Road traffic safety risk grading and vehicle speed limiting control method and system

Abstract

The invention discloses a road traffic safety risk grading and vehicle speed limit control method and system, and relates to the field of traffic control of road vehicles, and the method comprises the following steps: S1, carrying out the flexible grading of road traffic safety risks: S2, carrying out the vehicle speed limit control adjustment: (1) calculating a scene reference speed limit according to the grade of the road traffic safety risk and a road segment scene, a scene-vehicle type dual-adaptation speed limiting instruction is generated; (2) realizing pre-speed-limit control based on a risk pre-judgment result; (3) a vehicle-mounted warning system is linked while the speed limit is adjusted; and S3, establishing a full-link closed-loop optimization mechanism. According to the road traffic safety risk grading and vehicle speed limiting control method and system, the core defects of an existing road traffic safety risk grading and vehicle speed limiting control technology in the aspects of risk pre-judgment, scene adaptation, personalized control and closed-loop optimization are overcome, and a multi-dimensional optimization effect is formed; and the refined safety management and control capability and the traffic efficiency balance level in a complex traffic scene are integrally and remarkably improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of traffic control of road vehicles, in particular to a road traffic safety risk grading and vehicle speed limit control method and system. BACKGROUND

[0002] With the rapid development of intelligent transportation technology, road traffic safety risk grading and vehicle speed limit control has become a core means to improve traffic safety and optimize traffic efficiency. In the prior art, a dynamic speed limit control method is used to dynamically adjust the speed of vehicles on the highway, thereby reducing the risk of accidents and improving traffic safety. However, the prior art still has many key defects in practical application, and it is difficult to meet the fine safety control requirements in complex traffic scenarios. The specific deficiencies are as follows: Existing road safety risk assessment schemes mostly rely on real-time static data and passive mode of current state analysis, without establishing a risk trend prediction mechanism. The passive assessment scheme leads to delayed handling of sudden risks, and the secondary accident rate is higher than that of the prediction scheme, making the response lag behind the risk evolution. Secondly, the discrete grading logic with fixed threshold is generally used, without considering the differences in safety needs of different road sections and time periods. The grading standard is fixed, and the scene adaptability is poor. The existing system outputs a unified speed limit value, without considering the dual influence of vehicle performance differences and scene safety needs. Different vehicles under the same speed limit standard lead to a higher safety accident rate in special scenarios than the adaptive scheme. The existing technology only realizes a one-way control process of data collection-risk grading-speed limit issuance, without establishing a model iteration mechanism based on execution feedback. The model cannot be self-adaptively adjusted, leading to an increase in risk assessment error after long-term operation, a decrease in the matching degree of speed limit instructions and actual safety needs, and difficulty for the system to continuously adapt to changes in the traffic environment. SUMMARY

[0003] The present application relates to the field of traffic control of road vehicles, in particular to a road traffic safety risk grading and vehicle speed limit control method and system.

[0004] In a first aspect, the present application provides a road traffic safety risk grading and vehicle speed limit control method, comprising the following steps: S1, flexible grading of road traffic safety risk: ①Collect multi-source time-series data of roads, traffic, environment, and people and vehicles, and historical correlation data, construct a three-dimensional risk quantification model of "real-time state + historical law + trend prediction" based on a GRU time-series neural network, and take the three-dimensional risk quantification model as a basis for reference; ②Adjust the weight coefficient dynamically, introduce a scene correction factor, realize dynamic adaptive grading through "basic threshold + scene correction", and output real-time risk levels and future risk prediction results to classify road traffic safety risks; ③The parameters after grading are transmitted to a speed limit regulation unit; S2, vehicle speed limit control adjustment: ①Calculate a scene reference speed limit according to the road traffic safety risk level and the road scene, introduce a vehicle type correction coefficient in combination with a vehicle type safety attribute database, perform double adaptation mode, and then generate a scene-vehicle type double adaptation speed limit instruction; ②Realize pre-speed limit control based on the risk prediction result, and adjust the speed limit in a stepwise manner when the risk level changes; ③Link the vehicle-mounted warning system while adjusting the speed limit; S3, build a full-link closed-loop optimization mechanism, iteratively optimize the model parameters based on execution feedback data, and ensure long-term stable operation of the system.

[0005] In this embodiment, the multi-source time-series data collection of S1 includes basic static data and real-time dynamic data, wherein: The basic static data includes physical parameters (curvature, slope, number of lanes, and designed speed) of the road and attributes (highway / urban / rural, school / hospital / intersection) of the road section; The real-time dynamic data includes traffic flow time-series data (change trend of speed, distance, and vehicle density in the last 10 minutes), environmental time-series data (change of visibility, precipitation intensity, and road friction coefficient in the last 10 minutes), and people and vehicle state data (current driver fatigue, vehicle load, and braking performance); The historical correlation data includes accident occurrence rate, congestion frequency, and weather regularity data of the same road section and time period in the last 3 months.

[0006] In this embodiment, the multi-source time-series data collection of S1 constructs a three-dimensional model of time-series predictive risk quantification, and then performs dynamic adaptive grading: ①Set the basic grading interval based on the multi-source time-series data and historical correlation data: Low risk interval: 1-20; Lower risk interval: 21-40; Medium risk interval: 41-60; Higher risk interval: 61-80; High risk interval: 81-100; ②Introduce a scene correction factor, and adjust the threshold of grading dynamically in combination with the data of the scene factor; ③According to the results of dynamic adaptive grading, the current dynamic risk level and the risk prediction results in the next 10 minutes are output.

[0007] The introduction of the scene correction factor in the three-dimensional model of risk quantification is K; K includes a road section correction factor + a time period correction factor, and the dynamic adjustment of the grading threshold value follows the following formula: The construction of the three-dimensional risk quantification model of S1 follows the following formula: is the real-time risk value (calculated based on current dynamic data) is the historical associated risk value (calculated based on historical data in the same scene) is the risk prediction value in the next 5-10 minutes, obtained by training a GRU time series neural network, with a prediction accuracy of ≥85% is the dynamic weight, which is adaptively adjusted by training historical accident data.

[0008] In this embodiment, the vehicle speed limit control adjustment of S2 vehicle speed limit control adjustment realizes double adaptation speed limit, which specifically includes scene reference speed limit calculation and vehicle type adaptation adjustment. The scene reference speed limit calculation is to establish a reference speed limit model according to the risk level and the road scene, and the vehicle type adaptation adjustment is to establish a vehicle safety attribute database including braking distance, acceleration performance, and center of gravity height parameters, and introduce a vehicle type correction coefficient. Each vehicle type has a separate correction coefficient.

[0009] The scene reference speed limit calculation follows the following formula: is the road design speed is the total risk value is the minimum safe speed of the scene The vehicle type coefficient introduced in the vehicle type adaptation adjustment is: Small car: C=1.0, Medium passenger car / light truck: C=0.9, Heavy truck / dangerous goods transport vehicle: C=0.75, New energy vehicle (considering battery load): C=0.85, In this embodiment, after the S2 vehicle speed limit control adjustment, dynamic linkage and prediction control are finally performed: When the predicted risk is upgraded, the "pre-speed limit" is triggered to reduce the reference speed limit by 10% 3-5 minutes in advance to avoid sudden braking when the risk occurs; When the risk level suddenly changes from a lower risk to a high risk, a "step speed limit" is adopted, which is reduced by 8 km / h every 1 second, and the vehicle-mounted warning system is simultaneously linked to issue a warning; When the risk level decreases, a "slowly rising" strategy is adopted, which is increased by 5 km / h every 2 seconds to restore the vehicle speed and ensure the traffic efficiency; Special scene enhancement control: tunnel entrance, light mutation, additional 5km / h down, snow weather, road friction coefficient <0.4, additional 10km / h down.

[0010] In a second aspect, the present application provides a road traffic safety risk grading and vehicle speed limit control system, the control system comprising the following modules: data acquisition module, risk prediction grading module, speed limit control module, communication execution module, optimization iteration module.

[0011] In this embodiment, the data acquisition module: through radar, camera, information released by weather station, cooperated with vehicle terminal (GPS and OBD) and cloud database, multi-element time series data is collected, frequency ≥15Hz, data delay ≤30ms, main information including: road physical parameters, road section type, surrounding environment, time period attribute, special time node, real-time traffic flow, environmental state, vehicle dynamics and pedestrian / non-motor vehicle flow.

[0012] In this embodiment, the risk prediction grading module: through input collected information, GRU time series neural network is constructed, dynamic weight adjustment unit, grading threshold correction unit, then real-time risk level and future 10-minute risk prediction result are output, processing delay ≤80ms, the speed limit control module: based on scene reference speed limit model, vehicle type adaptation database, dynamic adjustment unit generates scene-vehicle type double adaptation speed limit instruction.

[0013] In this embodiment, the communication execution module: through 5G communication unit cooperated with cloud interaction interface, vehicle speed limit execution unit and data feedback unit are combined, realizing low-delay data interaction of roadside-vehicle-cloud, instruction transmission delay ≤40ms, receiving speed limit instruction controls vehicle speed, real-time feedback execution state (compliance rate, speed deviation).

[0014] In this embodiment, the optimization iteration module: feedback is given to data analysis unit and model parameter iteration unit, based on executed feedback data, accident rate, violation rate and traffic efficiency, risk model weight and speed limit parameter are iteratively optimized every 24 hours.

[0015] Compared with the prior art, the present application has the following advantages: The road traffic safety risk grading and vehicle speed limit control method and system solve the core defects of the existing road traffic safety risk grading and vehicle speed limit control technology in risk prediction, scene adaptation, individualized control and closed-loop optimization, form multi-dimensional optimization effect, and significantly improve the fine safety management and control ability and traffic efficiency balance level under complex traffic scenes.

[0016] 1. Furthermore, breaking through the passive evaluation mode that relies on real-time static data in existing technologies, we innovatively establish a risk trend prediction mechanism driven by "multi-source time-series data + neural network modeling". This achieves a core transformation from "passive response" to "proactive prediction". By establishing a risk trend prediction mechanism, we can perceive potential safety risks in advance, avoid the problem of delayed handling of sudden risks, effectively reduce the incidence of secondary accidents, and make risk response synchronized with or even ahead of the risk evolution process. This reduces the risk of chain reaction of accidents and realizes the essential leap of traffic safety management from "post-event remediation" to "pre-event prevention". This provides core technical support for refined safety management in complex traffic scenarios.

[0017] 2. Furthermore, a flexible risk classification system based on "scene perception + dynamic thresholds" is constructed, completely breaking the limitations of the traditional "one-size-fits-all" classification and significantly improving the accuracy of risk classification. At the road segment adaptation level, the system fully considers the inherent attributes of roads and the differences in safety sensitivity. At the time period adaptation level, the system accurately matches the traffic flow characteristics and safety requirements of different time periods, fully considering the differences in safety requirements of different road segments and different time periods, dynamically adjusting the classification thresholds to adapt to different road segments and time period scenarios, and constructing a flexible risk classification system. This achieves refined management with "one classification standard for one type of scenario," which can meet the high safety requirements of complex scenarios such as schools, intersections, and tunnels, while also taking into account the traffic efficiency of ordinary road segments. It provides core technical support for multi-dimensional and differentiated traffic safety management, solves the defects of existing classification standards being rigid and having poor scenario adaptability, and meets the refined management needs under complex traffic scenarios.

[0018] 3. Furthermore, at the scenario adaptation level, the system uses risk classification results as the core, combined with the inherent attributes of road segments and real-time environmental characteristics, to construct a dynamic scenario benchmark speed limit model. At the vehicle type adaptation level, the system has built a safety attribute database covering 6 major categories and 20 subcategories of vehicle types, corely recording key parameters such as driving distance, center of gravity height, acceleration performance, and rated load. The core advantage of dual-adaptive speed limit control lies in the "unity of personalized safety and optimal efficiency": for high-risk combination scenarios such as "school road sections + morning rush hour + heavy trucks", by accurately matching vehicle performance differences with scenario safety requirements, this mechanism avoids the problem of increased accident rates in special scenarios caused by uniform standards, and also eliminates the loss of traffic efficiency caused by excessive speed limits. It truly achieves the goal of realizing dual-adaptive speed limit control, ensuring personalized safety, and taking into account the refined management objectives of vehicle performance differences and scenario safety requirements. It provides core technical support for balancing road traffic safety and traffic efficiency in mixed traffic flow and complex scenarios, avoiding the problem of increased accident rates in special scenarios caused by uniform standards, and ensuring traffic safety while taking into account traffic efficiency.

[0019] 4. Furthermore, through a closed-loop control process encompassing "data acquisition - risk grading - speed limit adjustment - execution feedback - iterative optimization," and a continuous cycle of "acquisition - decision-making - execution - feedback - optimization," the system possesses self-evolution capabilities. It can dynamically adapt to changes in the traffic environment based on actual operational data, maintaining stable control accuracy and effectiveness over the long term. This closed-loop control process of data acquisition, risk grading, speed limit adjustment, and execution feedback iteration allows the system to continuously optimize the risk assessment model based on actual execution feedback. This avoids problems such as increased assessment errors and decreased matching between speed limit instructions and actual needs after long-term operation, ensuring the system can adapt to changes in the traffic environment and maintain stable control effectiveness in the long term. Attached Figure Description

[0020] Figure 1 This is a schematic diagram of the road traffic safety risk classification and vehicle speed limit control method of the present invention; Figure 2 This is a schematic diagram of the road traffic safety risk classification and vehicle speed limit control system of the present invention; Figure 3 This is a schematic diagram of multi-source time-series data acquisition according to the present invention; Figure 4 This is a schematic diagram of the three-dimensional model construction for risk quantification in this invention; Figure 5 This is a schematic diagram illustrating the dual-adaptive speed limit implementation of the vehicle speed limit control adjustment according to the present invention. Figure 6 This is a schematic diagram illustrating the dynamic linkage and predictive control of vehicle speed limit control adjustment according to the present invention. Detailed Implementation

[0021] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0022] This application provides a method and system for road traffic safety risk classification and vehicle speed limit control. Its core is flexible classification of road traffic safety risks: multi-source time-series data and historical correlation data of roads, traffic, environment, and vehicles are collected. A three-dimensional risk quantification model based on a GRU time-series neural network, incorporating real-time status, historical patterns, and trend prediction, is constructed as the basic reference. Weight coefficients are then dynamically adjusted, and a scenario correction factor is introduced. Dynamic adaptive classification is achieved through a combination of a basic threshold and scenario correction, outputting real-time risk levels and future risk predictions for road traffic safety risk classification. The classified parameters are transmitted to the speed limit control unit. Vehicle speed limit control adjustment: a scenario baseline speed limit is calculated based on the road traffic safety risk level and road segment scenario. A vehicle model correction coefficient is introduced using a vehicle model safety attribute database for a dual-adaptation mode, generating a scenario-vehicle dual-adaptation speed limit command. Pre-speed limit control is implemented based on the risk prediction results, employing a stepped speed limit adjustment when the risk level changes. The vehicle warning system is linked during speed limit adjustment. A full-link closed-loop optimization mechanism is established, iteratively optimizing model parameters based on execution feedback data to ensure long-term stable system operation.

[0023] Example 1: To better understand the above technical solution, the following will provide a detailed description of the technical solution in conjunction with the accompanying drawings and specific implementation methods. (Refer to...) Figure 1 As shown in the diagram, this is a diagram of road traffic safety risk classification and vehicle speed limit control methods. The road traffic safety risk classification and vehicle speed limit control methods include the following steps: In step S1, road traffic safety risks are flexibly classified: multi-source time-series data and historical correlation data of roads, traffic, environment, and vehicles and pedestrians are collected. A three-dimensional risk quantification model of "real-time status + historical pattern + trend prediction" is constructed based on GRU time-series neural network. The three-dimensional risk quantification model is used as the basic reference, and the weight coefficients are dynamically adjusted. A scene correction factor is introduced. Dynamic adaptive classification is achieved through "basic threshold + scene correction". The real-time risk level and future risk prediction results are output. The road traffic safety risks are classified, and the classified parameters are transmitted to the speed limit control unit through data transmission.

[0024] refer to Figures 3-4 As shown, Figure 3 It is a multi-source time-series data acquisition diagram. Figure 4This is a three-dimensional risk quantification model construction diagram. In its specific implementation, multi-source time-series data collection includes basic static data and real-time dynamic data. Basic static data includes the physical parameters of the road (curvature, slope, number of lanes, design speed) and the attributes of the road segment (highway / city / rural, school / hospital / intersection). Real-time dynamic data includes traffic flow time-series data (the trend of vehicle speed, distance between vehicles, and vehicle density changes in the past 10 minutes), environmental time-series data (the change of visibility, precipitation intensity, and road surface friction coefficient in the past 10 minutes), and human and vehicle status data (current driver fatigue, vehicle load, and braking performance). Historical correlation data includes the accident rate, congestion frequency, and weather pattern data of the same road segment and time period in the past 3 months.

[0025] It should be noted that basic static data is the underlying reference framework for risk assessment, reflecting the inherent safety characteristics of roads and road sections. It does not change dynamically over time and only needs to be collected initially and calibrated periodically. Real-time dynamic data is the "core sensitive factor" of risk evolution. It is necessary to capture the changing trend within the past 10 minutes to provide time-series characteristic support for risk prediction. The data has time-sensitivity requirements. Historical correlation data is the "pattern support" for risk prediction. By statistically analyzing data from the same scenario over the past 3 months, the correlation patterns between time period, weather and risk are explored to provide a basis for dynamic weight adjustment.

[0026] In addition, in specific implementation, the data collection frequency needs to be designed according to the dynamic characteristics of the data to avoid redundancy due to over-collection or feature loss due to under-collection. The basic static data collection frequency is initial full collection + quarterly calibration. The inherent attributes of the data are stable, and quarterly calibration can correct parameter changes after road maintenance (such as fine-tuning of curvature after pavement repair). The traffic flow time series data collection frequency is 10Hz (collected once every 0.1 seconds), which needs to capture subtle changes in vehicle speed and distance within the past 10 minutes (such as a sudden drop in vehicle speed due to congestion). 10Hz can meet the requirements for time series feature extraction and environmental time series data. The data collection frequency is 5Hz (once every 0.2 seconds), mainly for visibility and precipitation intensity changes that are relatively gradual compared to traffic flow. 5Hz can balance timeliness and sensor power consumption. The data collection frequency for human and vehicle status is 10Hz for driver fatigue and 5Hz for vehicle status. Fatigue needs to be judged in real time through facial features (10Hz ensures no missed judgments). The 5Hz frequency can capture dynamic changes in vehicle load / braking performance. The collection frequency for historical correlation data is statistically analyzed daily, summarized weekly, and updated monthly. Accident rate and congestion frequency need to be statistically analyzed over a long period of time. Daily statistics ensure details, and monthly updates ensure regularity and timeliness.

[0027] In practical implementation, the communication levels and processes for flexibly classifying road traffic safety risks are as follows: Perception layer communication: Roadside sensors (radar, camera, weather station) and roadside edge computing units (MEC) are directly connected via Ethernet with a transmission delay of ≤10ms, enabling local data preprocessing (noise reduction, format unification). Vehicle-to-infrastructure (V2I) communication: The roadside MEC and the vehicle terminal communicate via 5G-V2X (PC5 direct connection + Uu interface) to transmit real-time dynamic data (such as traffic flow and environmental status), with a command and data interaction latency of ≤30ms; Vehicle-to-cloud collaborative communication: The vehicle terminal communicates with the cloud database through the 5G Uu interface to upload vehicle and human status data and receive historical related data, with a transmission latency of ≤50ms; Cloud-edge collaborative communication: The cloud database and roadside MEC synchronize basic static data and historical related data through a dedicated network, and perform a full update every morning at midnight, with incremental data synchronized in real time.

[0028] Data synchronization mechanism: The strategy of "timestamp alignment + data verification" is adopted: all acquisition devices are synchronized with the same time (GPS + Beidou dual-mode time synchronization, error ≤1ms) to ensure the time consistency of multi-source data at the same time; Key data (such as road surface friction coefficient and driver fatigue signals) are transmitted with priority, occupying the priority channel of 5G-V2X communication bandwidth to avoid congestion and delays.

[0029] Meanwhile, multi-source time-series data collection constructs a three-dimensional model for predictive risk quantification based on time-series data, and then performs dynamic adaptive grading: ① Based on multi-source time-series data and historical correlation data, basic grading intervals are set: low-risk interval: 1-20; lower-risk interval: 21-40; medium-risk interval: 41-60; higher-risk interval: 61-80; high-risk interval: 81-100; ② Introduce a scene correction factor, and dynamically adjust the grading threshold in real time based on the scene factor data: The scenario correction factor introduced in the three-dimensional model for risk quantification is K; K includes a road segment correction factor and a time period correction factor. The dynamic adjustment of the grading threshold follows the following formula: Among them, the scene correction factor enables dynamic adaptation of the "tiered threshold". Existing technologies use fixed threshold levels (e.g., 80 points is considered high risk regardless of the scenario), resulting in poor scenario adaptability (e.g., 80 points may still be safe on highways, but 60 points is already considered high risk on school zones). This solution dynamically adjusts the threshold through a scenario correction factor K. The core logic is as follows: Design logic of scene correction factor: (Road Segment Correction) Focuses on "Inherent Attributes of Road Segments": Highly sensitive road segments (such as 500m around schools) require stricter risk assessment, so negative values ​​are used (lowering the classification threshold) to allow risks to be identified earlier; Lowly sensitive road segments (such as straight sections of highways) need to avoid excessive speed limits, so positive values ​​are used (increasing the classification threshold) to balance safety and efficiency; (Time Period Correction) Focuses on “Dynamic Characteristics of Time Periods”: High-risk periods (such as 23:00-0:00 at night) have poor lighting and drivers are prone to fatigue, so negative values ​​are used (lowering the threshold); low-risk periods (such as off-peak hours in the early morning) have sparse traffic, so positive values ​​are used (increasing the threshold).

[0030] ③ Based on the results of dynamic adaptive grading, output the current dynamic risk level and the risk prediction results for the next 10 minutes.

[0031] The construction of the three-dimensional risk quantification model follows the formula below: This is the real-time risk value (calculated based on current dynamic data). Historical associated risk values ​​(calculated based on historical data from the same scenario) The risk prediction values ​​for the next 5-10 minutes are obtained through training a GRU temporal neural network, with a prediction accuracy of ≥85%. The weights are dynamic and are adaptively adjusted by training with historical accident data.

[0032] Among them, represents real-time risk values, such as sudden heavy rain or severe driver fatigue, to ensure real-time risk response; For historical correlation risk values, anchor "scenario-based risk", such as the historical accident rate at a certain intersection during the morning rush hour being 3 times that during the off-peak hour, to avoid random deviations in real-time data; Predicting "future evolution risks," such as the GRU model predicting that traffic density will increase from 15 vehicles / 100 meters to 28 vehicles / 100 meters in 5 minutes, allows for early prevention and control.

[0033] The weights are not fixed values, but are trained in real time using historical accident data (such as LSTM / GRU neural networks). The model learns "which risk factors are more closely associated with the accident rate in different scenarios" and dynamically adjusts the weights.

[0034] The total risk value ranges from 0 to 100 points, with continuous and accurate values ​​to avoid the loss of precision caused by discretization and quantification. It is trained using a GRU time-series neural network, relying on nearly 10 minutes of time-series data (traffic flow, environmental change trends), with a prediction accuracy of ≥85%, ensuring the credibility of the predicted risk.

[0035] In step S2, vehicle speed limit control adjustment: the baseline speed limit is calculated based on the road traffic safety risk level and road segment scenario, and then a vehicle model correction coefficient is introduced by combining the vehicle model safety attribute database to perform a dual adaptation mode, thereby generating a scenario-vehicle model dual adaptation speed limit command; pre-speed limit control is implemented based on risk prediction results, and a step-by-step speed limit adjustment is adopted when the risk level changes; the vehicle warning system is linked at the same time as the speed limit adjustment.

[0036] refer to Figure 5 and Figure 6 As shown,Figure 5 It is a vehicle speed limit control adjustment that achieves dual-adaptive speed limit diagrams. Figure 6 After adjusting the vehicle speed limit control, dynamic linkage and predictive control are implemented. Specifically, the vehicle speed limit control adjustment achieves dual-adaptive speed limits, including scenario-based baseline speed limit calculation and vehicle model adaptation adjustment. Scenario-based baseline speed limit calculation establishes a baseline speed limit model based on risk level and road segment scenario. Vehicle model adaptation adjustment establishes a vehicle safety attribute database including braking distance, acceleration performance, and center of gravity height parameters, introducing a vehicle model correction coefficient. Each vehicle model has a separate correction coefficient. After adjusting the vehicle speed limit control, dynamic linkage and predictive control are finally implemented; risk prediction is performed. When upgrading, the "pre-speed limit" will be triggered 3-5 minutes in advance, reducing the base speed limit by 10% to avoid sudden braking in case of sudden risks; when the risk level changes abruptly from low risk to high risk, a "stepped speed limit" will be adopted, reducing the speed limit by 8 km / h every 1 second, and the on-board warning system will be activated simultaneously; when the risk level decreases, a "slow recovery" strategy will be adopted, increasing the speed by 5 km / h every 2 seconds to restore the vehicle speed and ensure traffic efficiency; special scenario enhanced control: at tunnel entrances, when the light changes suddenly, the speed limit will be reduced by an additional 5 km / h; in rainy or snowy weather, when the road surface friction coefficient is <0.4, the speed limit will be reduced by an additional 10 km / h.

[0037] The scenario baseline speed limit calculation follows the formula below: Design speed for roads Total risk value Minimum safe speed for the scenario The vehicle model coefficient introduced in the vehicle model adaptation adjustment is: Small cars: C=1.0, Medium-sized passenger vehicles / light-duty trucks: C=0.9, Heavy trucks / hazardous materials transport vehicles: C=0.75, New energy vehicles (considering battery load): C=0.85, It should be noted that dual-adaptive speed limit is the core of speed limit control. By using the logic of "calculating the baseline first and then adapting", it ensures that the speed limit value meets both road risk requirements and vehicle safety performance. The initial dual-adaptation speed limit is not fixed and needs to be dynamically adjusted according to the "predicted changes", "sudden changes" and "decreases" of the risk level to avoid sudden braking or excessive speed limit, while linking with the vehicle warning system to improve driver compliance; Pre-limit speed control: Prevents risk escalation in advance. The three-dimensional risk quantification model predicts that "the risk level will escalate in the next 5-10 minutes" and triggers the pre-limit speed 3-5 minutes in advance, allowing drivers to adapt to the speed change in advance and avoid sudden braking or rear-end collisions caused by a sharp reduction in speed when a risk occurs. Step-by-step speed limit reduction: In response to sudden changes in risk level, such as a sudden upgrade in risk level (e.g., from lower risk to higher risk), often due to emergencies such as sudden accidents or short-term heavy rainfall, the speed limit is reduced by 8 km / h every second until the target speed limit is reached.

[0038] Slow recovery strategy: Restore efficiency as risk decreases. As the risk level decreases (e.g., from high risk to medium risk, or from medium risk to low risk) and remains stable for more than 30 seconds (avoiding frequent adjustments), increase the speed limit by 5 km / h every 2 seconds until it returns to the dual-adaptive speed limit value corresponding to the current risk level.

[0039] In step S3, a full-link closed-loop optimization mechanism is established, and the model parameters are iteratively optimized based on the execution feedback data to ensure the long-term stable operation of the system.

[0040] In practice, the core of the end-to-end closed-loop optimization mechanism is to build a continuous iterative cycle of "data collection → feedback analysis → parameter optimization → verification and deployment → re-collection". Driven by the execution feedback data, it accurately optimizes the key parameters of the risk quantification model and speed limit control strategy, ensuring that the system adapts to changes in the traffic environment in the long term, such as road traffic characteristics, vehicle technology upgrades, and weather patterns.

[0041] Example 2: This invention provides a road traffic safety risk classification and vehicle speed limit control system, with reference to... Figure 2 As shown in the figure, this is a road traffic safety risk classification and vehicle speed limit control system. The control system includes the following modules: data acquisition module, risk prediction and classification module, speed limit control module, communication execution module, and optimization and iteration module. Data acquisition module: Collects multi-dimensional time-series data through information released by radar, cameras, weather stations, and in conjunction with the onboard terminal (GPS and OBD) and cloud database. The data is collected at a frequency of ≥15Hz and with a data latency of ≤30ms. The main information includes: road physical parameters, road segment type, surrounding environment, time period attributes, special time nodes, real-time traffic flow, environmental status, vehicle dynamics, and pedestrian / non-motorized vehicle flow. In practice, a "three-in-one" collaborative acquisition mode is adopted to ensure that no data dimensions are missed and that there is no transmission delay: Roadside perception layer: Equipment such as radar, cameras, and weather stations are responsible for collecting roadside data such as static road attributes, real-time traffic flow, and environmental conditions; Vehicle data acquisition layer: GPS and OBD vehicle terminals are responsible for collecting vehicle dynamics, driving trajectory and other vehicle terminal data; Cloud support layer: cloud database, responsible for storing and providing preset / historical data such as basic road attributes and historical correlation data (such as time period attributes and special time node patterns).

[0042] Basic road attribute data (static + semi-static) serves as the underlying reference for risk assessment. The data is collected primarily through initial collection and periodic calibration, with a collection frequency of "full initial collection + monthly calibration" to ensure that the data is consistent with the actual road condition.

[0043] Time attribute data (semi-static + dynamic) This type of data reflects the impact of time period characteristics on risk. It combines preset rules with real-time judgment and is collected at a frequency of 1Hz (updated once every 1 second). Real-time dynamic data (high-frequency dynamics) is a core sensitive factor in risk evolution. It is collected at a high frequency of ≥15Hz to ensure that subtle changing trends are captured.

[0044] The data acquisition module provides three core supports for subsequent stages through comprehensive, high-frequency, and low-latency multi-source time-series data acquisition: Provides complete input for risk quantification models: covering three types of risk factors: spatial, temporal, and dynamic, ensuring the comprehensiveness and accuracy of risk value calculation; Provide personalized basis for speed limit control strategy: Support vehicle model adaptation adjustment through data such as vehicle type, load, and braking performance; Support scenario-based speed limit calculation through data such as road segment type and pedestrian flow. Provide feedback data sources for the closed-loop optimization mechanism: collect data such as vehicle dynamics and accidents / hazards after the speed limit is implemented, and provide a basis for iterative optimization of model parameters.

[0045] Risk prediction and classification module: Constructs a GRU temporal neural network, dynamic weight adjustment unit, and classification threshold correction unit by inputting collected information, and then outputs the real-time risk level and risk prediction results for the next 10 minutes, with a processing latency of ≤80ms. Speed ​​limit control module: Generates scenario-vehicle dual-adaptive speed limit instructions based on the scenario benchmark speed limit model, vehicle model adaptation database, and dynamic adjustment unit. In practical implementation, the risk prediction and grading module is the "decision brain" of the system, responsible for transforming multi-source collected data into accurate risk levels and prediction results; the speed limit control module is the "execution instruction generation center," outputting personalized speed limit instructions based on risk decisions. The two work together to achieve the core logic of "accurate risk prediction - accurate speed limit adaptation," while strictly meeting the technical requirements of low latency and high adaptability. The module takes "multi-source time-series data" as input and outputs the real-time risk level (level 5) + the risk prediction result (level + risk evolution trend) within ≤80ms through the collaborative operation of GRU time-series neural network, dynamic weight adjustment unit, and graded threshold correction unit. In practice, the risk prediction and classification module outputs the real-time risk level and prediction result every 0.1 seconds. The speed limit control module receives and updates the speed limit command synchronously to ensure that the command matches the risk status in real time. The execution feedback data (such as vehicle speed deviation and compliance rate) of the speed limit control module is transmitted back to the risk prediction and classification module for iterative optimization of the GRU model and dynamic weights, forming a collaborative closed loop of "risk decision-making - speed limit execution - feedback optimization". When any module malfunctions, it automatically switches to the backup mode to ensure continuous system operation.

[0046] Communication execution module: Through the 5G communication unit and cloud interaction interface, together with the vehicle speed limit execution unit and data feedback unit, it realizes low-latency data interaction between roadside, vehicle and cloud, with command transmission latency ≤40ms. It receives speed limit commands to control vehicle speed and provides real-time feedback on execution status (compliance rate, speed deviation). In practical implementation, the communication execution module is the system's "transmission hub + execution terminal". The core uses a collaborative architecture of "5G communication unit + cloud interaction interface + vehicle execution / feedback unit" to open up the data interaction channel between the roadside, vehicle and cloud. It can not only achieve low-latency issuance of speed limit commands, but also complete real-time feedback of vehicle execution status. It is a key bridge connecting the decision-making layer (risk prediction and classification, speed limit control module) and the execution layer (vehicle) to ensure the closed loop of the entire "decision-transmission-execution-feedback" link. Its core positioning: Data transmission: As a "data relay station" for roadside (risk prediction and classification / speed limit control module), vehicle-mounted, and cloud, it realizes two-way low-latency communication; Command execution: Receives dual-adaptive speed limit commands from the speed limit control module and precisely controls the vehicle speed through the on-board unit; Status feedback: Real-time collection of status data such as compliance rate and speed deviation after vehicle speed limit enforcement, and reverse transmission to the cloud and closed-loop optimization module to support parameter iteration; The module adopts a dual-core architecture of "communication link + execution feedback". The communication link is responsible for low-latency data transmission, while the execution feedback is responsible for rate-limited landing and status acquisition.

[0047] The optimization iteration module provides feedback to the data analysis unit and the model parameter iteration unit. Based on the data feedback, such as accident rate, violation rate, and traffic efficiency, it iteratively optimizes the risk model weights and speed limit parameters every 24 hours. In practice, the optimization and iteration module is the "self-evolution hub" of the system. The core of it works in collaboration with the "data analysis unit + model parameter iteration unit" to deeply mine the value of execution feedback data (accident rate, violation rate, traffic efficiency) every 24 hours. It then iterates and optimizes the risk model weights and speed limit control parameters in a targeted manner to continuously improve the system's adaptability to the traffic environment and ensure a balance between accuracy, safety and traffic efficiency in long-term operation. Its core positioning: Data mining: Analyzing the deviation patterns and optimization space in execution feedback data, and locating the unreasonable aspects of risk models and speed limit parameters; Parameter iteration: Based on the analysis results, accurately adjust core parameters such as the dynamic weights of the risk quantification model and the vehicle type correction coefficient of the speed limit strategy; Performance upgrade: Through continuous iteration, the system can maintain high adaptability when road characteristics, traffic flow structure and weather patterns change, avoiding performance degradation after long-term operation; The advantages are: strong dynamic adaptability: through 24-hour periodic iteration, it adapts in real time to changes in the traffic environment (such as road traffic flow structure, weather patterns, and vehicle technology upgrades), solving the long-term performance degradation problem caused by the "fixed parameters" of traditional systems; High optimization accuracy: Problems are located based on multi-dimensional feedback data, and parameters are adjusted in a targeted manner to avoid blind optimization. After optimization, both safety and efficiency indicators are significantly improved. Balancing safety and efficiency: The dual-objective optimization algorithm ensures that both are taken into account, improving safety without sacrificing traffic efficiency, which meets the actual traffic management needs; Highly practical: The iteration process is standardized and automated, requiring no manual intervention, and has a sound verification and rollback mechanism, making the risks controllable and suitable for large-scale long-term operation.

[0048] Example 3: Risk Classification and Speed ​​Limit Control for Heavy Rain on Highways This embodiment is applied to a highway in heavy rain scenario. Addressing the shortcomings of existing technologies such as passive assessment, fixed classification, and uniform speed limits, the specific implementation is as follows: Risk prediction data collection: Data is collected in real time through millimeter-wave radar, video surveillance equipment, weather stations and vehicle OBD (on-board diagnostic system) deployed along the road. This includes real-time rainfall, road surface water depth, visibility, real-time vehicle speed, vehicle distance, vehicle braking frequency, historical rainstorm weather accident data for this road section (such as rainfall at the time of the accident and vehicle speed threshold), and rainfall trend prediction data for the next hour.

[0049] Risk trend prediction and dynamic classification: A risk prediction model is built based on the collected data. The correlation between rainfall, water depth, visibility and accident probability is obtained by training with historical accident data. Combined with future rainfall trend prediction, the risk evolution trend of the road section is predicted 15 minutes in advance. At the same time, fixed threshold classification is abandoned. According to the difference in traffic density during the morning peak (7:00-9:00), off-peak (10:00-16:00), and evening peak (17:00-19:00), a dynamic classification threshold is set. When the traffic density is high during the morning peak, a rainfall of 5 mm / h is judged as "medium risk". When the traffic density is low during the off-peak, a rainfall of 8 mm / h is judged as "medium risk".

[0050] Dual-adaptive speed limit command generation: Speed ​​limit commands are generated by combining vehicle performance differences (vehicle type obtained through vehicle-to-everything (V2X) such as small cars, heavy trucks, and new energy vehicles) with scenario requirements: small cars are limited to 80 km / h under "medium risk" conditions, heavy trucks are limited to 60 km / h due to their long braking distance, and new energy vehicles are limited to 70 km / h due to battery stability requirements in rainy weather. At the same time, the speed limit commands are accurately sent to the on-board unit (OBU) of the corresponding vehicle through the roadside unit (RSU) and displayed synchronously on the road electronic screen.

[0051] Closed-loop optimization and iteration: Real-time collection of vehicle speed limit command execution status (such as whether there is speeding or sudden braking) and actual accident occurrence on road sections. If it is found that there are still many sudden braking phenomena when the speed limit of heavy trucks is 60km / h under "medium risk", the "medium risk" speed limit threshold of this vehicle type is lowered to 55km / h, and the correlation parameter of "heavy truck-rainy braking distance" in the risk prediction model is updated to ensure that subsequent speed limit commands are more in line with actual safety needs.

[0052] Example 4: Risk Classification and Speed ​​Limit Control for Urban Expressways During Morning and Evening Peak Hours This embodiment is applied to the morning and evening rush hour scenarios on urban expressways (design speed 80km / h), focusing on the problems of poor scenario adaptability and lack of closed-loop optimization in existing technologies. The specific implementation is as follows: Multi-dimensional data collection: Traffic density, vehicle cutting frequency, and lane occupancy rate are collected through traffic detectors and video analysis equipment at expressway entrances and exits. This data is combined with historical morning and evening peak accident data (such as accident-prone road sections and peak times) and real-time traffic event information (such as temporary construction and vehicle breakdowns). At the same time, estimated travel time and route selection data are obtained through navigation platforms to help predict the risks caused by traffic congestion.

[0053] Dynamic grading and risk prediction: Grading logic is set according to the different characteristics of morning and evening peak hours. The morning peak (7:30-8:30) is mainly composed of commuter traffic with a single vehicle type. When the traffic density reaches 20 vehicles / km, it is judged as "high risk" (prone to rear-end collisions). The evening peak (18:00-19:00) includes commuter vehicles and freight vehicles, with a complex traffic composition. When the traffic density reaches 15 vehicles / km, it is judged as "high risk". At the same time, based on the trend of traffic density changes, it is predicted that the traffic density of a certain road section will increase from 18 vehicles / km to 22 vehicles / km in 10 minutes, triggering a "high risk" warning in advance.

[0054] Personalized speed limits and controls: Speed ​​limits are generated based on vehicle usage: commuter cars are limited to 60 km / h for "high-risk" situations, delivery trucks are limited to 55 km / h due to frequent lane changes, and emergency vehicles such as ambulances and fire trucks have priority routes generated based on real-time traffic conditions. At the same time, a "green wave" is created for emergency vehicles through traffic signal linkage. In addition, the speed limit is further reduced by 5 km / h on accident-prone curves to further improve safety.

[0055] Model Iteration and Optimization: If delivery vehicles are found to exceed speed limits by 15% (higher than the 8% of other vehicle types) in "high-risk" sections during the evening rush hour for a consecutive week, the reason is analyzed: due to the timeliness requirements of delivery, this vehicle type is highly sensitive to speed limits. Therefore, the strategy is adjusted: under "high-risk" conditions, alternative detour routes (auxiliary roads with less traffic) are planned for delivery vehicles, and their speed limit is slightly adjusted to 58km / h, which reduces the risk of speeding while taking into account delivery efficiency; and the correlation data of "delivery vehicle-traffic density-speeding probability" in the risk classification model is updated to improve the model's adaptability.

[0056] Example 5: Risk Classification and Speed ​​Limit Control for Icy and Snowy Weather on Mountain Roads in Winter This embodiment is applied to a winter icy and snowy weather scenario on mountain roads (with many sharp bends and slopes, designed speed 60km / h), overcoming the shortcomings of existing technologies that rely on passive response and uniform speed limits. The specific implementation is as follows: Special scenario data collection: Road surface temperature sensors and ice and snow detectors are used to collect road surface ice thickness and freezing point temperature. Drone aerial photography is used to obtain snow cover on sharp bends and long downhill sections. This is combined with historical winter accident data (such as truck rollaway accidents caused by ice and snow on long downhill sections and skidding accidents on sharp bends). At the same time, vehicle sensors are used to obtain vehicle tire type (ordinary tires, snow tires) and braking performance parameters.

[0057] Risk prediction and dynamic classification: Based on the correlation between road surface icing thickness and temperature, it is predicted that the road surface icing thickness will increase from 3mm to 5mm in the next 2 hours (with the freezing point temperature remaining below -2℃), and the risk level of the road section is raised from "low risk" to "medium risk" in advance; at the same time, classification thresholds are set according to the characteristics of the road section - a 2mm icing thickness on a long downhill section (slope ≥ 5%) is judged as "high risk", and a 4mm icing thickness on a straight section is judged as "high risk".

[0058] Vehicle-specific speed limit control: Speed ​​limit instructions are generated based on vehicle tire type and road characteristics: small cars equipped with snow tires are limited to 40km / h on "medium-risk" sharp curves, while small cars equipped with regular tires are limited to 30km / h; heavy trucks on long downhill "high-risk" sections are limited to 35km / h and are simultaneously given the instruction to "activate retarder"; and drivers are reminded of icy road conditions through roadside warning lights and voice broadcasts.

[0059] Closed-loop feedback optimization: If it is found that vehicles equipped with ordinary tires still have the risk of sideslip when the speed limit is 30km / h on "medium-risk" sharp curves, the speed limit for that model on "medium-risk" sharp curves will be lowered to 25km / h. At the same time, the weight of "ordinary tire-ice and snow road friction coefficient" will be strengthened in the risk prediction model. The model will be continuously optimized based on new operating data to ensure that the speed limit instruction is highly matched with the mountainous ice and snow road conditions and vehicle performance.

[0060] The contents not described in detail in this specification are existing technologies known to those skilled in the art.

[0061] Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments or make equivalent substitutions for some of the technical features. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for road traffic safety risk ranking and vehicle speed limit control, characterized in that: Comprise the following steps: S1, flexible classification of road traffic safety risk: ①Collect multi-source time series data and historical correlation data of road, traffic, environment, and man-machine, and build a three-dimensional risk quantification model based on GRU time series neural network for real-time state + historical law + trend prediction, and take the three-dimensional risk quantification model as the basis for reference; ②Adjust the weight coefficient dynamically, introduce a scene correction factor, and realize dynamic adaptive classification through "basic threshold + scene correction" to output real-time risk level and future risk prediction results, and classify road traffic safety risk; ③The parameters after classification are transmitted to the speed limit control unit; S2, vehicle speed limit control adjustment: ①Calculate the scene reference speed according to the road traffic safety risk level and the road scene, and then introduce the vehicle type correction coefficient combined with the vehicle type safety attribute database to realize double adaptation mode, and then generate scene-vehicle type double adaptation speed limit instruction; ②Realize pre-speed control based on risk prediction results, and adjust the speed in steps when the risk level changes; ③Link the vehicle warning system while adjusting the speed; S3, build a full-link closed-loop optimization mechanism, and iteratively optimize the model parameters based on execution feedback data to ensure long-term stable operation of the system.

2. The road traffic safety risk ranking and vehicle speed limit control method of claim 1, wherein: The S1, multi-source time series data collection includes basic static data and real-time dynamic data, wherein: Basic static data: including physical parameters of roads and attributes of road sections; Real-time dynamic data: including traffic flow time series data, environmental time series data, and man-machine state data; Historical correlation data: accident occurrence rate, congestion frequency, and weather regularity data of the same road section and time period in the past 3 months.

3. The road traffic safety risk ranking and vehicle speed limit control method of claim 2, wherein: The S1, multi-source time series data collection builds a three-dimensional model for time series predictive risk quantification, and then performs dynamic adaptive classification: ①Set the basic classification interval based on multi-source time series data and historical correlation data: Low risk interval: 1-20; Lower risk interval: 21-40; Medium risk interval: 41-60; Higher risk interval: 61-80; High risk interval: 81-100; ②Introduce a scene correction factor and combine the scene factor data to dynamically adjust the threshold of classification; ③Classify and output according to the results of dynamic adaptive classification, output the current dynamic risk level and the risk prediction results in the next 10 minutes.

4. The road traffic safety risk classification and vehicle speed limit control method of claim 3, wherein: The vehicle speed limit control adjustment of S2 vehicle speed limit control adjustment realizes double adaptation speed limit, which includes scene reference speed calculation and vehicle type adaptation adjustment. The scene reference speed calculation is to establish a reference speed model according to the risk level and the road scene, and the vehicle type adaptation adjustment is to establish a vehicle type safety attribute database including braking distance, acceleration performance, and gravity height parameters, and introduce a vehicle type correction coefficient. Each vehicle type has a separate correction coefficient.

5. The road traffic safety risk ranking and vehicle speed limit control method of claim 4, wherein: After the S2, vehicle speed limit control adjustment, dynamic linkage and prediction control are finally performed: When the predicted risk is upgraded, the reference speed is reduced by 10% 3-5 minutes in advance to avoid sudden braking in case of risk; When the risk level suddenly changes from low risk to high risk, use "step speed" to reduce the speed by 8 km / h every second, and link the vehicle warning system to issue a warning. When the risk level decreases, the "slow recovery" strategy is adopted, and the vehicle speed is increased by 5 km / h every 2 seconds to ensure traffic efficiency. Special scene enhancement control: tunnel entrance, light mutation, additional 5 km / h down, snow weather, road friction coefficient < 0.4, additional 10 km / h down.

6. A road traffic safety risk classification and vehicle speed limit control system for performing a road traffic safety risk classification and vehicle speed limit control method according to any one of claims 1 to 5, characterized by: The control system comprises the following modules: data acquisition module, risk prediction grading module, speed limit control module, communication execution module, and optimization iteration module.

7. A road traffic safety risk ranking and vehicle speed limiting control system as claimed in claim 6, characterised in that: The data acquisition module collects multi-element time series data through radar, camera, weather station information, vehicle-mounted terminal, and cloud database, with a frequency of ≥15 Hz and a data delay of ≤30 ms. The main information includes road physical parameters, road type, surrounding environment, time period attributes, special time nodes, real-time traffic flow, environmental status, vehicle dynamics, and pedestrian / non-motor vehicle flow.

8. A road traffic safety risk ranking and vehicle speed limiting control system according to claim 7, characterised in that: The risk prediction grading module constructs a GRU time series neural network, a dynamic weight adjustment unit, and a grading threshold correction unit based on the input collected information, and then outputs real-time risk level and 10-minute risk prediction results, with a processing delay of ≤80 ms.

9. A road traffic safety risk ranking and vehicle speed limiting control system according to claim 8, characterised in that: The speed limit control module generates scene-vehicle dual adaptation speed limit instructions based on a scene reference speed limit model, a vehicle type adaptation database, and a dynamic adjustment unit.

10. A road traffic safety risk ranking and vehicle speed limiting control system according to claim 9, characterised in that: The communication execution module realizes low-latency data interaction between roadside, vehicle, and cloud through a 5G communication unit and a cloud interaction interface, and combines a vehicle speed limit execution unit and a data feedback unit to achieve low-latency data interaction between roadside, vehicle, and cloud, with an instruction transmission delay of ≤40 ms. The optimization iteration module feeds back to the data analysis unit and model parameter iteration unit based on the execution feedback data, accident rate, violation rate, and traffic efficiency, and iteratively optimizes the risk model weight and speed limit parameters every 24 hours.

Images