusted upwards by 1.0 point. Through this tiered correction mapping, the correction magnitude can be precisely controlled according to the degree of link congestion, avoiding excessive correction that could cause a large influx of low-priority events into high-priority events, thus exacerbating link congestion.

[0095] The corresponding transmission strategy is determined based on the revised event score.

[0096] In this embodiment of the application, the corresponding transmission strategy and encryption scheme are determined based on the modified event score. The modified event score can be used as the input of the mapping table in step S105 to query the priority level to which the score belongs and its corresponding transmission strategy and encryption scheme. Then, the event is sent to the target vehicle through the target communication link according to the queried transmission strategy and encryption scheme.

[0097] In this embodiment, link quality data is predicted based on a communication link quality prediction model to obtain predicted load status information and predicted bandwidth utilization for the target communication link. Event scores are corrected based on at least one of the predicted load status information and predicted bandwidth utilization. A corresponding transmission strategy is determined based on the corrected event scores, achieving dynamic adaptation between event delivery priority and the actual carrying capacity of the link. Under link congestion conditions, the protection level of medium-level events can be automatically upgraded, allowing limited bandwidth resources to prioritize events with a higher impact on driving safety, effectively reducing the probability of high-risk warning delays or loss due to bandwidth contention. Simultaneously, through a correction mechanism based on prediction rather than actual measurement, the priority allocation of events can be adjusted in advance before link congestion actually occurs, avoiding the lag problem of responding only after congestion has occurred in existing technologies, and improving transmission reliability and adaptive scheduling capabilities in scenarios with limited communication resources.

[0098] In one example, to suppress repeated transmission strategy switching caused by frequent fluctuations in the event score S near the classification boundary when the vehicle is traveling at high speed, a graded anti-shake mechanism is built into the three-dimensional dynamic classification module. Specifically, the classification module incorporates a hysteresis filtering mechanism, setting upper and lower threshold hysteresis intervals. A hysteresis interval is formed between the up-shift threshold and the down-shift threshold. When the event score changes from a low interval to a high interval, a higher up-shift threshold is required to trigger a grade increase; conversely, when the event score changes from a high interval to a low interval, a lower down-shift threshold is required to trigger a grade decrease. Through this hysteresis interval, the system effectively filters out small fluctuations in the event score near the classification boundary, avoiding repeated classification switching caused by minor score changes.

[0099] Building upon this, a further duration constraint is added. That is, after an event score reaches the switching threshold, a level switch is not triggered immediately. Instead, the score must remain stable within the target level's corresponding score range for a preset duration (e.g., 300 milliseconds) before the grading module outputs a level switch command. This duration constraint ensures that score changes are continuous rather than instantaneous, further filtering out instantaneous score jumps caused by measurement noise or positioning jitter, effectively preventing frequent oscillations in the transmission strategy.

[0100] Building upon the aforementioned anti-shake mechanism, a link pre-locking mechanism is also introduced. Specifically, when the distance between the target vehicle and the event location is still outside the warning range (e.g., greater than 1.2km), the communication link quality prediction model pre-locks potentially needed backup link resources based on the current link quality prediction results. This means establishing logical connections for the links and allocating buffer resources in advance, but not activating data transmission. When the vehicle enters the warning range, if the event score fluctuates at the grade boundary, since the backup links have been pre-locked, there is no need to repeatedly perform link destruction and reconstruction operations during each grade switch; simply activating the locked backup links allows for a seamless switch of the transmission strategy.

[0101] In one example, to address the issue of traffic event data loss and the inability to issue high-risk warnings during the communication link switching window, the method before step S105 includes: If a communication link switch is detected, obtain the event score for each traffic event that has been calculated so far.

[0102] In this embodiment, the communication link switching process has a time window. From the disconnection of the old link to the complete establishment of the new link and the start of data transmission, it typically takes tens to hundreds of milliseconds (depending on the switching type and link establishment delay). During this window, communication between the system and the target vehicle is interrupted, and any traffic event data attempted to be sent will be lost due to transmission failure. Among these events attempted to be sent during the switching period, there may be Level 1 emergency events. Their loss will directly result in the driver not receiving evacuation information in time, potentially leading to serious safety accidents. Therefore, this application, upon detecting a communication link switch, immediately obtains the event scores of each currently calculated traffic event, forming a snapshot of the event list to be sent. This ensures that even in the worst-case scenario, where the old link is completely and instantaneously interrupted, the event list to be sent is safely saved and will not be lost due to subsequent communication interruptions, providing a complete data foundation for subsequent hierarchical caching and orderly recovery.

[0103] It should be noted that detecting a communication link switch can refer to an active link switch triggered by the communication link quality prediction model in step S102 predicting that the current main link QoS score is continuously lower than a preset threshold; or a link configuration switch driven by the scenario switch in step S103; or a link switch triggered after scenario re-checking confirms a misjudgment of the scenario.

[0104] Based on the event score, a caching strategy is determined for different traffic events, and the traffic events to be issued are cached according to the caching strategy.

[0105] In this embodiment, different event scores correspond to different levels of urgency and timeliness requirements. Level 1 emergency events need to be delivered immediately after the link is restored, while the timeliness requirements for Level 4 low-priority events are much lower than those for emergency events. If a uniform caching strategy is adopted for all events (e.g., uniformly stored in a regular cache and uniformly set with the same lifetime), high-priority events and low-priority events will be stored together and treated equally, potentially leading to a serious problem where Level 1 emergency events are overwritten and deleted by Level 4 events due to insufficient cache space. Therefore, by determining the caching strategy for different traffic events based on event scores, hierarchical storage and isolated management can be achieved. High-priority events are stored in a high-reliability cache area to receive the highest level of protection, while low-priority events are stored in a regular cache area with a limited lifetime. This ensures that storage resources for each priority level do not interfere with each other, and high-priority events are not affected by the crowding out of low-priority events, ensuring that the most urgent events receive the highest level of protection throughout the entire link switching process.

[0106] In one example, a mapping relationship between different event scoring ranges and caching strategies can be pre-configured. After obtaining the list of events to be sent, the corresponding caching scheme is determined by querying this mapping relationship based on the S value of each event. For example, this mapping relationship can be configured based on priority levels as follows: For level 1 priority events (S value of 12 to 15), they are stored in a high-speed memory cache protected against power failure. This cache uses independently powered non-volatile storage media, ensuring data is not lost after an abnormal power outage. This cache does not have a data eviction mechanism and will not actively delete any stored level 1 events due to insufficient space; it only actively clears the cache record after the event is successfully sent and a confirmation signal is received from the vehicle. For level 2 and level 3 priority events (S value of 6 to 11), they are stored in a regular circular cache. This cache uses a first-in, first-out circular buffer structure with a fixed cache space. When the cache is full, the earliest stored event record is automatically overwritten. For level 2 events, an additional copy can be retained in the backup link cache for resending via the backup link after the primary link switchover is completed. For low-priority events (level 4, with an S value of 3 to 5), they are stored in a short-term buffer. This buffer sets a time-to-live (TTL) period for each stored event, for example, configurable to 30 seconds. When an event's cache time exceeds the TTL threshold, the event is automatically removed from the cache and will not be resent. This mechanism effectively prevents low-value events from occupying cache space indefinitely during prolonged link interruptions.

[0107] Once the link switch is completed, the traffic events cached in each buffer are sequentially assigned a transmission strategy and encryption scheme based on the event score, and then transmitted to the target vehicle through the target communication link according to the transmission strategy and encryption scheme.

[0108] In this embodiment, after the link switch is completed, the new link is established and available, but communication bandwidth resources may still be limited, making it impossible to push all cached events instantly. To prioritize the allocation of scarce transmission resources to high-priority events with the greatest impact on traffic safety under limited bandwidth conditions, traffic events cached in each buffer are sequentially sent to the target vehicle according to event scores, from highest to lowest priority. This ensures that Level 1 emergency events arrive at the target vehicle first after the link is restored, while Level 4 events are automatically discarded if they have exceeded their time limit during the switchover, thus achieving rational utilization of bandwidth resources while ensuring traffic safety.

[0109] In one example, the retransmission order is as follows: First, the list of Level 1 events in the high-speed memory buffer under power failure protection is read, sorted by S value from high to low, and then sent out sequentially. After each Level 1 event is sent out, an acknowledgment signal is received from the vehicle. Only after receiving acknowledgment can the buffer record be cleared and the next Level 1 event be processed. If no acknowledgment signal is received, a retransmission mechanism is triggered (up to 3 retransmissions) to ensure that Level 1 events are not lost due to a single failed transmission. After all Level 1 events are processed, Level 2 events in the ordinary circular buffer are read, sorted by S value from high to low, and then sent out sequentially. The transmission strategy is hot backup of the primary and backup links. After Level 2 events are processed, Level 3 events in the ordinary circular buffer are processed. The transmission strategy is single transmission of the primary link plus periodic verification. After Level 3 events are processed, Level 4 events that have not yet expired in the short-term buffer are processed. The transmission strategy is single transmission of the primary link. Expired Level 4 events are automatically discarded and not retransmitted.

[0110] In one example, to address the potential shortage of cache space in high-density traffic event scenarios, the occupancy rate of each cache partition is monitored in real time, and the caching strategy is dynamically adjusted based on the occupancy rate: When the occupancy rate of the ordinary circular cache exceeds the preset occupancy rate, an active eviction strategy is automatically initiated, prioritizing the eviction of Level 3 events with the lowest S value and longest queuing time, freeing up storage space for newly arriving Level 2 events, and prioritizing the protection of higher priority events when cache space is insufficient. When the occupancy rate of the short-term cache exceeds a preset threshold (e.g., 70%), the lifetime of Level 4 events is automatically shortened from 30 seconds to 15 seconds, accelerating the automatic expiration and reclamation of low-value events and preventing low-priority events from consuming too many cache resources. When the occupancy rate of the high-speed memory cache protected against power failure exceeds a preset alarm threshold (e.g., 90%), it indicates that a large number of Level 1 emergency events are occurring in the current road segment. At this time, an alarm signal is sent to the link management module, requesting the immediate activation of all available links and an increase in transmission power to clear the Level 1 cache area as quickly as possible, preventing new Level 1 events from being unable to be stored due to insufficient cache space. Through this dynamic adjustment mechanism, cached resources can be adaptively reallocated in scenarios with drastic changes in event density, ensuring that the highest priority events still have sufficient cache space when subjected to high-density event impact.

[0111] In this embodiment, to avoid issues such as traffic event data packet loss, high-risk warning delays, and high- and low-priority events competing for transmission bandwidth during the communication link switching window, this application, upon detecting a communication link switch, first obtains the event scores of each traffic event that has already been calculated. Based on the event scores, a differentiated caching strategy is matched, and the traffic events to be sent are cached in partitions. This allows for the isolated storage of events of different risk levels during the window of link disconnection and transmission interruption, prioritizing the prevention of loss of emergency events while avoiding low-risk events from occupying core cache resources. Furthermore, after the link switch is completed and the new target communication link is ready, the cached data is read in descending order of event scores. A corresponding transmission strategy and encryption scheme are matched for each cached event, and hierarchical delivery is completed based on the target communication link. This enables priority re-sending of high-risk warnings, preventing a large number of low-priority congestion messages from crowding out link bandwidth and causing delays in emergency event delivery. This reduces the risk of secondary traffic accidents caused by link switching and improves the efficiency of traffic event transmission and the security of warnings.

[0112] The following describes the complete implementation process of the present invention in detail, using examples of highway fog and tunnel scenarios. It should be noted that the following scenarios are merely preferred embodiments of the present invention and are not intended to limit the scope of application of the present invention.

[0113] Scenario 1: Fog on a highway. This scenario illustrates the traffic incident classification and notification process of this invention under severe weather conditions.

[0114] First, multi-source data acquisition was performed. A roadside weather station detected a patchy fog event with visibility below 200 meters, marking this event as a severe weather warning. Millimeter-wave radar acquired vehicle information within the affected area, obtaining a target vehicle speed of 80 km / h and location coordinates of (120.315689, 31.456783). Simultaneously, the system collected nearly 10 seconds of link quality data for the 5G-Uu and PC5 links, with the 5G-Uu link exhibiting a latency of 20ms and a packet loss rate of 0.1%, and the PC5 link exhibiting a latency of 15ms and a packet loss rate of 0.3%.

[0115] Secondly, predictive scheduling is performed. The aforementioned link quality data is input into the communication link quality prediction model. The model predicts that the QoS score of the 5G-Uu link will be 9.2 points and the QoS score of the PC5 link will be 7.8 points within the next 3 seconds. Based on the prediction results, the system allocates the 5G-Uu link as the primary link and the PC5 link as the backup link. Since the fog event poses a serious threat to driving safety, the system classifies it as a Level 1 emergency priority and adopts a dual-link parallel transmission strategy, simultaneously sending out warning information through both the primary and backup links.

[0116] Then, a cross-scenario multi-dimensional collaborative switching is performed. The system identifies the current scenario as an adverse weather scenario, activates the adverse weather transmission mode, and maintains the positioning scheme as GNSS and IMU fusion positioning (expected accuracy better than 0.5 meters). At the same time, the weighting coefficients of the hierarchical strategy are updated, the scene adaptation dimension weighting coefficient γ is adjusted to 0.4, and the warning distance doubling strategy is activated, setting the warning distance coefficient k to 2.0.

[0117] Finally, the system executes a tiered alert distribution. The system calculates the event score according to a three-dimensional dynamic tiered model: the event semantic score E is 10 points (severe weather warning), the vehicle correlation score V is 9 points (vehicle approximately 1.5km from the fog area), and the scene adaptation score C is 10 points (highway with fog, complex environment, low visibility). Substituting these scores into the formula S = 0.3 × 10 + 0.3 × 9 + 0.4 × 10 = 9.7 points, the alert is rated as a Level 1 emergency event. The system uses a quantum encryption scheme to encrypt the warning information and transmits it in parallel via 5G-Uu and PC5 dual links at a push frequency of 10Hz to the target vehicle: Fog ahead 200 meters, recommended speed reduced to 60km / h, maintain a distance of ≥50 meters.

[0118] Scenario 2: Tunnel Scenario. This scenario illustrates the scene switching and event delivery process of this invention under GNSS signal-limited conditions.

[0119] First, scene trigger detection is performed. If the vehicle-mounted GNSS positioning module detects that the signal strength is below -105dBm for 3 consecutive seconds, the system determines that the vehicle has entered the tunnel based on the scene switching judgment conditions, thus triggering the tunnel scene switch.

[0120] Secondly, the system performs cross-scenario multi-dimensional collaborative switching. It switches the link configuration from the 5G-Uu main link + PC5 backup link in the highway scenario to the PLC main link + LoRa backup link in the tunnel scenario; it switches the positioning scheme from GNSS+IMU fusion positioning to IMU+LiDAR+visual SLAM fusion positioning, achieving a positioning accuracy of 3cm within the tunnel; and it adjusts the weighting coefficient γ of the hierarchical strategy to 0.4 to enhance the weight of scenario adaptability in event scoring.

[0121] Finally, the event transmission and distribution are executed. The system detects a road construction event within the tunnel, which is rated as a high-priority level 2 event. The system uses the PLC main link and the LoRa backup link for transmission, and uses a lightweight national cryptographic encryption scheme to encrypt the construction warning information. The warning content is pushed to the target vehicle: Tunnel construction 300 meters ahead, left turn lane closed, it is recommended to reduce speed to 40km / h. Actual testing shows that the transmission latency in this scenario is 85ms, and the distribution success rate reaches 99.5%.

[0122] Based on the traffic event classification and transmission strategy determination method provided in the above embodiments, this application also provides a specific implementation of the traffic event classification and transmission strategy determination device. Please refer to the following embodiments.

[0123] Figure 2 This paper illustrates a schematic diagram of a traffic event classification transmission strategy determination device according to an embodiment of this application. Figure 2 As shown, the traffic incident classification transmission strategy determination device 200 includes: The acquisition module 201 is used to acquire traffic event data, scene environment data, and link quality data within the target area. Prediction module 202 is used to predict the service quality of link quality data based on the communication link quality prediction model, obtain the service quality of the communication link, and determine at least one target communication link for issuing traffic events based on the service quality when performing communication link switching. The first determining module 203 is used to determine the target communication link, target positioning scheme and target hierarchical strategy corresponding to the target scene based on the mapping relationship between scene type, positioning scheme and hierarchical strategy, when scene switching is detected based on scene environment data. The second determining module 204 is used to determine the event score according to the target classification strategy, based on the traffic event type, the correlation between the vehicle and the traffic event, and the scene environment data; the traffic event type is determined based on the traffic event data; the correlation between the vehicle and the traffic event is determined based on the traffic event data and the target vehicle location, and the target vehicle location is determined based on the target positioning scheme; The third determining module 205 is used to determine the transmission strategy corresponding to different events based on the event score.

[0124] In one example, the prediction module 202 is specifically used to: pre-configure the mapping relationship between scenario type, link quality data, and communication link quality prediction model; when a scenario switch is detected, switch the link quality data to the target link quality data corresponding to the target scenario type, and switch the communication link quality prediction model to the target communication link quality prediction model corresponding to the target scenario type; input the target link quality data into the target communication link quality prediction model, predict the target link quality data, and obtain the service quality of the communication link.

[0125] In one example, the traffic event classification transmission strategy determination device 200 further includes: The first detection module is used to re-detect the scenario type when the service quality of the communication link is continuously lower than a preset threshold for a period of time that reaches a preset duration threshold. The fourth determination module is used to determine the target communication link, target positioning scheme, and target classification strategy corresponding to the switched scene when a scene switch is detected.

[0126] In one example, the traffic event classification transmission strategy determination device 200 further includes: The second detection module is used to obtain the event scores of each traffic event that has been calculated so far when a communication link switch is detected. The caching module is used to determine the caching strategy for different traffic events based on the event score, and to cache the traffic events to be issued according to the caching strategy. The fifth determination module is used to determine the corresponding transmission strategy and encryption scheme for the traffic events cached in each buffer area according to the event score after the link switch is completed, and then send them to the target vehicle through the target communication link according to the transmission strategy and encryption scheme.

[0127] In one example, the traffic event classification transmission strategy determination device 200 further includes: The first statistics module is used to determine the proportion of high-priority traffic incidents among all traffic incidents based on the incident score. The first statistics module is also used to generate event load weights according to the percentage and a preset mapping relationship, wherein the event load weights are positively correlated with the percentage.

[0128] The first adjustment module is used to adjust the sampling period of traffic event data, scene environment data, and link quality data within the target area according to the event load weight.

[0129] In one example, the second determining module 204 is specifically used to: obtain the distance value between the vehicle location and the location where the traffic event occurred; construct a distance confidence interval based on the distance value and the positioning error, wherein the positioning error is determined based on the switched positioning scheme; determine a first correlation degree based on the lower limit of the distance confidence interval; and determine a second correlation degree based on the upper limit of the distance confidence interval; wherein the target correlation degree is the correlation degree corresponding to the high-risk level in the first correlation degree and the second correlation degree; and calculate the event score according to the switched grading strategy for the traffic event type, the target correlation degree, and the scene environment data.

[0130] In one example, the traffic event classification transmission strategy determination device 200 further includes: The second statistical module is used to count the number of times the first correlation degree and the second correlation degree are different within a preset statistical period; The second adjustment module is used to record the current scenario type when the number of occurrences reaches a preset threshold, and to increase the sampling frequency of traffic event data and vehicle location when subsequent occurrences of the same scenario type are detected.

[0131] In one example, the third determining module 205 is specifically used to: predict link quality data based on the communication link quality prediction model to obtain the predicted load status information and predicted bandwidth utilization rate of the target communication link; correct the event score based on at least one of the predicted load status information and predicted bandwidth utilization rate; and determine the corresponding transmission strategy based on the corrected event score.

[0132] Figure 3 A schematic diagram of the hardware structure of an electronic device provided in an embodiment of this application is shown. The electronic device 300 includes a processor 301 and a memory 302 storing computer program instructions.

[0133] Specifically, the processor 301 may include a central processing unit (CPU), an application specific integrated circuit (ASIC), or one or more integrated circuits that can be configured to implement the embodiments of this application.

[0134] Memory 302 may include mass storage for data or instructions. For example, and not limitingly, memory 302 may include a hard disk drive (HDD), a floppy disk drive, flash memory, optical disk, magneto-optical disk, magnetic tape, or a Universal Serial Bus (USB) drive, or a combination of two or more of these. In one instance, memory 302 may include removable or non-removable (or fixed) media, or memory 302 may be a non-volatile solid-state memory.

[0135] In one example, memory 302 may be read-only memory (ROM). In one instance, the ROM may be a mask-programmed ROM, a programmable ROM (PROM), an erasable PROM (EPROM), an electrically erasable PROM (EEPROM), an electrically rewritable ROM (EAROM), or flash memory, or a combination of two or more of these.

[0136] Memory 302 may include read-only memory (ROM), random access memory (RAM), disk storage media device, optical storage media device, flash memory device, electrical, optical, or other physical / tangible memory storage device. Therefore, generally, memory includes one or more tangible (non-transitory) computer-readable storage media (e.g., memory devices) encoded with software including computer-executable instructions, and when the software is executed (e.g., by one or more processors), it is operable to perform the operations described with reference to the method according to one aspect of this application.

[0137] The processor 301 reads and executes computer program instructions stored in the memory 302 to implement a traffic event classification transmission strategy determination method in the above embodiment.

[0138] In one example, the electronic device may also include a communication interface 303 and a bus 304. For example, Figure 3 As shown, the processor 301, memory 302, and communication interface 303 are connected through bus 304 and complete communication with each other.

[0139] The communication interface 303 is mainly used to realize communication between various modules, devices, units and / or equipment in the embodiments of this application.

[0140] Bus 304 includes hardware, software, or both, that couples components of an online data traffic metering device together. For example, and not as a limitation, the bus may include an Accelerated Graphics Port (AGP) or other graphics bus, an Extended Industry Standard Architecture (EISA) bus, a Front Side Bus (FSB), a Hyper Transport (HT) interconnect, an Industry Standard Architecture (ISA) bus, an Infinite Bandwidth Interconnect, a Low Pin Count (LPC) bus, a memory bus, a Microchannel Architecture (MCA) bus, a Peripheral Component Interconnect (PCI) bus, a PCI-Express (PCI-X) bus, a Serial Advanced Technology Attachment (SATA) bus, a Video Electronics Standards Association Local (VLB) bus, or other suitable buses, or combinations of two or more of these. Where appropriate, bus 304 may include one or more buses. Although specific buses are described and illustrated in embodiments of this application, this application contemplates any suitable bus or interconnect.

[0141] Furthermore, in conjunction with the traffic event classification and transmission strategy determination method in the above embodiments, this application embodiment can provide a computer-readable storage medium for implementation. The computer-readable storage medium stores computer program instructions; when executed by a processor, these computer program instructions implement any of the traffic event classification and transmission strategy determination methods in the above embodiments.

[0142] This application also provides a computer program product, including a computer program that, when executed by a processor, implements any of the traffic event classification transmission strategy determination methods described in the above embodiments.

[0143] It should be clarified that this application is not limited to the specific configurations and processes described above and shown in the figures. For the sake of brevity, detailed descriptions of known methods are omitted here. In the above embodiments, several specific steps are described and shown as examples. However, the method process of this application is not limited to the specific steps described and shown. Those skilled in the art can make various changes, modifications, and additions, or change the order of steps, after understanding the spirit of this application.

[0144] The functional blocks shown in the above block diagram can be implemented as hardware, software, firmware, or a combination thereof. When implemented in hardware, they can be, for example, electronic circuits, application-specific integrated circuits (ASICs), appropriate firmware, plug-ins, function cards, etc. When implemented in software, the elements of this application are programs or code segments used to perform the required tasks. Programs or code segments can be stored on a machine-readable medium or transmitted over a transmission medium or communication link via data signals carried on a carrier wave. "Machine-readable medium" can include any medium capable of storing or transmitting information. Examples of machine-readable media include electronic circuits, semiconductor memory devices, read-only memory (ROM), flash memory, erasable read-only memory (EROM), floppy disks, compact disc read-only memory (CD-ROM), optical disks, hard disks, fiber optic media, radio frequency (RF) links, etc. Code segments can be downloaded via computer networks such as the Internet, intranets, etc.

[0145] It should also be noted that the exemplary embodiments mentioned in this application describe methods or systems based on a series of steps or apparatus. However, this application is not limited to the order of the above steps; that is, the steps can be performed in the order mentioned in the embodiments, or in a different order, or several steps can be performed simultaneously.

[0146] The aspects of this disclosure have been described above with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this disclosure. It should be understood that each block in the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, or other programmable data processing apparatus to produce a machine such that these instructions, executable via the processor of the computer or other programmable data processing apparatus, enable the implementation of the functions / actions specified in one or more blocks of the flowchart illustrations and / or block diagrams. Such a processor can be, but is not limited to, a general-purpose processor, a special-purpose processor, a special application processor, or a field-programmable logic circuit. It is also understood that each block in the block diagrams and / or flowcharts, and combinations of blocks in the block diagrams and / or flowcharts, can also be implemented by special-purpose hardware performing the specified functions or actions, or can be implemented by a combination of special-purpose hardware and computer instructions.

[0147] The above are merely specific embodiments of this application. Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the systems, modules, and servers described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here. It should be understood that the protection scope of this application is not limited thereto. Any person skilled in the art can easily conceive of various equivalent modifications or substitutions within the technical scope disclosed in this application, and these modifications or substitutions should all be covered within the protection scope of this application.

Claims

1. A method for determining a graded transmission strategy for traffic incidents, characterized in that, include: Acquire traffic incident data, scene environment data, and link quality data within the target area; Based on the communication link quality prediction model, the service quality of the link quality data is predicted to obtain the service quality of the communication link, and at least one target communication link for issuing traffic events is determined based on the service quality when performing communication link switching. Based on the mapping relationship between scene type, positioning scheme and hierarchical strategy, when scene switching is detected based on scene environment data, the target communication link, target positioning scheme and target hierarchical strategy corresponding to the target scene are determined. According to the target classification strategy, the event score is determined based on the traffic event type, the correlation between the vehicle and the traffic event, and the scene environment data; the traffic event type is determined based on the traffic event data; the correlation between the vehicle and the traffic event is determined based on the traffic event data and the target vehicle location, and the target vehicle location is determined based on the target positioning scheme. The transmission strategy corresponding to different events is determined based on the event score.

2. The method according to claim 1, characterized in that, The process of predicting the service quality of the communication link based on the communication link quality prediction model to obtain the service quality of the communication link includes: Pre-configure the mapping relationship between scenario types, link quality data, and communication link quality prediction models; Upon detecting a scene change, the link quality data is switched to the target link quality data corresponding to the target scene type, and the communication link quality prediction model is switched to the target communication link quality prediction model corresponding to the target scene type. The target link quality data is input into the target communication link quality prediction model to predict the service quality of the communication link.

3. The method according to claim 1 or 2, characterized in that, After predicting the service quality of the communication link based on the link quality prediction model, the process further includes: If the service quality of the communication link remains below a preset threshold for an extended period, the scenario type detection is performed again. Upon detecting a scene change, the target communication link, the target positioning scheme, and the target classification strategy corresponding to the changed scene are determined.

4. The method according to claim 1, characterized in that, Before determining the transmission strategy and encryption scheme corresponding to different events based on the event scores, the process also includes: If a communication link switch is detected, obtain the event score for each traffic event that has been calculated so far; Based on the event score, a caching strategy for different traffic events is determined, and the traffic events to be issued are cached according to the caching strategy. Upon completion of the link switch, the traffic events cached in each buffer zone are sequentially determined according to the event score, and the corresponding transmission strategy and encryption scheme are determined based on the event score. The events are then transmitted to the target vehicle through the target communication link according to the transmission strategy and encryption scheme.

5. The method according to claim 1, characterized in that, After determining the event score according to the target classification strategy, based on the traffic event type, the correlation between the vehicle and the traffic event, and the scene environment data, the process further includes: Based on the event scores, the proportion of high-priority traffic events among all traffic events is determined; Based on the stated proportion, an event load weight is generated according to a preset mapping relationship, and the event load weight is positively correlated with the stated proportion. The sampling period of traffic event data, scene environment data, and link quality data within the target area is adjusted according to the event load weight.

6. The method according to any one of claims 1, 2, 4 or 5, characterized in that, The step of determining an event score according to the target classification strategy, based on the type of traffic event, the correlation between the vehicle and the traffic event, and the scene environment data, includes: Obtain the distance value between the vehicle location and the location where the traffic incident occurred, and construct a distance confidence interval based on the distance value and the positioning error, wherein the positioning error is determined based on the switched positioning scheme; A first correlation degree is determined based on the lower limit of the distance confidence interval, and a second correlation degree is determined based on the upper limit of the distance confidence interval; wherein, the target correlation degree is the correlation degree corresponding to the high-risk level between the first correlation degree and the second correlation degree; According to the switched grading strategy, the traffic event type, the target correlation degree, and the scene environment data are calculated to obtain the event score.

7. The method according to claim 6, characterized in that, After determining the first correlation degree based on the lower limit of the distance confidence interval and the second correlation degree based on the upper limit of the distance confidence interval, the method further includes: Within a preset statistical period, the number of times that the first correlation degree and the second correlation degree are different is counted; When the number of occurrences reaches a preset threshold, the current scenario type is recorded, and when subsequent occurrences of the same scenario type are detected, the sampling frequency of the traffic event data and the vehicle location is increased.

8. The method according to claim 1, characterized in that, The step of determining the transmission strategy and encryption scheme corresponding to different events based on the event score includes: Based on the communication link quality prediction model, the link quality data is predicted to obtain the predicted load status information and predicted bandwidth utilization rate of the target communication link. The event score is corrected based on at least one of the predicted load status information and the predicted bandwidth utilization. The corresponding transmission strategy is determined based on the revised event score.

9. A device for determining a traffic incident classification and transmission strategy, characterized in that, include: The acquisition module is used to acquire traffic event data, scene environment data, and link quality data within the target area; The prediction module is used to predict the service quality of the link quality data based on the communication link quality prediction model, obtain the service quality of the communication link, and determine at least one target communication link for issuing traffic events based on the service quality when performing communication link switching. The first determining module is used to determine the target communication link, target positioning scheme, and target hierarchical strategy corresponding to the target scene based on the mapping relationship between scene type, positioning scheme, and hierarchical strategy, when a scene change is detected based on scene environment data. The second determining module is used to determine an event score according to the target classification strategy, based on the traffic event type, the correlation between the vehicle and the traffic event, and the scene environment data; the traffic event type is determined based on the traffic event data; the correlation between the vehicle and the traffic event is determined based on the traffic event data and the target vehicle location, and the target vehicle location is determined based on the target positioning scheme. The third determining module is used to determine the transmission strategy corresponding to different events based on the event score.

10. An electronic device, characterized in that, The device includes: a processor and a memory storing computer program instructions; when the processor executes the computer program instructions, it implements a method for determining a traffic event classification transmission strategy as described in any one of claims 1-8.

Images