Driving accompanying danger early warning and intervention method and system based on vehicle infrastructure cooperation

Abstract

The invention discloses an accompanying driving danger early warning and intervention method and system based on vehicle-road cooperation. The method comprises the steps that vehicle state, roadside environment, traffic participant and accompanying driving operation multi-source data are synchronously collected through a vehicle-mounted terminal, a roadside terminal, a surrounding vehicle terminal and an accompanying driving control multi-terminal; removing abnormal data through a 3 sigma criterion, supplementing missing data through a linear interpolation method, and obtaining standardized fusion data through a weighted fusion algorithm; identifying dangerous scenes and types based on a CNN-LSTM hybrid model, and quantifying danger levels; layered early warning information is output according to first-level danger, second-level danger and third-level danger; passive or active intervention is executed, and an intervention strategy is dynamically optimized through an intervention parameter adjustment formula; and storing all data in the cloud to form a dangerous event database and generate a redisk report. The system comprises multiple terminals, a cloud server and a data transmission module. According to the method, the safety and reliability of the accompanying driving scene are improved, and data support is provided for model and strategy optimization.

Description

Technical Field

[0001] This invention relates to the field of intelligent transportation technology, specifically to a method and system for hazard warning and intervention based on vehicle-road cooperation. Background Technology

[0002] Accompanying driving services are a crucial link in improving the driving skills of novice drivers and avoiding driving risks. Their core requirement is the real-time identification of dangerous scenarios during driving and the timely implementation of warnings or interventions to prevent accidents. Currently, safety in accompanying driving scenarios mainly relies on the manual observation and intervention of the accompanying driver, combined with simple alerts from single in-vehicle sensors (such as in-vehicle cameras or speed sensors). This presents the following significant technical challenges, and existing technologies have not yet provided effective solutions:

[0003] (1) Strong limitations in hazard identification: The existing accompanying driver warning method only relies on a single on-board sensor to collect the vehicle's own status data (such as vehicle speed and steering angle), and cannot obtain real-time dynamic information on the road environment, surrounding traffic participants (vehicles, pedestrians, non-motorized vehicles) and roadside facilities (traffic lights, zebra crossings, construction areas), resulting in a lag in the identification of complex dangerous scenarios such as meeting at intersections, overtaking in blind spots, and pedestrians crossing, with an identification accuracy rate of less than 70%, and prone to missed or false alarms.

[0004] (2) Insufficient timeliness of warnings: Existing technologies mostly issue warnings only when a dangerous scenario is about to occur or has already occurred. The warning interval is usually less than 1.5 seconds, leaving insufficient time for accompanying drivers or novice drivers to react. Moreover, the warning information is only a single sound and light reminder, without providing targeted guidance based on the level and type of danger, resulting in poor warning effectiveness.

[0005] (3) The intervention measures are passive and singular: the existing driving assistance intervention relies solely on the manual operation of the driving assistant (such as emergency braking and steering correction). Manual intervention has a reaction delay (average reaction time of 0.8-1.2 seconds) and lacks an active intervention mechanism. Although some simple systems have basic braking intervention functions, they do not combine vehicle-road cooperative data to make precise intervention parameter adjustments, which can easily lead to over-intervention (causing vehicle jerking) or under-intervention (failing to effectively avoid danger), resulting in low reliability.

[0006] (4) Data interaction is not smooth: The existing driving companion system lacks a standardized data interaction interface with roadside equipment and surrounding vehicles, which makes it impossible to achieve real-time synchronization and fusion of multi-source data of vehicle-road-person-vehicle. The data transmission delay is high (greater than 500ms) and the packet loss rate is high, which further reduces the timeliness of hazard identification and intervention and makes it difficult to meet the needs of complex road driving companion scenarios.

[0007] In summary, existing driving safety technologies suffer from shortcomings such as incomplete data collection, inaccurate hazard identification, untimely warnings, unintelligent intervention, and poor data interaction. These limitations prevent them from effectively addressing the complex hazard prevention and control issues in driving companion scenarios. Consequently, they suffer from low safety, poor reliability, and limited applicability. There is an urgent need for an efficient, accurate, and intelligent driving companion hazard warning and intervention solution based on vehicle-road cooperative technology. Summary of the Invention

[0008] To address the shortcomings of existing technologies, this invention discloses a method and system for hazard warning and intervention based on vehicle-road cooperation, in order to solve the problems mentioned in the background art.

[0009] To achieve the above objectives, the present invention provides the following technical solution: a method for hazard warning and intervention based on vehicle-road cooperation, comprising the following steps:

[0010] S1. Multi-source data acquisition: Simultaneously collect data on the status of the accompanying vehicle, roadside environment, surrounding traffic participants, and accompanying vehicle operation data through vehicle-mounted terminals, roadside terminals, accompanying driver control terminals, and surrounding vehicle terminals. All data is packaged in a standardized format.

[0011] S2. Data Preprocessing and Fusion: The multi-source data collected in step S1 is preprocessed by removing abnormal data and supplementing missing data. Then, the preprocessed data is fused using a weighted fusion algorithm to obtain standardized driving companion scenario fusion data.

[0012] S3. Hazard Identification and Level Determination: Based on the fused data obtained in step S2, the hazard identification model is used to identify the accompanying driving scenario in real time to determine whether there is a hazardous scenario and the type of hazard. If a hazardous scenario exists, the hazard level quantification value R is calculated using the formula R=P×C based on the probability of hazard occurrence P and the degree of hazard impact C. The hazard level is then divided into Level 1, Level 2, and Level 3 according to the value of R.

[0013] S4. Layered Warning: Based on the hazard identification results and hazard level in step S3, generate corresponding layered warning instructions and output warning information of corresponding priority to the novice driver and the accompanying driver through the driving control terminal.

[0014] S5. Intelligent Intervention: Based on the hazard level and hazard type, corresponding passive or active intervention measures are implemented. Data is collected and integrated in real time during the intervention process, and intervention parameters are dynamically adjusted through the intervention parameter adjustment formula.

[0015] S6. Data storage and review: All data from steps S1-S5 is stored on a cloud server to form a database of dangerous incidents during driving supervision, and a review report of driving supervision is generated.

[0016] Preferably, in step S1,

[0017] The accompanying vehicle's own status data includes real-time vehicle speed v, steering angle θ, brake pedal travel s, accelerator pedal travel t, vehicle position coordinates (x1, y1), and vehicle attitude angle α, with a collection frequency of 10-20Hz. Among them, the brake pedal travel s, together with the vehicle speed v, steering angle θ, and other data, constitutes the vehicle operation status dimension of the fused data F. The brake pedal travel s is a status feedback parameter for the execution of the intervention strategy, ensuring that the intervention measures are coordinated with the driver's operation and avoiding excessive intervention or conflict.

[0018] The roadside environmental data includes the status L of the traffic lights, the coordinates of the zebra crossing (x2, y2), the scope of the construction area, the speed limit v0, and the location and size of roadside obstacles, with a collection frequency of 5-10Hz;

[0019] The surrounding traffic participant data includes the location coordinates (x3, y3), real-time vehicle speed v3, driving direction β, and distance d from the accompanying vehicle of the surrounding vehicles, as well as the location coordinates (x4, y4), moving speed v4, and moving direction γ of the surrounding pedestrians, with a collection frequency of 5-10Hz;

[0020] The accompanying driver operation data includes the accompanying driver's braking operation signals, steering operation signals, voice command signals, and novice driver's operation action signals, with a collection frequency of 10Hz.

[0021] Preferably, in step S2, the formula for the weighted fusion algorithm is:

[0022] ,

[0023] Where F represents the fused data, n represents the number of data types, and w i The weight coefficients of the i-th class of data and satisfying =1,D i This represents the preprocessed data for the i-th class.

[0024] Preferably, in step S2, the outlier data removal adopts the 3σ criterion, that is, when a data point x deviates from the mean μ of this type of data and satisfies |x−μ|>3σ, where σ is the standard deviation of this type of data, it is determined to be outlier data and removed.

[0025] Missing data imputation uses linear interpolation: Let x be the valid data before and after the missing data point. k and x k+n If n≤3, then the missing data points are:

[0026] x k+i =x k +(x k+n -x k )× ,

[0027] Where i=1,2,...,n-1; n=4, the corresponding data types are vehicle data, roadside data, surrounding vehicle data, and driving companion operation data, respectively, and their corresponding initial weight coefficients are w1=0.35, w2=0.3, w3=0.25, and w4=0.1, respectively. When the packet loss rate of a certain type of data exceeds 10%, the corresponding weight coefficient is reduced by 50%, and the reduced weight is evenly distributed to other data types.

[0028] Preferably, in step S3, the hazard identification model is a CNN-LSTM hybrid model; the hazard types include blind spot overtaking hazard, intersection meeting hazard, pedestrian crossing hazard, speeding hazard, and operational error hazard; the hazard occurrence probability P ranges from 0 to 1, the hazard impact degree C ranges from 1 to 5, and the hazard level classification standard is: Level 1 hazard (minor hazard): 0.1≤R<0.3, Level 2 hazard (moderate hazard): 0.3≤R<0.6, Level 3 hazard (serious hazard): R≥0.6.

[0029] Preferably, in step S4, the specific method of layered warning is as follows: if it is a level one danger, output a voice reminder and a slight sound and light reminder on the vehicle instrument panel; if it is a level two danger, output precise voice guidance, sound and light reminder on the vehicle instrument panel, and slight vibration of the steering wheel; if it is a level three danger, output a high-frequency voice warning, a strong red light reminder on the vehicle instrument panel, vibration of the steering wheel and brake pedal, and send an emergency reminder signal to the driver assistance control terminal at the same time.

[0030] Preferably, in step S5, the formula for adjusting the intervention parameters is:

[0031] ,

[0032] Where a adj The adjusted braking acceleration is given by a0, which is the base braking acceleration of 3 m / s², and R, which is the hazard level quantification value. R is calculated from P × C, and the value of P depends on v. Therefore, v indirectly affects R by influencing P, ultimately determining the adjusted braking acceleration a. adj Size;

[0033] The specific methods of intelligent intervention are as follows:

[0034] 1) For Level 1 hazards, only passive intervention is implemented, that is, a warning message is issued, which is adjusted by the novice driver, while the accompanying person monitors the entire process;

[0035] 2) For Level 2 hazards, implement mild proactive intervention, including:

[0036] Automatically reduce vehicle speed to a safe speed: v s =v0×0.8,

[0037] Here, v0 represents the road speed limit, which is the data collected by the roadside terminal; the current speed v of the accompanying vehicle serves as a prerequisite for triggering this intervention, and the speed is only adjusted to v if v > v0. s And v and v s The difference directly determines the magnitude of the speed adjustment during intervention, where the speed adjustment amount = vv s .

[0038] At the same time, the steering angle is slightly adjusted with a correction amount of ≤5°, preserving the novice driver's operating authority, and the accompanying driver can manually remove the intervention;

[0039] 3) Level 3 hazards trigger emergency active intervention, including automatic emergency braking with a braking acceleration of 3-5 m / s², forced correction of steering direction with a correction amount ≤10°, locking the novice driver's operating authority, and only the accompanying driver can release the intervention through the emergency unlock command.

[0040] Preferably, in step S6, the data stored in the accompanying driving hazard event database includes raw collected data, preprocessed data, fused data, hazard identification results, early warning information, and intervention records; the accompanying driving review report includes the occurrence time, location, triggering cause, early warning information, intervention measures, and intervention effect of the hazard event, which is used to provide data support for hazard identification model optimization, weight coefficient adjustment, and accompanying driving strategy improvement.

[0041] This invention also provides a vehicle-road cooperative hazard warning and intervention system for implementing the aforementioned vehicle-road cooperative hazard warning and intervention method. The system includes an onboard terminal, a roadside terminal, surrounding vehicle terminals, a driver assistance control terminal, a cloud server, and a data transmission module. Each terminal communicates bidirectionally through the data transmission module.

[0042] The data transmission module adopts the V2X communication protocol combined with 5G communication technology, with a data transmission latency of ≤200ms and a packet loss rate of ≤3%.

[0043] The vehicle terminal is connected to the OBD interface and vehicle sensors of the accompanying vehicle, and is used to collect the vehicle's own status data and the novice driver's operation data, and to perform early warning and intervention operations.

[0044] The roadside terminal includes a roadside camera, a radar sensor, and a traffic light status acquisition module, which is used to collect roadside environmental data and perform preliminary preprocessing.

[0045] The surrounding vehicle terminal is used to collect driving status data of surrounding vehicles and realize data interaction between vehicles.

[0046] The accompanying driver control terminal is used to collect the operating instructions and voice commands of the accompanying driver, receive and display warning information, and provide emergency unlocking intervention function;

[0047] The cloud server includes a data storage unit, a data fusion unit, a hazard identification unit, an early warning generation unit, an intervention control unit, and a review and analysis unit, which are respectively used for data storage, data fusion, hazard identification and level determination, early warning instruction generation, intervention control, and review and analysis.

[0048] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0049] 1. This invention, by constructing a multi-source data acquisition architecture and weighted fusion algorithm for vehicle-road cooperation, and combining it with a dedicated hybrid neural network hazard identification model and quantitative judgment rules, breaks through the limitations of existing technologies that rely on a single vehicle sensor. It can effectively identify the dangerous characteristics of complex driving scenarios such as blind spot overtaking and pedestrian crossing, and significantly improve the comprehensiveness and accuracy of dangerous scenario identification.

[0050] 2. This invention generates hierarchical and progressive early warning instructions based on the analysis results of the hazard level quantification value and hazard type, and provides different intensity warning methods and priority feedback mechanisms. It overcomes the shortcomings of existing technologies in that the early warning is not timely and lacks specificity. It can provide novice drivers and accompanying drivers with more reaction time and clear handling guidance, and significantly improve the effectiveness of early warning.

[0051] 3. This invention designs a progressive control strategy that combines passive intervention with multi-level active intervention for different levels of dangerous scenarios. It also incorporates a dynamic adjustment mechanism for intervention parameters to achieve adaptive optimization of intervention intensity. Simultaneously, it enables low-latency and high-reliability interaction of multi-source data between vehicle, road, person, and vehicle, effectively solving the problems of slow response of manual intervention and inaccurate intervention of existing systems, thereby improving the safety and reliability of the accompanying driving process. Attached Figure Description

[0052] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used together with the embodiments of the invention to explain the invention and do not constitute a limitation thereof.

[0053] In the attached diagram:

[0054] Figure 1 This is a framework diagram of the accompanying driving hazard warning and intervention system based on vehicle-road cooperation of the present invention;

[0055] Figure 2 This is a flowchart of the steps of the accompanying driving hazard warning and intervention method based on vehicle-road cooperation of the present invention. Detailed Implementation

[0056] The preferred embodiments of the present invention will be described below with reference to the accompanying drawings. It should be understood that the preferred embodiments described herein are for illustration and explanation only and are not intended to limit the present invention.

[0057] This embodiment provides a method and system for warning and intervening in dangerous situations during driving companionship based on vehicle-road cooperation, which is applicable to urban road driving companionship scenarios. The specific structure and implementation steps are as follows:

[0058] The overall system structure is as follows Figure 1 As shown, it includes an in-vehicle terminal, a roadside terminal, surrounding vehicle terminals, a driver assistance control terminal, a cloud server, and a data transmission module. Each terminal achieves bidirectional communication through the data transmission module, and the specific layout is as follows:

[0059] Vehicle-mounted terminal:

[0060] Installation location: Center console of the vehicle, connected to the vehicle's OBD interface, speed sensor, steering sensor, attitude sensor, and vehicle camera.

[0061] Hardware configuration: The data acquisition unit uses an STM32F407 microcontroller, and the instruction execution unit includes an audible and visual alarm (operating voltage 12V, alarm volume ≥80dB), a steering wheel vibration motor (power 5W, vibration frequency 50Hz), and a braking intervention module (connected to the vehicle's braking system hydraulic circuit). The data transmission subunit uses a Balong 5000 5G communication chip, which supports the V2X communication protocol.

[0062] Functions: Collects vehicle status data and novice driver operation data, receives warning and intervention commands, and executes audible and visual reminders, vibration warnings, and braking / steering intervention operations.

[0063] Roadside terminal:

[0064] Deployment locations: one unit every 500 meters at intersections of urban roads and on both sides of zebra crossings.

[0065] Hardware configuration: The roadside camera is a 1080P high-definition infrared camera (night vision distance ≥50m), the radar sensor is a millimeter-wave radar (detection distance 0-100m, ranging accuracy ±0.5m), the traffic light status acquisition module is connected to the traffic signal control system through RS485 interface, and the data processing unit is an ARM Cortex-A9 processor (1GHz).

[0066] Function: Collect roadside environmental data, perform preliminary preprocessing (format conversion, data filtering), and then send it to the vehicle terminal and cloud server.

[0067] Surrounding vehicle terminals:

[0068] Installation location: Center console of small cars that participate in traffic.

[0069] Hardware components include a GPS+BeiDou dual-mode positioning module (positioning accuracy ±1m), a vehicle speed acquisition module (sampling accuracy ±1km / h), and a data transmission subunit (using the same 5G communication chip as the vehicle terminal).

[0070] Function: Collects driving status data of surrounding vehicles and sends it to the vehicle terminal and roadside terminal through the data transmission module to realize data interaction between vehicles.

[0071] Accompanying driver control terminal:

[0072] How to carry: Handheld by the accompanying driver, using a 10.1-inch tablet computer (1920×1200 resolution).

[0073] Hardware configuration: The operation panel is equipped with an emergency unlock button (mechanical button, pressing stroke ≥2mm) and a voice command button. The voice acquisition module uses a noise-canceling microphone (signal-to-noise ratio ≥60dB). The warning display module is a tablet touch screen. The emergency unlock module uses an independent communication channel to transmit unlock commands.

[0074] Functions: Collects the operation and voice commands of the accompanying driver, receives and displays warning information, and provides emergency unlocking intervention function.

[0075] Cloud server:

[0076] Software modules: The data storage unit uses a MySQL 8.0 database, while the data fusion unit, hazard identification unit, early warning generation unit, intervention control unit, and debriefing analysis unit are developed using the Python programming language and deployed using the TensorFlow framework to implement a CNN-LSTM hybrid model.

[0077] Functions: Enables data storage, multi-source data fusion, hazard identification and level determination, early warning instruction generation, intervention control, and debriefing report generation.

[0078] Data transmission module:

[0079] Communication protocol: V2X communication protocol is adopted, combined with 5G SA (standalone) communication technology.

[0080] Hardware components include a communication chip, an omnidirectional antenna (gain ≥8dBi), and a protocol parsing unit (dedicated V2X protocol parsing chip).

[0081] Performance metrics: Data transmission latency 150-200ms, packet loss rate ≤2%, supports concurrent data transmission from multiple terminals.

[0082] This embodiment presents an execution flow of a vehicle-road cooperative-based method for hazard warning and intervention during driving, such as... Figure 2 As shown, the specific steps are as follows:

[0083] S1: Multi-source data acquisition. After the system starts, each terminal synchronously acquires data at a preset frequency.

[0084] Data collected by the vehicle terminal: real-time vehicle speed v=45km / h, steering angle θ=0° (straight driving), brake pedal travel s=0mm, accelerator pedal travel t=20mm, vehicle position coordinates (x1,y1)=(116.403874,39.914885), vehicle attitude angle α=0°, novice driver did not brake or steer.

[0085] Data collected by the roadside terminal: Traffic light status L=1 (green light), zebra crossing location coordinates (x2, y2) = (116.403974, 39.914985), road speed limit v0 = 50km / h, no obstacles or construction areas on the roadside.

[0086] Data collected from surrounding vehicle terminals: 8 surrounding vehicles within 100m, with location coordinates (x3, y3) = (116.403774, 39.914885), real-time vehicle speed v3 = 35km / h, driving direction β = 0° (same direction), distance from accompanying vehicle d = 10m, and no pedestrians crossing.

[0087] Data collected by the accompanying driver control terminal: The accompanying driver did not brake or steer, and did not issue any voice commands. All data is packaged in a standardized format, with vehicle position coordinates uniformly in the WGS84 coordinate system, speed units uniformly in km / h, and angle units uniformly in °.

[0088] S2: The data preprocessing and fusion process includes:

[0089] Outlier data removal: Calculate the mean and standard deviation of each type of data. In this embodiment, all collected data satisfy |x-μ|≤3σ, there is no outlier data, and all are retained.

[0090] Weighted fusion: n=4, weight coefficients w1=0.35, w2=0.3, w3=0.25, w4=0.1, substituting into the formula F=w1D1+w2D2+w3D3+w4D4, where D1 is vehicle data (e.g., vehicle speed 45km / h), D2 is roadside data (e.g., speed limit 50km / h), D3 is surrounding vehicle data (e.g., distance 10m), and D4 is accompanying driver operation data (no operation command), the fused data F is calculated, which includes comprehensive information such as the fused vehicle speed 42km / h and the surrounding vehicle distance 10.2m.

[0091] S3: The specific process for hazard identification and level determination includes:

[0092] The fused data F is input into the CNN-LSTM hybrid hazard identification model. The model extracts spatial features (such as the accompanying vehicle traveling in the same direction as surrounding vehicles and sufficient distance) through the CNN layer and captures temporal features (such as stable vehicle speed and no sudden operations) through the LSTM layer. The output hazard identification result is no hazard (0), and no level judgment is required. The probability of hazard occurrence P is calculated based on parameters such as distance, vehicle speed, and operation actions in the fused data.

[0093] When the distance d between the accompanying vehicle and surrounding vehicles is less than 5m, and the speed difference between the two vehicles is |v-v3|>10km / h, P=0.7;

[0094] Where: v represents the real-time speed of the accompanying vehicle, which is the core data collected by the vehicle terminal; v3 represents the real-time speed of surrounding vehicles, which is the data collected by the terminals of surrounding vehicles.

[0095] The speed difference |v-v3| represents the key indicator for determining dangerous scenarios such as "close following + speed mismatch", which determines the value of P and then affects the determination of the danger level through R=P×C.

[0096] To verify the hazard identification and level determination logic, a scenario change is assumed: the accompanying vehicle's speed v = 55 km / h (speeding), the distance to surrounding vehicles d = 4 m (close following distance), and other data remain unchanged. The fused data F is then updated, and the model identifies the hazard type as "speeding hazard + close following hazard"; the probability of hazard occurrence is calculated as P = 0.7 (speeding and close following result in a high probability of an accident), and the hazard impact level C = 3, indicating a potential rear-end collision with moderate impact. Substituting these values ​​into the formula R = P × C = 0.7 × 3 = 2.1, since R ≥ 0.6, it is classified as a level three hazard.

[0097] S4: The specific steps for tiered early warning include:

[0098] For a "no danger" outcome: the system does not output any warning information and continuously monitors data changes. For a hypothetical "Level 3 danger" scenario: the warning generation unit generates a Level 3 warning command and sends it to the vehicle terminal and the accompanying driver control terminal. The vehicle terminal outputs a high-frequency voice warning, "Serious danger! Excessive speed and close following distance, please slow down and stop immediately!", flashes a bright red light on the vehicle's instrument panel (frequency 5Hz), and vibrates the steering wheel and brake pedal (amplitude 2mm). The warning display module of the accompanying driver control terminal displays "Level 3 danger: speeding + close following," and simultaneously issues a high-frequency audio and visual warning (volume 90dB, light flashing frequency 5Hz).

[0099] In the case of a Level 1 hazardous scenario (e.g., v=52km / h, d=8m): a Level 1 warning will be issued, with a voice reminder "Please pay attention to your speed, the speed limit ahead is 50km / h", and a slight audio-visual reminder will be given on the vehicle's instrument panel (30% brightness of the lights and 60dB volume).

[0100] In a Level 2 hazardous scenario, when v=53km / h and d=6m: a Level 2 warning is issued, with voice guidance saying "Moderate danger! Please reduce speed to below 50km / h and maintain a safe distance from surrounding vehicles," and the vehicle's instrument panel provides an audio and visual reminder, with headlight brightness at 60%, volume at 70dB, and slight vibration of the steering wheel with an amplitude of 1mm.

[0101] S5: The specific steps of intelligent intervention include:

[0102] For a "no danger" result: the system will not perform any intervention operations, the novice driver will drive autonomously, and the accompanying person will monitor the entire process.

[0103] For the hypothetical "Level 3 Hazard" scenario: The intervention control unit generates an emergency active intervention command and sends it to the command execution unit of the onboard terminal. Substituting the intervention parameter adjustment formula a... adj =a0×R / 0.6=3×2.1 / 0.6=10.5m / s². Considering the maximum safe braking acceleration of the vehicle is 5m / s², the actual braking acceleration is set to 5m / s², and automatic emergency braking is executed. At the same time, the steering direction is slightly corrected (correction amount 5°) to avoid collision with surrounding vehicles. The braking and steering operation permissions of the novice driver are locked, and only the accompanying driver can release the intervention through the emergency unlock button of the accompanying driver control terminal.

[0104] In a level 2 hazardous scenario: implement mild active intervention and automatically adjust the vehicle speed to v. s =50×0.8=40km / h, steering correction amount 3°, novice driver's operating rights are retained, and the accompanying driver can manually cancel the intervention through the accompanying driver control terminal.

[0105] In the case of a Level 1 dangerous scenario: only a warning message is issued, and no active intervention is performed. The novice driver can adjust the vehicle speed or distance independently.

[0106] S6: Specific steps for data storage and review include:

[0107] Data storage: All data in steps S1-S5, including raw collected data, preprocessed data, fused data, hazard identification results, early warning information and intervention records, are stored in a MySQL database on a cloud server to form a database of accompanying driving hazard events. The data retention period is 3 years.

[0108] Post-mortem Analysis: For the hypothetical "Level 3 Hazard" scenario, the post-mortem analysis unit generates a post-mortem report based on database data. The report includes: the time of the hazardous event (e.g., 2024-05-20 14:30:25), the location (e.g., coordinates 116.403874, 39.914885), the triggering cause (e.g., the accompanying vehicle exceeding the speed limit by 5 km / h and being too close to surrounding vehicles by 4 meters), warning information (e.g., Level 3 warning, high-frequency voice + strong light + vibration), intervention measures (e.g., emergency braking at 5 m / s², 5° steering correction), and the intervention effect (e.g., the vehicle decelerates to 30 km / h within 3 seconds, increasing the distance to surrounding vehicles to 8 meters, thus avoiding a collision). Simultaneously, the report analyzes operational problems of novice drivers, such as neglecting speed and following distance, and provides guidance to accompanying drivers on "strengthening speed control and following distance judgment training." It also provides data support for optimizing the hazard recognition model, such as adjusting the recognition threshold for speeding and close following scenarios.

[0109] This embodiment forms a closed loop of data collection, identification, early warning, intervention, and optimization through vehicle-road cooperative multi-source data fusion, intelligent hazard identification, hierarchical early warning and intervention, and data review. This effectively solves the pain points of existing driving companion technology and improves the safety and reliability of driving companion scenarios.

[0110] Finally, it should be noted that the above descriptions are merely preferred embodiments of the present invention and are not intended to limit the present invention. 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 hazard warning and intervention in driving companionship based on vehicle-road cooperation, characterized in that, Includes the following steps: S1. Simultaneously collect data on the status of the accompanying vehicle, roadside environment, surrounding traffic participants, and accompanying vehicle operation through the vehicle-mounted terminal, roadside terminal, accompanying driver control terminal, and surrounding vehicle terminals. S2. Perform preprocessing operations such as abnormal data removal and missing data supplementation on the multi-source data collected in step S1, and then fuse the preprocessed data through a weighted fusion algorithm to obtain standardized driving companion scenario fusion data. S3. Based on the fused data obtained in step S2, the accompanying driving scenario is identified in real time through the hazard identification model to determine whether there is a dangerous scenario and the type of hazard. If there is a dangerous scenario, the hazard level quantification value R is calculated based on the probability of hazard occurrence P and the degree of hazard impact C using the formula R=P×C. The hazard level is then divided into level one, level two, and level three according to the value of R. S4. Based on the hazard identification results and hazard level in step S3, generate corresponding hierarchical warning instructions and output warning information of corresponding priority to the novice driver and the accompanying driver through the accompanying driver control terminal. S5. Based on the hazard level and hazard type, implement corresponding passive or active intervention measures. During the intervention process, collect and integrate data in real time, and dynamically adjust the intervention parameters through the intervention parameter adjustment formula. S6. Store all data from steps S1-S5 on a cloud server to form a database of dangerous incidents during driving supervision, and generate a driving supervision review report.

2. The method for hazard warning and intervention based on vehicle-road cooperation as described in claim 1, characterized in that: In step S1, The accompanying vehicle's own status data includes the vehicle's real-time speed v, steering angle θ, brake pedal travel s, accelerator pedal travel t, vehicle position coordinates (x1, y1), and vehicle attitude angle α. The roadside environmental data includes the status L of the traffic lights, the coordinates of the zebra crossing (x2, y2), the scope of the construction area, the speed limit v0, and the location and size of roadside obstacles. The surrounding traffic participant data includes the location coordinates (x3, y3), real-time vehicle speed v3, driving direction β, and distance d from the accompanying vehicle of the surrounding vehicles, as well as the location coordinates (x4, y4), moving speed v4, and moving direction γ of the surrounding pedestrians; The accompanying driver operation data includes the accompanying driver's braking operation signals, steering operation signals, voice command signals, and operation action signals of the novice driver.

3. The method for hazard warning and intervention based on vehicle-road cooperation as described in claim 1, characterized in that: In step S2, the formula for the weighted fusion algorithm is: ， Where F represents the fused data, n represents the number of data types, and w i The weight coefficients of the i-th class of data and satisfying =1,D i This represents the preprocessed data for the i-th class.

4. The method for hazard warning and intervention based on vehicle-road cooperation as described in claim 1, characterized in that: In step S2, outlier data removal adopts the 3σ criterion, that is, when a data point x deviates from the mean μ of this type of data and satisfies |x −μ|>3σ, where σ is the standard deviation of this type of data, it is judged as outlier data and removed. Missing data imputation uses linear interpolation: Let x be the valid data before and after the missing data point. k and x k+n If n≤3, then the missing data points are: x k+i =x k +(x k+n −x k )× , Where i=1,2,...,n-1; n=4, the corresponding data types are vehicle data, roadside data, surrounding vehicle data, and driving companion operation data, respectively.

5. The method for hazard warning and intervention based on vehicle-road cooperation as described in claim 1, characterized in that: In step S3, the hazard identification model is a CNN-LSTM hybrid model; the hazard types include blind spot overtaking hazard, intersection meeting hazard, pedestrian crossing hazard, speeding hazard, and operational error hazard.

6. The method for hazard warning and intervention based on vehicle-road cooperation as described in claim 1, characterized in that: In step S4, the specific method of tiered warning is as follows: if it is a level one danger, output a voice reminder and a slight sound and light reminder on the vehicle instrument panel; If it is a level 2 hazard, the system will provide precise voice guidance, audio and visual alerts on the in-vehicle instrument panel, and slight vibrations in the steering wheel. If the danger level is level three, a high-frequency voice warning will be issued, a bright red light will illuminate on the vehicle's instrument panel, and the steering wheel and brake pedal will vibrate. At the same time, an emergency warning signal will be sent to the driver assistance control terminal.

7. The method for hazard warning and intervention based on vehicle-road cooperation as described in claim 1, characterized in that: In step S5, the formula for adjusting the intervention parameters is: ， Where a adj The adjusted braking acceleration is given by a0, where a0 is the base braking acceleration of 3 m / s², and R is the hazard level quantification value. The specific method of intelligent intervention is as follows: 1) For Level 1 hazards, only passive intervention is implemented, that is, a warning message is issued, which is adjusted by the novice driver, while the accompanying person monitors the entire process; 2) For Level 2 hazards, implement mild proactive intervention, including: Automatically reduce vehicle speed to a safe speed: v s =v0×0.8; Slightly adjust the steering angle with a correction amount of ≤5°, retain the novice driver's operating rights, and the accompanying driver can manually remove the intervention; 3) Level 3 hazards trigger emergency active intervention, including automatic emergency braking with a braking acceleration of 3-5 m / s², forced correction of steering direction with a correction amount ≤10°, locking the novice driver's operating authority, and only the accompanying driver can release the intervention through the emergency unlock command.

8. The method for hazard warning and intervention based on vehicle-road cooperation as described in claim 1, characterized in that: In step S6, the data stored in the accompanying driving hazard event database includes raw collected data, preprocessed data, fused data, hazard identification results, early warning information, and intervention records; the accompanying driving review report includes the occurrence time, location, triggering cause, early warning information, intervention measures, and intervention effect of the hazard event, which is used to provide data support for hazard identification model optimization, weight coefficient adjustment, and accompanying driving strategy improvement.

9. A vehicle-road cooperative driving hazard warning and intervention system, used to implement the vehicle-road cooperative driving hazard warning and intervention method according to any one of claims 1-8, characterized in that: It includes vehicle-mounted terminals, roadside terminals, surrounding vehicle terminals, driver assistance control terminals, cloud servers, and data transmission modules; each terminal achieves bidirectional communication through the data transmission module. The data transmission module adopts the V2X communication protocol combined with 5G communication technology; The vehicle terminal is connected to the OBD interface and vehicle sensors of the accompanying vehicle, and is used to collect the vehicle's own status data and the novice driver's operation data, and to perform early warning and intervention operations. The roadside terminal includes a roadside camera, a radar sensor, and a traffic light status acquisition module, which is used to collect roadside environmental data and perform preliminary preprocessing. The surrounding vehicle terminal is used to collect driving status data of surrounding vehicles and realize data interaction between vehicles. The accompanying driver control terminal is used to collect the operating instructions and voice commands of the accompanying driver, receive and display warning information, and provide emergency unlocking intervention function; The cloud server includes a data storage unit, a data fusion unit, a hazard identification unit, an early warning generation unit, an intervention control unit, and a review and analysis unit, which are respectively used for data storage, data fusion, hazard identification and level determination, early warning instruction generation, intervention control, and review and analysis.

Images