Decentralized risk decision sending method and system

Abstract

The invention relates to the technical field of driving early warning, and provides a decentralized risk decision sending method and system, and the method comprises the steps: calculating a current risk value of each vehicle terminal through first road condition data collected by a vehicle terminal and second road condition data collected by a roadside edge node based on all driving objects in a monitoring range; and calculating a predicted risk value corresponding to each current risk value through the risk early warning model, determining risk information when the predicted risk value is greater than or equal to a risk threshold value, and sending a corresponding risk decision instruction to the vehicle terminal based on the risk information. Tasks of data acquisition, risk value calculation and the like are dispersed to the vehicle terminal and the roadside edge node, so that the delay and calculation pressure of data transmission are effectively reduced, the real-time performance and response speed of risk early warning are improved, and the traffic accident risk caused by early warning lag is reduced.

Description

Technical Field

[0001] This application relates to the field of driving warning technology, and in particular to a decentralized risk decision sending method and system. Background Technology

[0002] In the field of intelligent connected vehicles, the vehicle-road-cloud integration system can achieve risk warning for vehicle driving through the collaborative interaction of vehicles, road infrastructure and cloud platforms. For example, it can integrate multi-sensory data from vehicle terminals and roadside, and with the help of the powerful computing power of the cloud.

[0003] Currently, most existing vehicle-road-cloud integrated systems employ a centralized control architecture for risk warning. This involves collecting traffic data from local areas and uploading it to a cloud platform. The cloud platform then integrates global information from different regions, performs complex risk assessments and calculations, and finally issues control or warning commands to relevant vehicles, thereby achieving risk warning.

[0004] However, this centralized early warning method through cloud platforms is prone to decision-making delays when faced with massive amounts of high-concurrency real-time traffic data, making it difficult to meet the timeliness of risk warning responses and resulting in a lag in risk warnings. Summary of the Invention

[0005] This application provides a decentralized risk decision sending method and system to solve the problem of lag in centralized early warning through cloud platforms.

[0006] In a first aspect, embodiments of this application provide a decentralized risk decision transmission method, including: The system acquires first road condition data collected by a vehicle terminal, and acquires second road condition data collected by a roadside edge node based on all driving objects within the monitoring range of the roadside edge node. The roadside edge nodes are spaced apart at the edge of the road, the vehicle terminal is installed on the vehicle, the driving objects include the vehicle, and the roadside edge nodes and the vehicle terminal respectively store risk warning models. Based on the second road condition data and the first road condition data of each vehicle terminal within the monitoring range, calculate the current risk value of each vehicle terminal within the monitoring range; Based on the risk warning model, the predicted risk value of each vehicle terminal is calculated according to the current risk value of each vehicle terminal. If the predicted risk value of any of the vehicle terminals exceeds the risk value threshold, risk information is determined based on the first predicted risk value that exceeds the risk value threshold. Based on the risk information, a corresponding risk decision instruction is sent to each vehicle terminal within the monitoring range, so that each vehicle terminal executes the corresponding risk decision instruction.

[0007] In some feasible embodiments, the step of calculating the predicted risk value for each vehicle terminal based on the risk warning model and the current risk value of each vehicle terminal includes: A local dynamic risk map corresponding to the roadside edge node is generated based on the current risk value of each vehicle terminal; The risk warning model calculates the predicted risk value for each vehicle terminal based on a preset number of local dynamic risk maps.

[0008] In some feasible embodiments, after the step of generating the local dynamic risk map corresponding to the roadside edge node based on the current risk value of each vehicle terminal, the method further includes: A global dynamic risk map is generated based on the local dynamic risk map corresponding to each roadside edge node at the same time. The global dynamic risk map is uploaded to the cloud platform to perform iterative updates on the risk warning model of the cloud platform, resulting in updated model parameters; the cloud platform is communicatively connected to the roadside edge nodes and the vehicle terminal. The updated model parameters are sent to each roadside edge node and each vehicle terminal within the monitoring range of each roadside edge node through the cloud platform, so as to update the model parameters of the risk warning model of the roadside edge node and the model parameters of the risk warning model of each vehicle terminal.

[0009] In some feasible embodiments, the step of sending a corresponding risk decision instruction to each vehicle terminal within the monitoring range based on the risk information includes: Acquire historical driving behavior data for each vehicle terminal within the monitoring range; Based on the historical driving behavior data corresponding to each vehicle terminal, calculate the vehicle score for each vehicle terminal. The vehicle terminal with the highest vehicle score is identified as the target vehicle terminal. Based on the risk information, risk decision instructions are generated for each vehicle terminal within the monitoring range through the target vehicle terminal; Send a corresponding risk decision instruction to each vehicle terminal within the monitoring range.

[0010] In some feasible embodiments, the step of, in the presence of multiple vehicle terminals with the highest vehicle ratings, further includes: Determine the computing power level corresponding to the vehicle terminal with the highest vehicle score; The vehicle terminal with the highest computing power level and the highest vehicle score is identified as the target vehicle terminal.

[0011] In some feasible embodiments, after the step of generating risk decision instructions for each vehicle terminal within the monitoring range based on the risk information via the target vehicle terminal, the method further includes: When there is a conflict between risk decision instructions corresponding to at least two of the vehicle terminals, the risk decision instructions and the conflict type corresponding to the risk decision instructions are sent to the cloud platform. Based on the risk information and the conflict type, the cloud platform regenerates the risk decision instructions corresponding to each vehicle terminal.

[0012] In some feasible embodiments, after the step of sending the corresponding risk decision instruction to each vehicle terminal within the monitoring range, the method further includes: Obtain feedback from each vehicle terminal within the monitoring range regarding each risk decision instruction, including whether to agree to execute or refuse to execute; If all vehicle terminals provide feedback agreeing to execute, control each vehicle terminal within the monitoring range to execute its corresponding risk decision instruction. If at least one of the vehicle terminals responds with a refusal to execute, a conflicting risk decision instruction is identified.

[0013] In some feasible embodiments, after the step of controlling each vehicle terminal within the monitoring range to execute its corresponding risk decision instruction when all vehicle terminals provide feedback of agreement to execution, the method further includes: Obtain the execution result of each vehicle terminal executing its corresponding risk decision instruction; The execution result is sent to the cloud platform; Based on the execution results, the risk warning model of the cloud platform is iteratively updated to obtain the updated model parameters.

[0014] In some feasible embodiments, calculating the current risk value of each vehicle terminal within the monitoring range based on the first and second road condition data of each roadside edge node includes: The location coordinates, driving speed, driving acceleration, and distance between the vehicle terminal and each driving object within the monitoring range are determined based on the first road condition data and the second road condition data. The current risk value of each vehicle terminal is calculated based on its location coordinates, driving speed, driving acceleration, and the distance between the vehicle terminal and each driving object within the monitoring range.

[0015] Secondly, embodiments of this application provide a decentralized risk decision transmission system, including: A risk warning model is stored in the vehicle terminal. The vehicle terminal is configured to collect first road condition data and send the first road condition data to the roadside edge node corresponding to the vehicle terminal. Multiple roadside edge nodes are spaced apart along the road edge. Each roadside edge node stores a risk warning model and is configured to collect second road condition data based on all driving objects within the monitoring range of the roadside edge node. Based on the second road condition data and the first road condition data of each vehicle terminal within the monitoring range, the current risk value of each vehicle terminal within the monitoring range is calculated. Based on the risk warning model, a predicted risk value of each vehicle terminal is calculated according to the current risk value of each vehicle terminal. If the predicted risk value of any vehicle terminal exceeds a risk value threshold, risk information is determined based on the first predicted risk value exceeding the risk value threshold. The vehicle terminal or the roadside edge node is further configured to send a corresponding risk decision instruction to each vehicle terminal within the monitoring range based on the risk information, so that each vehicle terminal executes the corresponding risk decision instruction.

[0016] In some feasible embodiments, the roadside edge node is further configured as follows: A local dynamic risk map corresponding to the roadside edge node is generated based on the current risk value of each vehicle terminal; The risk warning model calculates the predicted risk value for each vehicle terminal based on a preset number of local dynamic risk maps.

[0017] In some feasible embodiments, it further includes: a cloud platform that stores a risk warning model, and the cloud platform is communicatively connected to the roadside edge node and the vehicle terminal; The roadside edge node is also configured to generate a global dynamic risk map based on the local dynamic risk map corresponding to each roadside edge node at the same time, and upload the global dynamic risk map to the cloud platform. The cloud platform is configured to perform iterative updates on the risk warning model of the cloud platform through the global dynamic risk map to obtain updated model parameters, and send the updated model parameters to each of the roadside edge nodes and each of the vehicle terminals located within the respective monitoring range of each of the roadside edge nodes. Attached Figure Description

[0018] To more clearly illustrate the technical solution of this application, the drawings used in the embodiments will be briefly introduced below. Obviously, for those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0019] Figure 1 A flowchart illustrating a decentralized risk decision sending method provided in this application embodiment; Figure 2 A top view of the vehicle terminal and roadside edge node provided in an embodiment of this application; Figure 3 A flowchart for calculating the predicted risk value and updating the model parameters is provided for an embodiment of this application; Figure 4 A flowchart illustrating the synchronization of model parameters via a cloud platform, provided as an embodiment of this application; Figure 5 A flowchart illustrating the sending of risk decision instructions via a target vehicle terminal, provided in an embodiment of this application; Figure 6 This is a flowchart illustrating how the target vehicle terminal is determined based on computing power level, as described in an embodiment of this application. Figure 7 This is a flowchart illustrating the re-decision-making process for conflicting risk decision instructions, as described in an embodiment of this application. Figure 8 This is a flowchart illustrating how a risk decision instruction is determined based on feedback from each vehicle terminal, as described in this application. Figure 9 This is a flowchart illustrating how the model parameters are updated based on the execution result of a risk decision instruction, as described in an embodiment of this application. Figure 10 This is a flowchart for calculating the current risk value of the vehicle terminal in an embodiment of this application. Detailed Implementation

[0020] The embodiments will now be described in detail, examples of which are illustrated in the accompanying drawings. When the following description relates to the drawings, unless otherwise indicated, the same numbers in different drawings represent the same or similar elements. The embodiments described below do not represent all embodiments consistent with this application. They are merely examples of systems and methods consistent with some aspects of this application as detailed in the claims.

[0021] The Vehicle-Road-Cloud Integration (VCR) system can collect road condition data from multiple perspectives through collaborative interaction between vehicles, road infrastructure such as roadside edge nodes, and the cloud platform, thereby providing risk warnings for each vehicle on the road based on the road condition data.

[0022] Typically, a vehicle-road-cloud integrated system uploads road condition data collected by vehicle terminals and roadside edge nodes to a cloud platform simultaneously. This allows the cloud platform to leverage its computing power to perform centralized fusion processing, risk analysis, and decision generation on road condition data from different sources.

[0023] However, this centralized architecture has certain limitations in practical applications. For example, when massive amounts of road condition data are uploaded to the cloud simultaneously, data transmission latency increases. In some high-risk scenarios with extremely high real-time requirements, such as emergency avoidance and sudden accident response, the process of uploading road condition data to the cloud platform and then receiving a decision from the cloud platform takes a long time, which cannot meet the needs of vehicles to take timely evasive actions, resulting in a lag in risk warnings.

[0024] To address the aforementioned issues, this application provides a decentralized risk decision transmission method and system. This system enhances the real-time performance of risk warnings in the vehicle-road-cloud integrated system through collaborative decision-making and distributed processing by vehicle terminals, roadside edge nodes, and a cloud platform. The decentralized risk decision transmission system can include vehicle terminals and roadside edge nodes.

[0025] Figure 1 A flowchart illustrating a decentralized risk decision sending method provided in this application embodiment.

[0026] See Figure 1 This application provides a decentralized risk decision sending method, comprising: S100: Acquire first road condition data collected by the vehicle terminal, and acquire second road condition data collected by the roadside edge node based on all driving objects within the monitoring range of the roadside edge node.

[0027] In some embodiments, first road condition data can be acquired by the vehicle through a vehicle terminal, wherein the vehicle terminal is installed in the vehicle. The first road condition data may include the vehicle's own driving status data and surrounding environment data. Driving status data may include the vehicle's driving information, which can be obtained through the vehicle's GPS positioning module. Surrounding environment data can be collected by sensors such as cameras, millimeter-wave radar, and lidar mounted on the vehicle, and may include the relative distance and speed of vehicles ahead or to the sides, lane line information, traffic light status, the position of pedestrians and non-motorized vehicles, and their corresponding movement trends. Before acquiring the first road condition data, the vehicle terminal needs to obtain permission to collect the data. After obtaining permission, the vehicle terminal can begin collecting the first road condition data based on that permission.

[0028] In some embodiments, roadside edge nodes can be deployed at key locations along the road edge, such as intersections, tunnel entrances / exits, curves, or both sides of the road. Roadside edge nodes can integrate various detection devices, such as high-definition cameras and microwave radar, enabling multi-dimensional data collection of all vehicles within the monitoring range to obtain second road condition data. These vehicles can include motor vehicles, non-motor vehicles, and pedestrians. The second road condition data can include the real-time location coordinates, speed, and trajectory of each vehicle.

[0029] Figure 2 This is a top view of the vehicle terminal and roadside edge node provided in an embodiment of this application.

[0030] See Figure 2 Roadside edge nodes are set at the edge of the road, and collect second road condition data for each driving object within the monitoring fan-shaped monitoring range. For roads with two-way lanes, the monitoring range can cover the cross-sectional area of ​​the two-way lanes corresponding to the roadside edge nodes, so as to collect second road condition data simultaneously based on each driving object (such as vehicles) in the two-way lanes through the roadside edge nodes.

[0031] It should be noted that the distance between the roadside edge node and the road can be adjusted to change the coverage of the road by the monitoring range of the roadside edge node. The shape of the monitoring range can also be other shapes. This embodiment does not specifically limit the distance between the roadside edge node and the road, or the shape of the monitoring range.

[0032] In some embodiments, the vehicle terminal and roadside edge nodes each store a risk warning model. This risk warning model can be a Spatio-Temporal Graph Neural Network (STGNN) model, which includes a spatial convolutional module and a temporal convolutional module. The spatial convolutional module can capture the spatial dependencies between vehicle terminals using Chebyshev graph convolution. The temporal convolutional module can extract the temporal evolution features of the local risk map and can expand the receptive field through dilated convolution to capture long-term temporal dependencies.

[0033] In some embodiments, during the pre-training phase of the risk warning model, historical traffic data, accident case data, and various road condition data can be used as training data to train the initial model of the risk warning model. When a preset convergence condition is met, such as when the loss output by the initial model is less than or equal to a loss threshold, the pre-trained risk warning model is obtained. The risk warning model can then perform risk prediction for each vehicle terminal and output the predicted risk value for each vehicle terminal.

[0034] S200: Based on the second road condition data and the first road condition data of each vehicle terminal within the monitoring range, calculate the current risk value of each vehicle terminal within the monitoring range.

[0035] In some embodiments, vehicle terminals located within the monitoring range of a roadside edge node can send the collected first road condition data to the corresponding roadside edge node. The roadside edge node can calculate the current risk value of each vehicle terminal within its monitoring range based on the first road condition data uploaded by each vehicle terminal and the second road condition data it has collected.

[0036] The current risk value represents the probability that the vehicle corresponding to the vehicle terminal poses a real-time traffic hazard to the vehicle in its current driving state, given the surrounding traffic environment. By calculating the current risk value for each vehicle terminal, a quantitative basis can be provided for subsequent risk warnings and decisions based on the current risk value, helping the vehicle terminal to more accurately perceive the immediate traffic risk situation.

[0037] It should be noted that calculating the current risk value of a vehicle terminal based on the first and second road condition data requires ensuring that the vehicle terminal is within the monitoring range of the roadside edge node. This ensures that the roadside edge node can successfully receive the first road condition data uploaded by the vehicle terminal and then combine it with the second road condition data it has collected to calculate the current risk value of the vehicle terminal.

[0038] Based on the vehicle's driving status, the vehicle terminal will pass through the monitoring range of multiple roadside edge nodes. For example, when the vehicle terminal moves from the monitoring range of the first roadside edge node to the monitoring range of the second roadside edge node, the vehicle terminal will stop sending the real-time collected first road condition data to the first roadside edge node and start sending the real-time collected first road condition data to the second roadside edge node.

[0039] S300: Based on the risk warning model, calculate the predicted risk value of each vehicle terminal according to the current risk value of each vehicle terminal.

[0040] In some embodiments, since the roadside edge nodes and vehicle terminals each store a risk warning model, the corresponding predicted risk value can be calculated based on the current risk value of each vehicle terminal using the risk warning model of each roadside edge node and / or the risk warning model of the vehicle terminal.

[0041] The predicted risk value represents the level of risk that a vehicle terminal may face within a preset future timeframe. This preset timeframe can be flexibly set according to the actual application scenario. For example, for complex road conditions, a relatively short preset timeframe, such as 1 second or 3 seconds, can be set. For simple road conditions, a relatively long preset timeframe, such as 5 seconds or 10 seconds, can be set.

[0042] In some embodiments, after calculating the current risk value corresponding to each vehicle terminal, the roadside edge node can send its corresponding current risk value to each vehicle terminal so that the vehicle terminal can calculate its predicted risk value based on its own current risk value according to the risk warning model.

[0043] In some embodiments, after calculating the current risk value corresponding to each vehicle terminal, the roadside edge node can store the current risk value in the risk warning model of the roadside edge node, so as to calculate the predicted risk value of each vehicle terminal through the risk warning model.

[0044] In some embodiments, to improve the accuracy of calculating predicted risk values, the risk warning model can extract feature information of the current road segment based on first road condition data and / or second road condition data, and combine the feature information of the current road segment with the current risk value of the vehicle terminal for comprehensive calculation, outputting the predicted risk value of the vehicle terminal at a preset future time. For example, in calculating the predicted risk value, the roadside edge node can combine the first road condition data uploaded by all vehicle terminals within the monitoring range. Since each vehicle's location is different, the first road condition data collected by the corresponding vehicle terminal is also different. For example, for a two-way road, the first road condition data collected by the vehicle terminal on this side is different from that collected by the vehicle terminal on the opposite side. If there is a pedestrian in a blind spot that the vehicle terminal on the opposite side cannot detect, then the location of the pedestrian in the blind spot can be determined by combining the first road condition data collected by the vehicle terminal on this side. The predicted risk value of the vehicle terminal on the opposite side will be increased based on the pedestrian in the blind spot to avoid collisions between the vehicle on the opposite side and the pedestrian in the blind spot, thereby improving the accuracy of calculating the predicted risk value of each vehicle terminal and reducing the occurrence of traffic accidents.

[0045] S400: If the predicted risk value of any of the vehicle terminals exceeds a risk value threshold, risk information is determined based on the first predicted risk value that exceeds the risk value threshold.

[0046] If a roadside edge node detects that the predicted risk value of any vehicle terminal exceeds the risk value threshold, the roadside edge node can determine that there is a vehicle with a safety risk within the monitoring range within a preset time. Therefore, the roadside edge node can determine the risk information based on a first predicted risk value, which is a predicted risk value that exceeds the risk value threshold.

[0047] Within the monitoring range, there can be multiple first predicted risk values, indicating that multiple vehicles corresponding to these multiple first predicted risk values ​​all pose a safety risk. Risk information can be represented as specific risk types, such as an accident ahead, icy road ahead, or pedestrians in a blind spot.

[0048] Based on these risk types, risk information can also include risk levels. Risk levels can be classified according to the difference between the first predicted risk value and the risk value threshold. For example, if the difference is relatively small, the risk level is mild; if the difference is relatively large, the risk level is high. The number of risk levels can be determined according to the actual situation; this embodiment does not specifically limit the number of levels.

[0049] S500: Based on the risk information, send a corresponding risk decision instruction to each vehicle terminal within the monitoring range, so that each vehicle terminal executes the corresponding risk decision instruction.

[0050] After obtaining risk information, risk decision instructions can be generated for each vehicle terminal based on the risk information, and the risk decision instructions can be sent to the corresponding vehicle terminal so that the vehicle terminal can execute the risk decision instructions, thereby predicting traffic risks in advance.

[0051] Each vehicle terminal can receive different risk decision instructions. For example, in the event of an accident on the road ahead, the risk decision instruction received by the vehicle terminal closest to the accident location may be "reduce speed by 80%", while the risk decision instruction received by the vehicle terminal relatively far from the accident location may be "change lanes to the left of the current lane". This allows for differentiated risk warning strategies based on the different positions of different vehicles on the road, generating risk decision instructions that are adapted to road conditions and safe for each vehicle terminal.

[0052] In some embodiments, risk decision instructions can be generated by roadside edge nodes or by one of the vehicle terminals within the monitoring range of a roadside edge node. During the generation of risk decision instructions for each vehicle terminal, the vehicle terminal's driving information, such as current speed, acceleration, and direction of travel, as well as the vehicle type (e.g., sedan, truck, bus), can be considered to generate the corresponding risk decision instructions. For example, a large truck, due to its large footprint, cannot change lanes to other vehicles; therefore, a deceleration risk decision instruction can be generated for the truck's terminal. Conversely, a sedan, with its relatively smaller footprint, can generate a lane-changing risk decision instruction.

[0053] In some embodiments, for a vehicle terminal, multiple risk decision commands can be generated. For example, the first risk decision command is to reduce speed by 50%, and the second risk decision command is to change lanes to the right. Through multiple simple risk decision commands, the vehicle corresponding to the vehicle terminal can be controlled to perform complex driving actions.

[0054] Based on the above technical solution, risk warning models are stored at roadside edge nodes and vehicle terminals respectively. The current risk value and predicted risk value of each vehicle terminal are calculated in real time based on the first road condition data collected by the vehicle terminal and the second road condition data collected by the roadside edge node. When the predicted risk value exceeds the threshold, the roadside edge node or vehicle terminal directly sends a risk decision command to the vehicle terminal within the monitoring range. This avoids the data transmission delay and calculation delay caused by uploading all road condition data to the cloud platform for centralized processing, improves the real-time performance and response speed of risk warning, and effectively reduces the risk of traffic accidents caused by delayed warnings.

[0055] Figure 3 This is a flowchart illustrating the calculation of predicted risk values ​​and the updating of model parameters, provided for embodiments of this application.

[0056] See Figure 3 In some embodiments, step S300 may include the following steps: S310: Generate a local dynamic risk map corresponding to the roadside edge node based on the current risk value of each vehicle terminal.

[0057] In some embodiments, after calculating the current risk value of each vehicle terminal, the roadside edge node can generate a local dynamic risk map corresponding to the roadside edge node based on the current risk value of each vehicle terminal. The local dynamic risk map can intuitively represent the real-time risk distribution of each vehicle terminal within the monitoring range. In the local dynamic risk map, each vehicle terminal can be represented as a node, and the node position can correspond to the position of the vehicle terminal within the monitoring range. The interaction strength between any two nodes can be represented as an edge, and the weight value of the edge can be determined based on the driving conditions between the two vehicle terminals, such as the relative distance, relative speed, driving direction angle, and historical interaction records between the two vehicle terminals.

[0058] For example, if two vehicles are in the same lane, close to each other, and have a positive relative speed (i.e., the rear vehicle's speed is greater than the front vehicle's speed), the edge weight will increase accordingly to indicate a risk of a rear-end collision. If the two vehicles are traveling in opposite directions and the distance between them gradually increases, the edge weight will decrease, indicating a lower risk of interaction between the two vehicles.

[0059] S320: Using the risk warning model, calculate the predicted risk value for each vehicle terminal based on a preset number of local dynamic risk maps.

[0060] The roadside edge node can input a preset number of local dynamic risk maps into the risk warning model of the roadside edge node. The risk warning model can identify the points and edges between the local dynamic risk maps, determine the dynamic correlation between multiple vehicle terminals within the monitoring range, and thus calculate the predicted risk value of each vehicle terminal.

[0061] The preset number is the number of local dynamic risk maps generated by roadside edge nodes within a preset time period in the future. The risk warning model can identify the dynamic correlation between different vehicle terminals in each local dynamic risk map in the order of generation of the local dynamic risk maps, thereby improving the accuracy of calculating the predicted risk value of each vehicle terminal.

[0062] In some embodiments, the roadside edge node can also send the local dynamic risk map to each vehicle terminal, so that the vehicle terminal can calculate its own predicted risk value based on the local dynamic risk map through its own risk warning model. The method of calculating the predicted risk value can be referred to the above embodiments, and will not be repeated here.

[0063] Based on the above technical solutions, the real-time risk distribution of each vehicle terminal within the monitoring range can be intuitively determined through the local dynamic risk map. Combined with the generation order of the local dynamic risk map, the risk warning model can more accurately predict the risk trend of vehicle terminals in the future period, thereby improving the accuracy and reliability of the predicted risk value.

[0064] In some embodiments, the decentralized risk decision sending system further includes a cloud platform that stores a risk warning model. The risk warning model on the cloud platform uses the same model parameters as the risk warning models of the roadside edge nodes and the vehicle terminals. Therefore, the cloud platform can iteratively train and update the model parameters of the risk warning model, and synchronize these parameters to the roadside edge nodes and vehicle terminals.

[0065] See also Figure 3 After step S310, the following steps may also be included: S330: Generate a global dynamic risk map based on the local dynamic risk map corresponding to each of the roadside edge nodes at the same time.

[0066] In some embodiments, multiple roadside edge nodes are provided at road intervals. At the same time, each roadside edge node generates a local dynamic risk map based on the current risk value calculated by the respective roadside edge node based on the vehicle terminals within its corresponding monitoring range, in order to determine the risk situation corresponding to each monitoring range.

[0067] By fusing these locally generated dynamic risk maps at the same time, a global dynamic risk map representing the overall risk situation can be generated. During the fusion process, for multiple locally generated dynamic risk maps that may have overlapping areas, redundant map data can be offset by methods such as weighted averaging. For multiple locally generated dynamic risk maps that do not have overlapping areas, coordinate alignment can be performed based on the locally generated dynamic risk maps to stitch them together, ultimately obtaining the global dynamic risk map.

[0068] S340: Upload the global dynamic risk map to the cloud platform to perform iterative updates on the risk warning model of the cloud platform and obtain the updated model parameters.

[0069] In some embodiments, roadside edge nodes can upload a global dynamic risk map to a cloud platform. By leveraging the computing resources of the cloud platform, the risk warning model on the cloud platform can be iteratively updated. This allows the risk warning model to learn the risk distribution characteristics of the current road based on the latest global dynamic risk map input in real time, and update the risk parameters of the risk warning model, thereby improving the adaptability and prediction accuracy of the risk warning model to complex traffic scenarios.

[0070] It should be noted that the global dynamic risk map includes local dynamic risk maps corresponding to multiple roadside edge nodes. Therefore, the cloud platform can capture the characteristic relationships between different roadside edge nodes through the local dynamic risk maps corresponding to different roadside edge nodes, thereby completing the training of the cloud platform's risk prediction model. For example, if a traffic accident occurs on a certain road segment, another road located upstream of that road will also experience congestion and other impacts due to the traffic accident. Thus, through global judgment, the model parameters of the roadside edge nodes and vehicle terminals are updated.

[0071] S350: The updated model parameters are sent to each roadside edge node and each vehicle terminal within the monitoring range of each roadside edge node through the cloud platform, so as to update the model parameters of the risk warning model of the roadside edge node and the model parameters of the risk warning model of each vehicle terminal.

[0072] Figure 4 This is a flowchart illustrating the synchronization of model parameters via a cloud platform, as provided in an embodiment of this application.

[0073] See Figure 4 The cloud platform can communicate with roadside edge nodes and vehicle terminals to send updated model parameters to each roadside edge node and each vehicle terminal within the monitoring range of each roadside edge node. This allows for the synchronous updating of the model parameters of the risk warning model of the roadside edge node and the risk warning model of each vehicle terminal, ensuring that the roadside edge node and vehicle terminal can input predicted risk values ​​based on the latest risk distribution characteristics.

[0074] Based on the above technical solutions, this embodiment generates a global dynamic risk map by integrating the local dynamic risk maps of each roadside edge node. This enables the cloud platform to update the model parameters of the risk warning model based on the global dynamic risk map and distribute the updated model parameters to the roadside edge nodes and vehicle terminals, thereby continuously improving the adaptability and accuracy of the distributed risk warning model.

[0075] In some embodiments, risk decision instructions can be generated by roadside edge nodes and vehicle terminals. When a risk decision instruction is generated by a vehicle terminal, it is necessary to determine the target vehicle terminal for generating the risk decision instruction among multiple vehicle terminals.

[0076] Figure 5 This is a flowchart illustrating the sending of risk decision instructions via a target vehicle terminal, as provided in an embodiment of this application.

[0077] See Figure 5 Step S500 may include the following steps: S510: Obtain historical driving behavior data for each vehicle terminal within the monitoring range.

[0078] In some embodiments, historical driving behavior data for each vehicle terminal within the monitoring range can be obtained through roadside edge nodes. This historical driving behavior data may include records of various driving operations performed by the vehicle terminal within a preset historical time period, as well as its compliance with traffic rules. By acquiring the historical driving behavior of each vehicle terminal within the monitoring range, it is possible to determine the driving habits, risk preferences, and degree of compliance with traffic rules of the driver corresponding to each vehicle terminal from multiple dimensions, thereby facilitating the determination of a vehicle score for each vehicle terminal based on the historical driving behavior data.

[0079] S520: Calculate the vehicle score for each vehicle terminal based on the historical driving behavior data corresponding to each vehicle terminal.

[0080] In some embodiments, roadside edge nodes can calculate a vehicle score for each vehicle terminal based on its historical driving behavior data. The evaluation index system may include driving conduct, adherence to traffic rules, etc. The driving conduct dimension may include indicators such as the frequency of rapid acceleration, the frequency of sudden braking, and the number of consecutive lane changes. For example, if the proportion of rapid accelerations to total driving time for a vehicle terminal within a preset historical time period exceeds a set threshold, the vehicle score for that vehicle terminal will be lowered.

[0081] The degree of compliance with traffic rules can include the number of traffic violations such as running red lights, speeding, and illegal parking, and the vehicle score will be reduced based on the corresponding traffic violations. For example, if a vehicle at a certain terminal has no traffic violations in the past month and its driving behavior remains consistently good, the vehicle score for that terminal can be increased.

[0082] S530: The vehicle terminal with the highest vehicle score is identified as the target vehicle terminal.

[0083] In some embodiments, the roadside edge node can compare vehicle scores within its monitoring range to identify the vehicle terminal with the highest score as the target vehicle terminal. For example, if the predicted risk value calculated by the roadside edge node or any vehicle terminal within its monitoring range is greater than or equal to a risk value threshold, the roadside edge node will retrieve and sort the vehicle scores of all vehicle terminals within its monitoring range in real time. If vehicle A has a score of 95, vehicle B has a score of 88, and vehicle C has a score of 92, then vehicle A can be identified as the target vehicle terminal, and the roadside edge node will prioritize sending risk information to vehicle A.

[0084] In some embodiments, based on the vehicle's movement state, the target vehicle terminal may move out of the monitoring range corresponding to the roadside edge node. In this case, the roadside edge node can re-execute steps S510-S530 to change the target vehicle terminal in real time according to the changes in vehicles within the monitoring range.

[0085] In some embodiments, after determining the target terminal, alternative vehicle terminals can be determined based on the order of vehicle scores from high to low. If the target vehicle terminal is unable to generate risk decision instructions, the alternative vehicle terminal can take over the generation of risk decision instructions from the target vehicle terminal.

[0086] S540: Based on the risk information, generate a risk decision instruction for the vehicle terminal within the monitoring range through the target vehicle terminal.

[0087] After the target vehicle terminal receives risk information, it can generate risk decision instructions for vehicle terminals within the monitoring range based on the risk information. At this time, the roadside edge node can send the first road condition data uploaded by each vehicle terminal within the monitoring range to the target vehicle terminal, so that the target vehicle terminal can perform comprehensive analysis based on more comprehensive road condition information, and generate corresponding risk decision instructions for each vehicle terminal within the monitoring range according to the risk type and preset decision rules. These risk decision instructions will provide corresponding risk decision instructions based on vehicle terminals at different locations within the monitoring range to ensure the relevance and executability of the instructions.

[0088] S550: Send the corresponding risk decision instruction to the vehicle terminal within the monitoring range.

[0089] After generating risk decision instructions for vehicle terminals within the monitoring range, the target vehicle terminal can send corresponding risk decision instructions to each vehicle terminal within the monitoring range of the roadside edge node corresponding to the target vehicle terminal.

[0090] In some embodiments, risk decision instructions for different vehicle terminals may include different execution sequences. For example, for vehicle terminals whose predicted risk value is higher than a risk value threshold, the execution priority of the risk decision instruction may be set to the highest, requiring immediate execution of evasive actions such as deceleration and lane changing. For ease of description, the area where vehicle terminals with predicted risk values ​​higher than the risk value threshold are located will be referred to as the risk source.

[0091] For vehicle terminals located close to the risk source, their own risk decision-making instructions can be executed after the corresponding risk decision-making instructions are executed at the vehicle terminal corresponding to the risk source. For vehicle terminals located far from the risk source, the execution of risk decision-making instructions can be delayed, or only a warning prompt can be executed, provided safety is ensured. In this way, based on the execution sequence, it is possible to avoid traffic congestion or increased risk probability that may be caused by all vehicle terminals executing risk decision-making instructions simultaneously, thereby improving the safety and efficiency of the overall traffic flow.

[0092] Based on the above technical solutions, this embodiment evaluates the decision reliability of each vehicle terminal through vehicle scoring, and uses the vehicle terminal with the highest reputation value as the target vehicle terminal for instruction generation. This can improve the credibility and acceptability of risk decision instructions. At the same time, it utilizes the distributed computing power of vehicle terminals to share the decision load of roadside edge nodes, avoids data-driven processing, and improves the efficiency of decision instruction generation.

[0093] Figure 6 This is a flowchart illustrating how the target vehicle terminal is determined based on computing power level in an embodiment of this application.

[0094] See Figure 6 In some embodiments, if multiple vehicle terminals with the highest vehicle ratings exist simultaneously, step S530 may further include the following steps: S560: Determine the computing power level corresponding to the vehicle terminal with the highest vehicle score.

[0095] In some embodiments, roadside edge nodes can obtain the computing power level of the vehicle terminal with the highest vehicle rating. The computing power level represents the computing processing capability of the vehicle terminal and can be determined based on the hardware configuration of the vehicle terminal. For example, the computing power level can be divided into three levels: high, medium, and low. Vehicle terminals equipped with high-performance multi-core processors can be classified as high-level, vehicle terminals equipped with medium-performance processors can be classified as medium-level, and vehicle terminals equipped with basic-performance processors can be classified as low-level.

[0096] S570: The vehicle terminal with the highest vehicle score and the highest computing power level is identified as the target vehicle terminal.

[0097] After determining their respective computing power level, each vehicle terminal can send its computing power level to the roadside edge node. This allows the roadside edge node to compare the computing power levels of each vehicle terminal and identify the vehicle terminal with the highest computing power level as the target vehicle terminal. For example, if there are two vehicle terminals with the highest reputation scores, one with a high computing power level and the other with a medium computing power level, then the vehicle terminal with the high computing power level will be directly identified as the target vehicle terminal.

[0098] Based on the above technical solutions, when vehicles have the same score, the vehicle terminal with stronger computing power is selected first to execute the risk decision-making instruction generation task. This ensures the timeliness and accuracy of instruction calculation in complex risk scenarios, avoids decision delays due to insufficient computing power, and improves the processing efficiency of risk decisions.

[0099] Figure 7 This is a flowchart illustrating the re-decision-making process for conflicting risk decision instructions, as described in an embodiment of this application.

[0100] See Figure 7 In some embodiments, after step S500, the following step is further included: S600: When there is a conflict between risk decision instructions corresponding to at least two of the vehicle terminals, the risk decision instruction and the conflict type corresponding to the risk decision instruction are sent to the cloud platform.

[0101] Each risk decision instruction requires a vote from all vehicle terminals within the monitoring range to ensure consensus on each instruction before execution. If a risk decision instruction from one vehicle terminal conflicts with that of another, a disagreement will arise in the voting process among all vehicle terminals.

[0102] Therefore, when risk decision commands from at least two vehicle terminals conflict, the target vehicle terminal can send the conflicting risk decision commands and their corresponding conflict types to the cloud platform, allowing the cloud platform to arbitrate the conflicting risk decision commands. Conflict types can include driving path conflicts, speed control conflicts, and emergency avoidance priority conflicts, among others.

[0103] S700: Through the cloud platform, risk decision instructions corresponding to each vehicle terminal are regenerated based on the risk information and the conflict type.

[0104] The cloud platform can combine global risk information and conflict types to regenerate risk decision instructions for each vehicle terminal. It can also retrieve historical decision data—decision data with the same conflict type within a preset historical time period—as a reference. Based on global risk information, it can regenerate risk decision instructions for each vehicle terminal, enabling unified decision-making through the cloud platform and preventing traffic accidents caused by distributed decision-making disagreements.

[0105] Based on the above technical solutions, when conflicts arise in local decisions, arbitration and re-decision-making can be carried out through the cloud platform based on global risk information and conflict type, which can resolve conflicts between risk decision instructions and ensure the consistency and security of risk decisions.

[0106] Figure 8 This is a flowchart illustrating how a risk decision instruction is determined based on feedback from each vehicle terminal, as described in this application.

[0107] See Figure 8 In some embodiments, step S550 may include the following steps: S551: Obtain feedback from each vehicle terminal within the monitoring range regarding each risk decision instruction.

[0108] After the target vehicle terminal generates a risk decision instruction for each vehicle terminal within the monitoring range, all vehicle terminals within the monitoring range need to provide feedback on each risk decision instruction. The feedback may include agreeing to execute or refusing to execute.

[0109] For example, if a vehicle terminal believes that a risk decision instruction generated by a target vehicle terminal has issues such as path planning conflicts or insufficient safety distance, it can respond with "refuse to execute". If it considers the instruction reasonable and feasible, it can respond with "agree to execute". By collecting feedback from all vehicle terminals within its monitoring range on each risk decision instruction, roadside edge nodes or target vehicle terminals provide data support for subsequent conflict arbitration and unified decision-making, further improving the rationality and accuracy of decision-making.

[0110] S552: If all the feedback from the vehicle terminals is "agree to execute", control each of the vehicle terminals within the monitoring range to execute its corresponding risk decision instruction.

[0111] For a risk decision instruction, if all vehicle terminals within the monitoring range generate feedback agreeing to execute the risk decision instruction, then the vehicle terminals corresponding to the risk decision instruction can execute the risk decision instruction according to the specified execution order. If all feedback for each risk decision instruction is in agreement to execute, the target vehicle terminal can control each of the vehicle terminals within the monitoring range to execute its respective risk decision instruction.

[0112] S553: ​​If the feedback from at least one of the vehicle terminals is a refusal to execute, determine that there is a conflicting risk decision instruction.

[0113] If at least one vehicle terminal refuses to execute a risk decision instruction, it indicates that different vehicle terminals have differing opinions on the risk decision instruction. In this case, the target vehicle terminal can identify the risk decision instruction as a conflicting risk decision instruction and upload each feedback corresponding to the risk decision instruction to the cloud platform for unified decision-making in the future.

[0114] In some embodiments, if a vehicle terminal fails to send feedback on the risk decision instruction to the target vehicle terminal within a preset time, the target vehicle terminal can determine that there is a conflict in the risk decision instruction.

[0115] Based on the above technical solutions, this embodiment uses a distributed consensus mechanism to confirm the feasibility of each risk decision instruction through feedback from each vehicle terminal. This avoids execution resistance that may be caused by unilateral decision-making, improves the success rate of instruction execution, and promptly identifies conflicting instructions when the feedback is a refusal to execute, providing a basis for subsequent conflict handling.

[0116] Figure 9 This is a flowchart illustrating how the model parameters are updated based on the execution results of risk decision instructions, as described in this application.

[0117] See Figure 9 In some embodiments, after step S552, the following steps are also included: S554: Obtain the execution result of each vehicle terminal executing its corresponding risk decision instruction.

[0118] In some embodiments, after each vehicle terminal executes its corresponding risk decision instruction, the target vehicle terminal monitors the execution results fed back by each vehicle terminal. The execution results may include the execution status of the risk decision instruction, such as successful execution, partial execution, or execution anomaly.

[0119] After the vehicle terminal executes the risk decision instruction, if the execution status is successful, the vehicle terminal will send a successful execution result to the target vehicle terminal. If the vehicle terminal does not execute the risk decision instruction within a preset time, the vehicle terminal will send an execution result indicating an execution error to the target vehicle terminal. If part of the driving actions of the risk decision instruction are successfully executed, but another part of the driving actions are not completed within the preset time, the vehicle terminal will send a partial execution result to the target vehicle terminal.

[0120] S555: Send the execution result to the cloud platform.

[0121] In some embodiments, after obtaining the execution results of each vehicle terminal, the target vehicle terminal can send the execution results of each vehicle terminal to the cloud platform through the communication connection between the target vehicle terminal and the cloud platform, so that the cloud platform can uniformly store the execution status of the risk decision instructions of each vehicle terminal.

[0122] S556: Based on the execution result, perform iterative updates on the risk warning model of the cloud platform to obtain the updated model parameters.

[0123] In some embodiments, the driving status and location of different vehicle terminals change based on different risk decision instructions. Therefore, the cloud platform needs to iteratively update the cloud platform's risk warning model based on the execution results of each vehicle terminal.

[0124] After receiving the execution results from each vehicle terminal, the cloud platform can iteratively update the risk warning model based on the execution results to obtain the updated model parameters. Then, the cloud platform can send the updated model parameters to the vehicle terminal and the roadside edge node respectively through the communication connection with the vehicle terminal and the communication connection with the roadside edge node, so as to update the model parameters of the risk warning model of the vehicle terminal and the model parameters of the risk warning model of the roadside edge node respectively.

[0125] Based on the above technical solutions, this embodiment obtains the execution results of risk decision instructions executed by each vehicle terminal and uploads them to the cloud, enabling the cloud platform to iteratively update the risk warning model based on the execution results, forming a closed-loop update mechanism, and improving the practicality and scenario adaptability of the risk warning model.

[0126] Figure 10 This is a flowchart for calculating the current risk value of the vehicle terminal in an embodiment of this application.

[0127] See Figure 10 In some embodiments, step S200 may include the following steps: S210: Determine the position coordinates, driving speed, driving acceleration, and distance between the vehicle terminal and each driving object within the monitoring range based on the first road condition data and the second road condition data.

[0128] In some embodiments, roadside edge nodes can determine the position coordinates, speed, and acceleration of vehicle terminals by parsing first road condition data uploaded by vehicle terminals within the monitoring range and combining it with second road condition data collected by themselves. After determining the position coordinates of each vehicle terminal within the monitoring range, the roadside unit node can also determine the position coordinates of each driving object within the monitoring range using the second road condition data, and calculate the position coordinates of the vehicle terminals and the position coordinates of each driving object to obtain the distance between the vehicle terminals and each driving object within the monitoring range.

[0129] S220: Calculate the current risk value of each vehicle terminal based on its position coordinates, driving speed, driving acceleration, and the distance between the vehicle terminal and each driving object within the monitoring range.

[0130] In some embodiments, roadside edge nodes can calculate the current risk value of each vehicle terminal at its corresponding location coordinates using risk field quantization, wherein the formula for risk field quantization is: ； in, Let i be the current risk value of vehicle terminal i. Let be the position coordinates of vehicle terminal i. For vehicle terminals i driving speed, For vehicle terminals i The acceleration of the vehicle. This refers to the distance between the vehicle terminal and the object being driven. The first adjustment parameter is... For the second adjustment parameter and This is the third adjustment parameter. The first, second, and third adjustment parameters are all set based on empirical values. The first adjustment parameter adjusts the sensitivity of the vehicle terminal's speed to risk value calculation; the second adjustment parameter adjusts the sensitivity of the vehicle terminal's acceleration to risk value calculation; and the third adjustment parameter adjusts the sensitivity of the vehicle terminal's distance from each traveling object to risk value calculation.

[0131] Based on the above formula, the current risk value of each vehicle terminal can be calculated according to the vehicle terminal's position coordinates, driving speed, driving acceleration, and the distance between the vehicle terminal and each driving object within the monitoring range.

[0132] Based on the above technical solutions, this embodiment uses the formula corresponding to risk field quantification to calculate the current risk value of the vehicle terminal based on the vehicle terminal's position coordinates, driving speed, driving acceleration, and the distance between the vehicle terminal and each driving object within the monitoring range. This improves the accuracy of calculating the current risk value through multiple dynamic parameters.

[0133] This application embodiment also provides a decentralized risk decision sending system, including a vehicle terminal installed in a vehicle, the vehicle terminal storing a risk warning model, the vehicle terminal being configured to collect first road condition data and send the first road condition data to the roadside edge node corresponding to the vehicle terminal.

[0134] In some embodiments, the vehicle terminal may include environmental perception sensors, such as cameras and radar, for collecting first road condition data. The vehicle terminal may also include a V2X communication unit for communicating with roadside edge nodes.

[0135] The decentralized risk decision-making and transmission system also includes multiple roadside edge nodes spaced at intervals along the road edge. Each roadside edge node stores a risk warning model and is configured to: collect second road condition data from all driving objects within the monitoring range of the roadside edge node; calculate the current risk value of each vehicle terminal within the monitoring range based on the second road condition data and the first road condition data of each vehicle terminal within the monitoring range; calculate the predicted risk value of each vehicle terminal based on the risk warning model and the current risk value of each vehicle terminal; and determine the risk information based on the first predicted risk value exceeding the risk value threshold if the predicted risk value of any vehicle terminal exceeds the risk value threshold.

[0136] In some embodiments, each roadside edge node may include a roadside unit (RSU) and an edge computing server. The roadside unit is used to establish short-range communication connections with vehicle terminals, such as through a PC5 interface. The edge computing server is used to connect to other roadside edge nodes via wired fiber optic cables and to perform computational tasks for the roadside edge nodes, such as generating risk decision instructions and calculating the current risk value of each vehicle terminal within the monitoring range.

[0137] In some embodiments, the vehicle terminal or roadside edge node is further configured to send a corresponding risk decision instruction to each vehicle terminal within the monitoring range based on risk information, so that each vehicle terminal executes the corresponding risk decision instruction.

[0138] In some embodiments, the roadside edge node is further configured to: generate a local dynamic risk map corresponding to the roadside edge node based on the current risk value of each vehicle terminal, and calculate the predicted risk value of each vehicle terminal based on a preset number of local dynamic risk maps using a risk warning model.

[0139] In some embodiments, the decentralized risk decision sending system further includes a cloud platform, which stores a risk warning model and is communicatively connected to roadside edge nodes and vehicle terminals. The roadside edge nodes are further configured to generate a global dynamic risk map based on the local dynamic risk map corresponding to each roadside edge node at the same time, and upload the global dynamic risk map to the cloud platform through the communication connection between the roadside edge nodes and the cloud platform.

[0140] The cloud platform is configured to iteratively update the risk warning model of the cloud platform through a global dynamic risk map, obtain updated model parameters, and send the updated model parameters to each roadside edge node and each vehicle terminal within the monitoring range of each roadside edge node, thereby updating the model parameters of the risk warning model of the roadside edge node and the model parameters of the risk warning model of each vehicle terminal.

[0141] As can be seen from the above technical solutions, this application provides a decentralized risk decision transmission method and system. It calculates the current risk value for each vehicle terminal using first road condition data collected by vehicle terminals and second road condition data collected by roadside edge nodes from all driving objects within the monitoring range. Then, using a risk warning model, it calculates the predicted risk value corresponding to each current risk value. If the predicted risk value is greater than or equal to a risk threshold, the risk information is determined, and a corresponding risk decision instruction is sent to the vehicle terminal based on the risk information. This application effectively reduces data transmission latency and computational pressure by distributing tasks such as data collection and risk value calculation to vehicle terminals and roadside edge nodes, improving the real-time performance and response speed of risk warnings, and reducing the risk of traffic accidents caused by delayed warnings.

[0142] Similar parts between the embodiments provided in this application can be referred to mutually. The specific implementation methods provided above are only a few examples under the overall concept of this application and do not constitute a limitation on the scope of protection of this application. For those skilled in the art, any other implementation methods extended from the solution of this application without creative effort shall fall within the scope of protection of this application.

Claims

1. A decentralized risk decision transmission method, characterized in that, include: The system acquires first road condition data collected by a vehicle terminal, and acquires second road condition data collected by a roadside edge node based on all driving objects within the monitoring range of the roadside edge node. The roadside edge nodes are spaced apart at the edge of the road, the vehicle terminal is installed on the vehicle, the driving objects include the vehicle, and the roadside edge nodes and the vehicle terminal respectively store risk warning models. Based on the second road condition data and the first road condition data of each vehicle terminal within the monitoring range, calculate the current risk value of each vehicle terminal within the monitoring range; Based on the risk warning model, the predicted risk value of each vehicle terminal is calculated according to the current risk value of each vehicle terminal. If the predicted risk value of any of the vehicle terminals exceeds the risk value threshold, risk information is determined based on the first predicted risk value that exceeds the risk value threshold. Based on the risk information, a corresponding risk decision instruction is sent to the vehicle terminals within the monitoring range, so that each vehicle terminal executes the corresponding risk decision instruction.

2. The decentralized risk decision sending method according to claim 1, characterized in that, The step of calculating the predicted risk value for each vehicle terminal based on the risk warning model and the current risk value of each vehicle terminal includes: A local dynamic risk map corresponding to the roadside edge node is generated based on the current risk value of each vehicle terminal; The risk warning model calculates the predicted risk value for each vehicle terminal based on a preset number of local dynamic risk maps.

3. The decentralized risk decision sending method according to claim 2, characterized in that, After the step of generating the local dynamic risk map corresponding to the roadside edge node based on the current risk value of each vehicle terminal, the method further includes: A global dynamic risk map is generated based on the local dynamic risk map corresponding to each roadside edge node at the same time. The global dynamic risk map is uploaded to the cloud platform to perform iterative updates on the risk warning model of the cloud platform, resulting in updated model parameters; the cloud platform is communicatively connected to the roadside edge nodes and the vehicle terminal. The updated model parameters are sent to each roadside edge node and each vehicle terminal within the monitoring range of each roadside edge node through the cloud platform, so as to update the model parameters of the risk warning model of the roadside edge node and the model parameters of the risk warning model of each vehicle terminal.

4. The decentralized risk decision sending method according to claim 3, characterized in that, The step of sending a corresponding risk decision instruction to the vehicle terminal within the monitoring range based on the risk information includes: Acquire historical driving behavior data for each vehicle terminal within the monitoring range; Based on the historical driving behavior data corresponding to each vehicle terminal, calculate the vehicle score for each vehicle terminal. The vehicle terminal with the highest vehicle score is identified as the target vehicle terminal. Based on the risk information, a risk decision instruction for the vehicle terminal within the monitoring range is generated through the target vehicle terminal; Send the corresponding risk decision instruction to the vehicle terminal within the monitoring range.

5. The decentralized risk decision sending method according to claim 3, characterized in that, In the case where there are multiple vehicle terminals with the highest vehicle ratings, the following is also included: Determine the computing power level corresponding to the vehicle terminal with the highest vehicle score; The vehicle terminal with the highest computing power level and the highest vehicle score is identified as the target vehicle terminal.

6. The decentralized risk decision sending method according to claim 4, characterized in that, After the step of generating risk decision instructions for each vehicle terminal within the monitoring range based on the risk information via the target vehicle terminal, the method further includes: When there is a conflict between risk decision instructions corresponding to at least two of the vehicle terminals, the risk decision instructions and the conflict type corresponding to the risk decision instructions are sent to the cloud platform. Based on the risk information and the conflict type, the cloud platform regenerates the risk decision instructions corresponding to each vehicle terminal.

7. The decentralized risk decision sending method according to claim 6, characterized in that, After the step of sending the corresponding risk decision instruction to each vehicle terminal within the monitoring range, the method further includes: Obtain feedback from each vehicle terminal within the monitoring range regarding each risk decision instruction, including whether to agree to execute or refuse to execute; If all vehicle terminals provide feedback agreeing to execute, control each vehicle terminal within the monitoring range to execute its corresponding risk decision instruction. If at least one of the vehicle terminals responds with a refusal to execute, a conflicting risk decision instruction is identified.

8. The decentralized risk decision sending method according to claim 7, characterized in that, After the step of controlling each vehicle terminal within the monitoring range to execute its corresponding risk decision instruction when all vehicle terminals provide feedback agreeing to the execution, the method further includes: Obtain the execution result of each vehicle terminal executing its corresponding risk decision instruction; The execution result is sent to the cloud platform; Based on the execution results, the risk warning model of the cloud platform is iteratively updated to obtain the updated model parameters.

9. The decentralized risk decision transmission method according to claim 1, characterized in that, The calculation of the current risk value of each vehicle terminal within the monitoring range, based on the first and second road condition data of each roadside edge node, includes: The location coordinates, driving speed, driving acceleration, and distance between the vehicle terminal and each driving object within the monitoring range are determined based on the first road condition data and the second road condition data. The current risk value of each vehicle terminal is calculated based on its location coordinates, driving speed, driving acceleration, and the distance between the vehicle terminal and each driving object within the monitoring range.

10. A decentralized risk decision transmission system, characterized in that, include: A risk warning model is stored in the vehicle terminal. The vehicle terminal is configured to collect first road condition data and send the first road condition data to the roadside edge node corresponding to the vehicle terminal. Multiple roadside edge nodes are spaced apart along the edge of the road. Each roadside edge node stores a risk warning model and is configured to collect second road condition data based on all driving objects within the monitoring range of the roadside edge node. Based on the second road condition data and the first road condition data of each vehicle terminal within the monitoring range, calculate the current risk value of each vehicle terminal within the monitoring range; Based on the risk warning model, the predicted risk value of each vehicle terminal is calculated according to the current risk value of each vehicle terminal. If the predicted risk value of any of the vehicle terminals exceeds the risk value threshold, risk information is determined based on the first predicted risk value that exceeds the risk value threshold. The vehicle terminal or the roadside edge node is further configured to send a corresponding risk decision instruction to each vehicle terminal within the monitoring range based on the risk information, so that each vehicle terminal executes the corresponding risk decision instruction.

11. The decentralized risk decision sending system according to claim 10, characterized in that, The roadside edge node is also configured as follows: A local dynamic risk map corresponding to the roadside edge node is generated based on the current risk value of each vehicle terminal; The risk warning model calculates the predicted risk value for each vehicle terminal based on a preset number of local dynamic risk maps.

12. The decentralized risk decision sending system according to claim 11, characterized in that, Also includes: A cloud platform that stores a risk warning model and is communicatively connected to the roadside edge nodes and the vehicle terminals; The roadside edge node is also configured to generate a global dynamic risk map based on the local dynamic risk map corresponding to each roadside edge node at the same time, and upload the global dynamic risk map to the cloud platform. The cloud platform is configured to perform iterative updates on the risk warning model of the cloud platform through the global dynamic risk map to obtain updated model parameters, and send the updated model parameters to each of the roadside edge nodes and each of the vehicle terminals located within the respective monitoring range of each of the roadside edge nodes.

Images