Method for automatically monitoring and identifying dangerous driving behavior based on driving scene

Abstract

The invention provides a method for automatically monitoring and identifying dangerous driving behaviors based on a driving scene, and the method achieves the comprehensive monitoring of a vehicle driving state and a vehicle driving road section through the installation of a satellite positioning device, the collection of CAN bus data, the deployment of high-precision map road network data, and the data fusion. By setting a behavior threshold value and combining time sequence analysis and map data, the judgment accuracy is enhanced, a real-time monitoring algorithm engine is developed, and dangerous driving behaviors are accurately monitored and recognized in a complex and changeable driving scene. According to the method, dangerous driving behaviors of the driver can be corrected in time, dangerous driving events are recorded and managed at the same time, the running safety risk is reduced, the driving safety is improved, and losses caused by traffic accidents are reduced. Various dangerous driving scenes are covered, a driver is helped to reasonably control the vehicle speed, digitization of motorcade management and transparency of risk management are achieved, and the requirement for supervision compliance is met.

Description

Technical Field

[0001] The present invention relates to the technical field of vehicle safe driving, and in particular to a method for automatically monitoring and identifying dangerous driving behaviors based on driving scenarios. Background Art

[0002] Commercial vehicles are motor vehicles used for commercial road transport, primarily for the transportation of people and goods for compensation. In the commercial vehicle transport industry, commercial vehicles typically carry passengers or important supplies. Accidents often result in significant casualties and property damage, and the transport company will be held responsible.

[0003] In the existing technology, although vehicles are required to install satellite positioning devices and connect to the networked joint control system, there are still many safety issues that cannot be resolved: there are blind spots in safety management during driving, and only the real-time location of the vehicle can be monitored, but it is impossible to know whether the vehicle is speeding, driving while fatigued, whether the driving section is dangerous, whether it encounters rainy or snowy days, etc. The system cannot give the driver targeted safe driving warning reminders and guidance; for corporate entities, there is a lack of effective management and control measures while the driver is driving, and only empirical speculation and reminders can be used to accurately remind and supervise the driver's safe driving; there is a lack of understanding of the driver's overall driving behavior and a lack of management measures for the driver, and it is impossible to have a comprehensive understanding of the driving risks of the driver during the entire driving process, nor to conduct assessments and evaluations based on the driver's overall driving behavior to motivate the driver to improve bad driving behavior. Summary of the Invention

[0004] In order to solve the above technical problems, the present invention proposes a method for automatically monitoring and identifying dangerous driving behaviors based on driving scenarios.

[0005] The present invention aims to reduce the safety risks of commercial vehicles, thereby improving driving safety and reducing losses caused by traffic accidents. A convenient method is proposed, comprising the following steps: S1. Data collection and fusion. This includes the following sub-steps:

[0006] S101: Install satellite positioning device; Satellite positioning devices are installed on operating transport vehicles to continuously collect the vehicle's real-time location information at a high frequency, and obtain satellite positioning data to accurately depict the vehicle's driving trajectory.

[0007] S102: Collect CAN bus data; Connect the vehicle's CAN bus interface to the vehicle's electronic control system to collect vehicle status and driving operation signals.

[0008] S103: Deploy high-precision map road network data; Deploy high-precision map road network data containing detailed road information, and dynamically match relevant information of the current road according to the real-time location of the vehicle.

[0009] S104: Data fusion; Satellite positioning data is de-noised and then integrated with map road network data. Based on the vehicle's real-time location, road information such as speed limits, curve curvature, and slope are extracted from the map data and correlated with the vehicle's CAN signal data.

[0010] S2. Set threshold model rules and monitor calculations. This includes the following sub-steps:

[0011] S201: Setting behavior threshold; Behavioral thresholds are set based on driving safety standards, such as braking for ≥10 seconds, or driving continuously for 4 hours.

[0012] S202: Combined time series analysis; Combined with time series analysis of continuous behavioral risks, if sudden acceleration, sudden deceleration, or speeding occurs ≥3 times within 5 consecutive minutes, it is judged as "aggressive driving".

[0013] S203: Enhance determination accuracy by combining map data; Relevant data from the map is compared with actual driving data from a specific driving process. For example, if the speed limit on a road is 60 km / h and the monitored speed exceeds 60 km / h, it is considered speeding. Combining map data can enhance the accuracy of dangerous behavior judgments and avoid misjudgments due to road changes.

[0014] S204: Develop real-time monitoring algorithm engine; Based on the threshold model rules set above, a real-time monitoring algorithm engine is developed. The engine processes the uploaded vehicle data in the form of data streams and determines in real time whether each operating vehicle has triggered the threshold of bad driving behavior.

[0015] S3. Real-time monitoring, calculation, and identification of dangerous driving behaviors. This includes the following sub-steps:

[0016] S301: Rule engine deployment; Use the Drools rule engine to execute rules by priority and detect dangerous driving behaviors in real time.

[0017] S302: Determination and early warning; Determine if dangerous driving behavior starts recording and issuing warnings, continuously monitor dangerous driving behavior, and determine whether the dangerous driving behavior has been resolved based on driving data. Or, when it is impossible to continuously monitor dangerous driving behavior or an abnormal situation occurs, forcibly end the monitoring of dangerous driving behavior. Regardless of whether it ends normally or abnormally, the complete dangerous driving event will be pushed to the platform for record storage.

[0018] S303: Cycle monitoring; When there is no similar dangerous driving behavior monitoring in progress, monitoring will be restarted according to the rule engine to achieve cyclic monitoring.

[0019] S4. Real-time warning and correction of the driver. This includes the following sub-steps:

[0020] S401: Push warning information; Dangerous driving incidents are identified and pushed to the platform. The platform configures corresponding speech tags and speech content, and uses voice broadcasts to remind the driver according to the speech content to correct the driver's bad driving behavior and guide him to drive safely.

[0021] S402: Notify the fleet manager; When the warning information is pushed, the reminder information will be notified to the fleet manager. After receiving the warning information, the vehicle's driving status and driver behavior can be viewed in real time through the monitoring system. If necessary, the driver can be communicated with through the on-board intercom system to require him to correct the bad driving behavior immediately.

[0022] S5. Record and manage dangerous driving incidents. This includes the following sub-steps:

[0023] S501: Recording and statistics; Based on statistics of dangerous driving incidents of drivers over a certain period of time, driver ranking assessments are conducted and parallel month-on-month comparisons are made.

[0024] S502: Motivation and guidance; If the driver improves his bad driving behavior, he can be given certain rewards to encourage and guide the driver to drive safely, forming a closed-loop management mechanism of "monitoring-early warning-rectification".

[0025] Furthermore, in step S104 of data collection and fusion, the satellite positioning device continuously collects the real-time position information of the vehicle at a high frequency to ensure accurate depiction of the vehicle's driving trajectory.

[0026] Furthermore, step S202 is combined with the time series analysis step to analyze the data using a fixed time window, calculate relevant statistical features, and compare the statistical features with the corresponding set thresholds to ensure accurate identification of continuous dangerous driving behavior patterns.

[0027] Furthermore, the rule engine deployment in step S301 adopts a priority queue scheduling mechanism, which can dynamically adjust the rule execution order. When high-risk behavior is detected, the rule execution priority is automatically increased to ensure immediate judgment and warning of high-risk behavior.

[0028] Furthermore, in the judgment and warning step S302, after it is determined that the dangerous driving behavior has been resolved, the end time of the event is automatically recorded, and the complete event record is pushed to the management platform for storage and analysis to ensure the integrity and availability of the data.

[0029] Furthermore, in step S401 of pushing the warning information, the platform-configured speech tags and speech content can be customized according to different types of dangerous driving behaviors to ensure the pertinence and effectiveness of the warning information.

[0030] Furthermore, in the recording and statistics step S501, monthly statistics are collected on the driver's dangerous driving incidents to provide quantitative data support for the driver's performance evaluation and training.

[0031] The present invention proposes a method for automatically monitoring and identifying dangerous driving behaviors based on driving scenarios. Based on information such as vehicle positioning, CAN bus data, and map road network data, the method adopts big data real-time monitoring algorithm engine technology, innovatively covering a variety of dangerous driving scenarios, giving drivers safety reminders and guidance, reducing driving safety risks, and improving driving safety; in the field of operating vehicle management, by real-time monitoring and retaining dangerous driving data, it helps companies meet regulatory compliance requirements, accurately remind and supervise drivers to drive safely, and avoid penalties due to violations; at the same time, it realizes the digitization of fleet management and the transparency of risk management. Companies can obtain the dangerous driving behavior data of each vehicle through the background, understand the driver's driving behavior throughout the entire driving process, analyze the risky behavior, conduct targeted safety training and assessment of drivers, and reduce losses caused by traffic accidents. BRIEF DESCRIPTION OF THE DRAWINGS

[0032] Figure 1 The figure is a schematic diagram of the overall workflow of a method for automatically monitoring and identifying dangerous driving behaviors based on driving scenarios according to the present invention.

[0033] Figure 2 This is a schematic diagram of the workflow of step S3 of a method for automatically monitoring and identifying dangerous driving behaviors based on driving scenarios of the present invention.

[0034] Figure 3 This is a schematic diagram of the workflow of step S5 of a method for automatically monitoring and identifying dangerous driving behaviors based on driving scenarios of the present invention. DETAILED DESCRIPTION

[0035] In order to provide a further understanding of the purpose, structure, features, and functions of the present invention, the present invention is described in detail below with reference to the embodiments.

[0036] A method for automatically monitoring and identifying dangerous driving behaviors based on driving scenarios includes the following steps:

[0037] S1. Data collection and fusion. This includes the following sub-steps:

[0038] S101: Install satellite positioning device; Satellite positioning devices installed on operational transport vehicles collect real-time location information at high frequency, including longitude, latitude, altitude, speed, and heading angle, to accurately depict the vehicle's trajectory. This data is transmitted in real time to a backend server via a wireless 5G communication module. The satellite positioning device integrates a wireless communication module, which connects to the positioning module via a serial UART interface. The satellite positioning device receives signals from multiple satellites to calculate the vehicle's geographic location, including coordinate data such as longitude, latitude, and altitude, as well as kinematic data such as speed and heading angle. The positioning module packages the collected satellite positioning data according to the TCP / IP communication protocol and transmits it to the wireless communication module. Upon receiving the data, the wireless communication module transmits the satellite positioning data to the backend server using wireless signals from the mobile network.

[0039] By installing a satellite positioning device, vehicles can be precisely located. High-frequency data collection can record the vehicle's driving trajectory in detail, providing accurate spatial information for subsequent driving status analysis. For example, while a vehicle is in motion, a Beidou satellite positioning device can accurately and continuously collect real-time location information and record the vehicle's driving trajectory on the highway, helping to more accurately determine whether the vehicle is traveling on dangerous roads or has deviated from its normal route.

[0040] S102: Collect CAN bus data; The vehicle's electronic control system is connected to the vehicle's CAN bus interface via the vehicle's controller area network (CAN) bus interface. Specifically, the vehicle's OBD-II (on-board diagnostic) interface is connected to the CAN bus interface. The data acquisition device is connected to the OBD-II interface via a dedicated cable. The data acquisition device contains a CAN bus controller and transceiver for communication with the vehicle's CAN bus network. During the connection process, ensure that the data acquisition device's power supply is properly connected to the vehicle's power supply and that the communication line is free of interference. The vehicle's electronic control system transmits various vehicle status and driver operation signals via the CAN bus. These signals are transmitted on the bus in the form of CAN frames. The data acquisition device's CAN bus controller monitors data transmission on the bus. When a data frame is detected, it receives the relevant data frame based on preset filtering conditions (such as frame ID). For example, to collect vehicle acceleration and deceleration signals, the data acquisition device receives relevant CAN frames from the vehicle's powertrain. The data acquisition device collects CAN bus data at a frequency of 20 Hz, meaning it collects vehicle status and driver operation signals 20 times per second. To achieve this frequency, the data acquisition device has an internal timer that controls the data acquisition interval. Each time data is collected, the collector will parse the data in the CAN frame and extract specific physical values, such as acceleration, deceleration, steering angle, throttle opening, brake pedal status, gear information, etc.

[0041] The parsed data is transmitted to the onboard terminal via the vehicle network. The terminal performs preliminary data processing, such as format conversion and compression, to reduce storage space. The terminal then stores the processed data in local memory and sends it to the backend server as needed. This data transmission also utilizes a wireless 5G communication module, similar to the wireless communication in S101, sending data to the backend server via the mobile network.

[0042] Collecting CAN bus data can capture various vehicle status information and driver operation details. This data reflects the driver's real-time control behavior, such as the depth of the accelerator pedal and the strength of the brakes. Analysis of this data allows for a more comprehensive assessment of the driver's driving safety. For example, by analyzing the brake pedal status and deceleration, it is possible to accurately determine whether sudden braking was involved, thereby providing the driver with targeted driving improvement recommendations.

[0043] S103: Deploy high-precision map road network data; High-precision maps containing detailed road attribute information are deployed through cloud services, dynamically matching relevant information about the current road based on the vehicle's real-time location. High-precision map data is provided by professional map providers, including detailed road attributes such as road type (e.g., expressway, urban arterial road, rural road), speed limit, curve radius, slope, and traffic sign and signal light locations. The collected data is then processed and modeled to generate high-precision map data, which is stored in GeoJSON format on cloud servers. To ensure data currency, map data is regularly updated, for example, to adjust map data based on factors such as road construction and changes in traffic regulations.

[0044] Install a map data access software API on the vehicle's onboard terminal or backend server. While the vehicle is driving, the onboard terminal uses the vehicle's real-time location, obtained via satellite positioning, to send a data request to the cloud server via the wireless network. Upon receiving the request, the cloud server extracts the corresponding road segment data from a high-precision map database based on the vehicle's location information and sends it back to the onboard terminal. The onboard terminal then parses and displays the data.

[0045] The vehicle terminal matches the vehicle's real-time location with road data in the high-precision map to obtain relevant road information. For example, by comparing the vehicle's latitude and longitude coordinates with the road vector data in the map, it can determine the type of road, speed limit, slope, and other road information.

[0046] The detailed road information provided by high-precision maps can help accurately judge the vehicle's driving environment. For example, the location of traffic signs and signal lights in the map can help better understand the traffic rules and environment in which the vehicle is located, avoiding misjudgment of the driver's behavior.

[0047] S104: Data fusion; The satellite positioning data is integrated with the map road network data. Specifically, the raw satellite positioning data is subjected to noise reduction processing to generate a smooth trajectory sequence. Then, based on the real-time position of the vehicle, the relevant information of the corresponding current road section is extracted from the map road network data and associated with the vehicle's CAN signal data.

[0048] Because satellite signals may be affected by environmental factors such as buildings and mountains, the collected satellite positioning data may contain noise. To improve data quality, a Kalman filter algorithm is used to reduce noise in satellite positioning data. The Kalman filter calculates the optimal estimate based on previous estimates and current measurements, thereby reducing the impact of noise.

[0049] Parse the collected CAN bus data to extract specific physical quantity features, such as acceleration, brake pedal state, steering angle, etc. Both satellite positioning data and CAN bus data have their own timestamps. On the background server, compare the timestamps of the two types of data to find the pair of satellite positioning data and CAN bus data with the closest timestamps. Calculate the difference Δt between the timestamp of the satellite positioning data and the timestamp of the CAN bus data. If Δt is less than a certain threshold (100 milliseconds), it is considered that these two pieces of data are collected simultaneously and can be directly fused; if Δt is large, timestamp alignment is required. Use the linear interpolation algorithm to interpolate the CAN bus data. Assume that the acquisition time of the satellite positioning data is t1, and the acquisition times of two adjacent CAN bus data are t0 and t2 (t0 < t1 < t2), and the corresponding physical quantity feature values are x0 and x2. Then the interpolated physical quantity feature value x1 at time t1 can be calculated by the following formula: , in this way, the CAN bus data value aligned with the timestamp of the satellite positioning data can be obtained.

[0050] In a scenario, an operating vehicle is driving on a section of the road. The satellite positioning data shows that the vehicle's position is at longitude 116.54°, latitude 39.87°, altitude 22 meters, vehicle speed 110 km / h, and heading angle 120°; the CAN bus collects the brake pedal state for 4 seconds (sampling rate 20 Hz); then search the surrounding road network centered on the vehicle's position to determine the lane information, and the high-precision map returns the speed limit of 120 km / h and slope of 1% for this section of the road. During the data fusion process, the positioning data and CAN signals will be aligned by interpolation; finally, a data fusion vector {longitude 116.54°, latitude 39.87°, altitude 22 meters, speed: 110 km / h, heading angle: 120°, brake duration: 4 s, road speed limit: 120 km / h, slope: 1%} is generated.

[0051] S2. Set the threshold model rules and monitor the calculation; In this method, the setting of the threshold model rules is a key step based on comprehensive consideration of multi-source information, aiming to ensure the accurate identification of bad driving behaviors while taking into account the complexity and particularity of different driving scenarios. It specifically includes the following sub-steps:

[0052] S201: Set the behavior threshold; For different bad driving behaviors during driving, set corresponding behavior thresholds based on driving safety standards. For example, set the following behavior thresholds: a brake state duration ≥ 10 s is determined as long-term braking, a continuous driving time ≥ 4 hours is determined as fatigue driving, and a deceleration > 2 m / s² is determined as emergency braking.

[0053] The above threshold ranges are for illustration purposes only. In actual situations, the thresholds should be set based on local traffic regulations, a large amount of driving data and accident case analysis, and driving safety standards. Thresholds set based on regulations and a large amount of data can accurately identify various bad driving behaviors, provide a reliable basis for subsequent monitoring and early warning, and provide a judgment standard for the subsequent identification of dangerous driving behaviors. Only when the data of the vehicle's driving behavior exceeds the corresponding threshold will it be judged as dangerous driving behavior.

[0054] S202: Combined time series analysis; Time series analysis is a method used to analyze time series data, identifying patterns and trends within the data. In this method, it is used to identify continuous dangerous driving behaviors. The core approach is to aggregate discrete events into systemic risk patterns along the temporal dimension. In specific implementation, vehicle driving data is collected in real time via satellite positioning devices and CAN bus interfaces. The collected data undergoes preliminary cleaning and preprocessing to remove noise and outliers, ensuring data accuracy and completeness. The preprocessed data is then arranged in chronological order to form a time series dataset. Each data point contains information about the vehicle's driving state at a specific moment, such as speed, acceleration, and braking status. A timestamp is added to each data point to accurately record the time of data collection. An appropriate time window length, such as 5 or 10 minutes, is selected based on different driving behaviors and business requirements. Within each time window, statistical features of the vehicle driving data are calculated, such as the number of sudden accelerations and braking, the duration of speeding, maximum acceleration, maximum deceleration, and average speed. A sliding window approach is used to perform feature calculations on each window of the time series data. After each sliding window movement, the statistical feature value within the current window is recalculated and compared with a pre-set behavior threshold. If the statistical feature value exceeds the corresponding threshold, the corresponding dangerous driving behavior is determined to have occurred within the time window. This behavior is marked and the type, occurrence time, and duration of the behavior are recorded. At the same time, the corresponding warning mechanism is triggered to remind the driver to pay attention to driving safety.

[0055] Taking fatigue driving as an example, time series analysis can be used to detect driver fatigue. Consider a commercial vehicle traveling on a highway, with its travel time monitored in real time. When the vehicle's travel time approaches four hours, time series analysis is combined with a focus on analyzing the vehicle's driving data over the next 10 minutes. If the vehicle's speed is relatively stable but with continuous minor speed fluctuations and directional deviations, the driver is considered likely fatigued, triggering a fatigue driving warning and prompting the driver to take a break.

[0056] Time series analysis can aggregate discrete driving behavior events over time, identifying patterns of continuous dangerous driving behavior over a period of time. For example, multiple instances of rapid acceleration and braking within a short period of time can be detected, potentially indicating aggressive driving habits, while single instances of rapid acceleration or braking may simply be isolated incidents. It can also effectively distinguish between occasional driver actions and habitual dangerous driving behaviors, avoiding misjudgments caused by interference from momentary road conditions. For example, a driver may brake suddenly during an emergency evasive maneuver, but this does not necessarily indicate dangerous driving habits. Time series analysis comprehensively considers a driver's behavior over a period of time, improving the accuracy and comprehensiveness of dangerous driving behavior identification.

[0057] S203: Enhance determination accuracy by combining map data; To more accurately judge driving behavior, it's necessary to combine map data with road information to enhance accurate judgments. For example, map data provides speed limits for roads, and the vehicle's actual speed is compared against these limits to determine whether the driver is speeding. Map data also identifies road construction areas and special sections (such as school areas, hospital areas, and accident-prone areas). When a vehicle enters these areas, it issues early warnings, prompting drivers to exercise caution and monitoring. Traffic sign and signal light information provided by the map data, combined with the vehicle's actual driving behavior and location, can be used to determine whether the driver is complying with traffic regulations. By leveraging map data to accurately capture key road information, the system enables more precise judgments of dangerous driving behavior.

[0058] The following is an example: Suppose a commercial vehicle is driving on an urban road with a speed limit of 50km / h. There is a school next to the road, and the speed limit in the school area is 30km / h. At 100 meters away from the school area, the map data sends an early warning message to the vehicle system to remind the driver that he is about to enter the school area. When the vehicle enters the school area, if the vehicle speed exceeds 30km / h, it is immediately judged as speeding and a warning is triggered to remind the driver to slow down. If there is a traffic sign indicating to stop and give way on this road section, and the vehicle passes directly without stopping as required, the driver is judged to have violated the traffic rules, the violation is recorded, and the driver is reminded. In this process, the map data provides the road speed limit, the location of the school area, and the traffic sign information. Combined with the actual driving speed and location information of the vehicle, it can accurately judge the dangerous driving behaviors of speeding and violating traffic rules.

[0059] S204: Develop real-time monitoring algorithm engine; The real-time monitoring algorithm engine is a system module capable of real-time data processing and analysis. It was developed to process uploaded vehicle data as a data stream and determine in real time whether adverse driving behavior thresholds have been triggered. For example, upon receiving vehicle speed data, acceleration data, braking status data, and road speed limit data, it is immediately compared with the set thresholds. Similarly, to identify sudden acceleration and braking, it is necessary to monitor acceleration data changes in real time and compare them with the set thresholds.

[0060] Specifically, based on the threshold model rules for driving behavior established in the above steps, corresponding algorithmic logic is developed to analyze data in real time. The fused data from step S104 is then integrated into the algorithm engine, tested in actual driving scenarios, and optimized based on the results. Setting the threshold model rules involves preliminarily setting thresholds for various types of undesirable driving behaviors based on regulations, safety standards, and data analysis results. Time series analysis is then used to determine the time window length and statistical characteristics for identifying continuous patterns of dangerous driving behavior. Furthermore, map data is combined with driving behavior data to determine how to leverage road attribute information (such as speed limits and special zones) to adjust thresholds and enhance judgment accuracy. Specifically, using the single-scenario threshold from step S201 as the minimum basic judgment unit, discrete events at the base layer are aggregated through the time window from step S202. The base thresholds are then dynamically adjusted (e.g., speed limits are reduced in school zones) using the map road network data from step S203, ensuring that the judgment criteria for the same behavior in different scenarios are tailored to the actual risk.

[0061] Finally, the real-time stream processing framework Apache Flink was used to process the data stream. During the stream processing process, a sliding window function was defined to aggregate and analyze the data within a time range. The sliding window time granularity was set to 5 minutes, with a sliding step size of 1 minute. Within the window, the vehicle's driving data was aggregated and calculated to extract statistical features, such as the number of occurrences of sudden acceleration, sudden deceleration, and speeding. By comparing these statistical features with preset thresholds, it was determined whether the driver had a continuous pattern of dangerous driving behavior. For example, for sudden acceleration or braking, the acceleration changes within 5 minutes were monitored in real time and compared with the set sudden acceleration threshold or sudden braking threshold to determine the number of sudden acceleration or braking behaviors within 5 minutes.

[0062] S3. Real-time monitoring, calculation, and identification of dangerous driving behaviors. This includes the following sub-steps:

[0063] S301: Rule engine deployment; Install the Drools rule engine on the vehicle monitoring system's backend server. Configure the rule engine's runtime environment, including setting the rule file loading path and logging level.

[0064] The Drools rule engine executes rules based on priority, detecting single dangerous behaviors in real time, analyzing continuous behaviors using a sliding window, and determining dangerous driving behaviors by combining map and road network data. A data transmission channel is established to input the fused vehicle data from S104 into the Drools rule engine in real time. Data objects are inserted into the rule engine's working memory through the Drools API. The rule engine performs pattern matching on the data inserted into the working memory according to pre-set rules. When the data meets the rule conditions, the rule engine executes the corresponding action, such as generating an event record or triggering an alert.

[0065] To deploy the rule engine, specifically, you must first define rules for various types of dangerous driving behaviors in the Drools rule engine, including conditions and actions. In the rule conditions, by combining the road segment attributes of the map data and accessing the map information in the fused data, the rules can dynamically adjust the judgment conditions based on different road conditions and environments. For example, in dangerous scenarios such as heavy rain or icy roads, the speeding threshold can be lowered, or the emergency braking threshold can be relaxed on downhill sections. In addition, each rule is assigned a priority based on business needs and the degree of danger. Use the salience property of the Drools rule engine to set the execution priority of the rule. This ensures that when multiple rule conditions are met simultaneously, the high-priority rule is executed first.

[0066] For example, in other dangerous scenarios like heavy rain, icy roads, or heavy fog, the condition for the "Speeding" rule is "Vehicle speed > road section speed limit x 60%," with the action "Triggering a speeding warning." The condition for the "Aggressive Driving" rule is "Consecutive sudden braking greater than or equal to three times within five minutes," with the action "Triggering an aggressive driving warning." The execution priority of the rules is then set based on business needs, and the integrated vehicle data is fed into the Drools rule engine in real time. Finally, the Drools engine matches the input data against pre-set rules. Once the data meets the rule conditions, the actions of recording the event and triggering an alert are immediately executed. The following is an example of the content of a Drools rule file: / / Rule 1: Speeding in hazardous weather rule "OverSpeed_DangerousWeather" salience 10 / / Execution priority when $f: Frame( speed>roadLimit * 0.6, / / speed > 60% of the speed limit $ts: timestamp, $weather: weatherCondition in ("heavyRain", "snow", "fog") / / Judge weather conditions ) then insert(new OverSpeedEvent($weather, $ts, $f.speed)); / / Action: Generate an overspeed event end / / Rule 2: Speeding in Normal Weather rule "OverSpeed_NormalWeather" salience 7 when $f: Frame( speed>roadLimit, $ts: timestamp, $weather: weatherCondition not in ("heavyRain", "snow", "fog") ) then insert(new OverSpeedEvent($weather, $ts, $f.speed)); end / / Rule 3: Aggressive Driving rule "HardBraking_Sequence" salience 5 when $window: SlidingWindow(length: 5m) / / 5-minute sliding window $brakeList: List(size>= 3) from collect( Frame(deceleration>2) over $window ) then insert(new AggressiveDrivingEvent("Aggressive Driving", $window.getStartTime())); end Consider a scenario where a commercial truck is traveling at 57 km / h on a road with a 100 km / h speed limit. Within five minutes, four decelerations are detected: 2.1 m / s², 2.3 m / s², 2.3 m / s², and 2.2 m / s². Map data indicates that the road is experiencing heavy rain. The vehicle's real-time speed, location, speed limit, and deceleration data are fused and fed into the Drools engine. Based on pre-set rules, the engine first matches the higher-priority "Speeding" rule. Finding that the speed of 57 km / h does not exceed 60% of the speed limit (i.e., 60 km / h), the triggering condition for the "Speeding" rule is not met, and no action is taken. The engine then matches the "Aggressive Driving" rule. Since the deceleration reached -3 m / s² or less four times within five minutes, indicating four sudden braking events, the "Aggressive Driving" rule's triggering condition is met. The engine immediately inserts a sudden braking event and generates a corresponding warning.

[0067] S302: Determination and early warning; The judgment and warning mechanism is a core function of this method, designed to promptly alert drivers to correct undesirable driving behaviors and fully record the incident for subsequent analysis. When the rules engine determines that dangerous driving behavior has occurred, an event record is immediately created. This record contains key information such as the event type, time of occurrence, unique vehicle identifier, and current location. Once the event record is created, a specific warning message is delivered to the driver via the vehicle's 5G communication module via voice broadcast. For example, a warning message stating that the current speed has exceeded the speed limit and that the driver should immediately reduce speed is required. Vehicle data is then continuously monitored in real time to determine whether the dangerous behavior has been resolved, for example, by whether the speed has dropped below the speed limit. Simultaneously, the warning message is sent to the fleet management center platform, which notifies fleet managers via a push notification service. If the dangerous behavior persists or an anomaly occurs (such as signal loss, equipment failure, or network interruption), monitoring is terminated and the complete event record is pushed to the management platform to ensure a complete record of all events, providing data support for subsequent analysis. Whether the event terminates normally or abnormally, the complete dangerous driving event is pushed to the platform for storage.

[0068] For example, upon receiving a speeding warning and determining that a vehicle is speeding, a speeding event record is immediately created and sent to the fleet management center platform. The system then continuously monitors the vehicle's speed, combining driving data acquired via the vehicle's CAN bus. If the driver slows down to below the speed limit, the dangerous behavior is considered resolved and the end time of the event is recorded. If the driver does not slow down, the warning continues. If an anomaly such as signal loss occurs, monitoring is terminated and the complete event record is pushed to the platform.

[0069] S303: Cycle monitoring; A cyclic monitoring mechanism ensures continuous monitoring of vehicle driving status to avoid missing any potentially dangerous driving behaviors. A timer is set in the monitoring system to periodically (as needed, such as every 5 seconds, every 10 seconds, etc.) check whether there are currently any dangerous driving behavior monitoring tasks. Each monitoring task is assigned a status mark indicating its execution status (in progress, completed, paused). During each check, the task status mark is used to determine whether a similar monitoring task is currently in progress. If no similar monitoring task is currently in progress, the monitoring system automatically restarts monitoring according to the rules set by the rule engine. When restarting monitoring, the vehicle and map data required for monitoring are updated to ensure the accuracy and timeliness of monitoring.

[0070] For example, after a speeding incident involving a commercial vehicle ends, the monitoring system detects that no other similar dangerous driving behaviors are currently being monitored. It then restarts monitoring according to the rules set by the rule engine. At this point, it continues to monitor the vehicle's driving data in real time, obtaining the vehicle's latest location and speed information, as well as the latest map data for the current road section. The system then continues to monitor the vehicle's driving data according to the rules set by the rule engine to identify new dangerous driving behaviors.

[0071] The loop monitoring mechanism ensures the monitoring system continuously monitors the vehicle's driving status, preventing any potential dangerous driving behavior from being missed. Even if dangerous driving behavior disappears briefly and then reappears during driving, it can be detected and triggered in a timely manner, providing uninterrupted safety protection.

[0072] S4. Real-time warning and correction of the driver. This includes the following sub-steps:

[0073] S401: Push warning information; After identifying a dangerous driving incident, the platform retrieves the corresponding event record created in S3 and pushes it to the platform. The platform then configures the corresponding script tags and content, extracting the corresponding script tags and content from the configured script library. The script library is a table stored in a database that associates various types of dangerous driving behaviors with corresponding voice prompts. For example, the script for "speeding" is set to "You are speeding, please reduce your speed," and the script for "aggressive driving" is set to "Please control your speed, avoid sudden acceleration and braking, and maintain safe driving." The platform then provides voice reminders to the driver based on the script content, correcting poor driving behavior and guiding them to drive safely.

[0074] At the same time, warning information is sent to the fleet management center platform via a wireless 5G communication network. Warning information is transmitted in a specific JSON format, containing key data fields such as event type, vehicle information, and geographic location, allowing the platform to accurately parse and display this information.

[0075] For example, if a commercial vehicle's speed exceeds a threshold for 10 consecutive seconds, the rules engine identifies it as "speeding," triggering an alert and sending a voice prompt to the driver via the vehicle's terminal: "You are speeding, please reduce your speed." Simultaneously, the alert is sent to the fleet management center, allowing managers to promptly understand the situation and take appropriate measures. For another example, if the rules engine identifies a commercial vehicle driver as engaging in "aggressive driving," the platform configures a corresponding callout tag, "Aggressive Driving Reminder," with the following message: "Please control your speed, avoid sudden acceleration and braking, and maintain safe driving."

[0076] The platform broadcasts reminders in voice according to the set script, which can quickly attract the driver's attention. The preset script provides the driver with corresponding driving suggestions, allowing him to understand the current dangerous driving behavior in time and correct it, effectively reducing the risk of accidents.

[0077] S402: Notify the fleet manager; The platform simultaneously sends a warning message to fleet managers. Upon receiving the warning, the information is immediately displayed on the management interface. The interface presents the information in both a map and a list format. The map identifies the location of the vehicle involved in the dangerous driving behavior, while the list details the type of incident, vehicle information, and time of occurrence, allowing managers to quickly understand the situation. The platform also sends a notification message to managers, alerting them to new warning events requiring attention. The notification includes basic information about the incident and the vehicle's ID, ensuring timely attention.

[0078] The manager uses the monitoring system to view the vehicle's real-time driving status and historical driving data. Combining these data, the manager verifies the incident's authenticity and marks the verification result in the handling report. If necessary, the manager can communicate directly with the driver through the in-vehicle intercom system. The in-vehicle intercom system utilizes a wireless communication network to enable two-way voice communication between the manager and the driver. The manager can provide immediate guidance and corrective advice to the driver, emphasize the importance of safe driving, and request that the driver immediately correct any undesirable driving behavior.

[0079] For example, when a driver of a commercial vehicle received an "aggressive driving warning" voice message, the fleet manager also received the warning. The manager checked the vehicle's driving status in real time through the monitoring system and found that the driver was indeed frequently accelerating and braking suddenly. The manager marked the result of the action report as confirmed and then communicated with the driver through the in-vehicle intercom system, emphasizing the importance of safe driving and requiring immediate correction.

[0080] S5. Record and manage dangerous driving incidents. This includes the following sub-steps:

[0081] S501: Recording and statistics; For the determined dangerous driving events, the fleet manager obtains the corresponding event records created in S3 and performs periodic statistics based on the recorded dangerous driving events of the drivers.

[0082] In one scenario, a monthly statistical analysis of dangerous driving incidents by a commercial vehicle driver was performed. In one month, Driver A exceeded the speed limit five times (for a total of 165 seconds), engaged in sudden braking three times, and engaged in two aggressive driving behaviors. This recording and statistical analysis provided data support for driver performance evaluation and training, helping fleet managers understand driver behavior improvement progress.

[0083] S502: Motivation and guidance; Based on the recorded and statistical data on dangerous driving incidents, a driver ranking assessment is conducted to determine the number of dangerous driving behaviors. Month-over-month comparisons are also conducted for each driver, comparing the number of dangerous driving behaviors this month with the number of behaviors last month. Drivers who improve their bad driving behavior can be rewarded to encourage and guide safe driving, forming a closed-loop management mechanism of "monitoring-warning-rectification."

[0084] In one scenario, according to statistical data, in a certain area, driver A had the fewest dangerous driving incidents this month, ranking first among all drivers in the area. The fleet manager gave driver A certain rewards, such as issuing a "Safe Driving Model" certificate and additional bonus performance, to encourage him to continue to maintain good driving habits. A month-on-month comparison found that driver B's dangerous driving behavior this month decreased by 40% compared with the previous month. The fleet manager gave him a certain bonus to commend his significant progress. At the same time, for those drivers who had more dangerous driving behaviors, targeted safety training courses will be pushed to guide them to improve their driving behavior.

[0085] Through incentive and guidance measures, drivers' awareness of safe driving can be effectively improved, a good management closed loop can be formed, and the overall safety risk of the fleet can be continuously reduced.

[0086] Through the above steps, the present invention realizes a method for automatically monitoring and identifying dangerous driving behaviors based on driving scenarios, reducing the safety risks of commercial vehicles, improving driving safety, and reducing losses caused by driving safety accidents.

[0087] The present invention has been described with reference to the above embodiments. However, the above embodiments are merely exemplary embodiments of the present invention. It should be noted that the disclosed embodiments do not limit the scope of the present invention. On the contrary, modifications and improvements that do not depart from the spirit and scope of the present invention are intended to be protected by the present invention.

Claims

1. A method for automatically monitoring and identifying dangerous driving behaviors based on driving scenarios, characterized in that: The following steps are involved: S1: Data acquisition and fusion; S2: Set threshold model rules and monitor calculations; S3: Real-time monitoring, calculation and identification of dangerous driving behaviors; S4: Real-time warning and correction of the driver; S5: Recording and management of dangerous driving incidents by drivers.

2. The method for automatically monitoring and identifying dangerous driving behaviors based on driving scenarios according to claim 1, characterized in that: The step S1 includes the following sub-steps: S101: Install satellite positioning device; S102: Collect CAN bus data; S103: Deploy high-precision map road network data; S104: Data fusion; The data fusion in step S104 specifically involves acquiring satellite positioning data, fusing it with map road network data, and associating it with CAN signal data to generate a fused data frame containing relevant information about the vehicle itself and its location.

3. The method for automatically monitoring and identifying dangerous driving behaviors based on driving scenarios according to claim 1, characterized in that: The step S2 includes the following sub-steps: S201: Setting behavior threshold; S202: Combined time series analysis; S203: Enhance determination accuracy by combining map data; S204: Develop real-time monitoring algorithm engine; Step S204: Develop a real-time monitoring algorithm engine, set threshold model rules based on the above steps, and use a real-time stream processing framework to process the uploaded data of each vehicle in real time to determine whether the bad driving behavior threshold is triggered in real time.

4. The method for automatically monitoring and identifying dangerous driving behaviors based on driving scenarios according to claim 1, characterized in that: The step S3 includes the following sub-steps: S301: Rule engine deployment; S302: Determination and early warning; S303: Cycle monitoring; In step S301, the rule engine is deployed using the Drools rule engine to define rules for various types of dangerous driving behaviors, including conditions and actions, and to set a priority for each rule through the salience attribute.

5. The method for automatically monitoring and identifying dangerous driving behaviors based on driving scenarios according to claim 4, characterized in that: In the judgment and warning of step S302, regardless of whether the process of monitoring dangerous driving behavior ends normally or abnormally, the complete dangerous driving event is pushed to the platform for record and storage. Specifically, after the dangerous driving behavior is determined, a warning reminder is started and a dangerous driving event record is immediately created and sent to the fleet management center platform. The dangerous driving behavior is continuously monitored and whether the dangerous driving behavior has been resolved is determined in combination with the driving data. When the dangerous driving behavior cannot be continuously monitored or an abnormal situation occurs, the dangerous driving behavior monitoring is forcibly terminated and the complete event record is pushed to the platform for storage.

6. The method for automatically monitoring and identifying dangerous driving behaviors based on driving scenarios according to claim 1, characterized in that: Step S4 includes the following sub-steps: S401: Push warning information; S402: Notify the fleet management personnel; In step S402, after receiving the warning information, the manager checks it on the management interface and determines the dangerous driving behavior; If necessary, the manager needs to communicate directly with the driver through the in-vehicle intercom system, provide the driver with immediate guidance and corrective suggestions, and require the driver to correct bad driving behavior immediately.

7. The method for automatically monitoring and identifying dangerous driving behaviors based on driving scenarios according to claim 1, characterized in that: Step S5 includes the following sub-steps: S501: Recording and statistics; S502: Motivation and guidance; In step S502, the incentive and guidance are specifically provided based on the ranking of the number of dangerous driving behaviors and the improvement of the driver.

Images