An active safety supervision method and system for intelligent networked operating vehicles

Abstract

The application discloses an active safety supervision method and system for intelligent network-connected operating vehicles, which obtains vehicle dynamics, operating business, services and external multi-source data; through intelligent analysis and early warning engine, comprehensive scoring, multi-scale digital twin is combined to complete early warning verification and regulation information generation; after being presented through a visual interface, the risk level is determined, early warning and the optimal processing mode are pushed, cross-scale linkage supervision is realized, the false alarm and missed alarm rate is reduced, the supervision foresight and effectiveness are improved, and the method is suitable for various operating scenes.

Description

Technical Field

[0001] This invention belongs to the field of intelligent connected vehicle supervision technology, specifically relating to a method and system for supervising the operational safety of intelligent connected commercial vehicles. Background Technology

[0002] In the field of connected vehicles, when vehicles are transitioning from testing and demonstration to formal commercial operation, regulators typically rely on video surveillance and manual patrols, GPS-based trajectory monitoring, post-event data statistics and reporting, and vehicle status and fault code monitoring to ensure the operational safety of vehicles in real and complex road environments.

[0003] Currently, vehicle monitoring suffers from the following problems: Continuous manual review and interpretation of video surveillance footage and vehicle trajectories is not only inefficient but also highly susceptible to missed critical anomalies due to staff fatigue. This means that most problems are only discovered after serious consequences have occurred, completely failing to meet the modern regulatory needs for early warning and proactive intervention. Furthermore, real-time vehicle operation data, daily business operation data, and service quality data are stored in different information systems, creating severe data silos. This makes it difficult to identify complex operational safety risks that cross systems and dimensions.

[0004] The existing regulatory system primarily focuses on basic parameters such as speed and location, failing to establish effective correlations with the core business characteristics of different operating vehicles. There is a lack of dedicated monitoring mechanisms and evaluation standards for key business indicators that can truly reflect operational status, such as real-time passenger volume of buses and delivery vehicle task completion rates.

[0005] Therefore, how to build a data fusion and intelligent analysis system that can adapt to various operating scenarios, break through the technical limitations of the existing regulatory model, and achieve proactive discovery, accurate identification, and scientific quantification of operational safety risks of intelligent connected vehicles has become one of the key issues that urgently need to be addressed in the field of vehicle networking. Summary of the Invention

[0006] To address the above problems, the technical solution adopted by the present invention is as follows:

[0007] A proactive safety supervision method for intelligent connected commercial vehicles, comprising the following steps:

[0008] S1. Obtain vehicle dynamic data;

[0009] Data collected via vehicle-mounted terminals or data interfaces includes vehicle CAN bus data, GPS location information, autonomous driving system control modes, and system fault codes.

[0010] S2. Obtain vehicle operation business data, determine the vehicle type, and collect data based on the vehicle route coverage and target area.

[0011] S3. Obtain vehicle service data and external data; based on the enterprise customer service system, obtain vehicle service data including valid complaint records and handling process information for vehicles or operational tasks; obtain vehicle traffic violation data through external data interfaces.

[0012] S4. Based on the intelligent analysis and early warning engine, perform data analysis and data mining on vehicle dynamic data, vehicle operation business data, vehicle service data and external data, and obtain analysis results based on business scenarios.

[0013] S5. Input the analysis results into the regulatory visualization interface to obtain the vehicle situation diagram and traffic heat map;

[0014] The situational map includes buses and delivery vehicles distinguished by different colors and icons on the map, and vehicles in warning and alarm states are highlighted; the traffic heat map shows the traffic status of the road network within the target area.

[0015] S6. Determine the vehicle risk level based on the vehicle situation map and traffic heat map, and push early warning information and a list of optimal handling methods to vehicles whose risk level exceeds the threshold.

[0016] Furthermore, CAN bus data includes vehicle speed, acceleration, steering angle, braking status, and door or cargo box door opening / closing status signals.

[0017] Furthermore, vehicle operation data is acquired, and after determining the vehicle type, data is collected based on the vehicle's route coverage and target area. This includes: if the operating vehicle type is a bus, then the bus dispatch system is used to obtain the schedule, planned route, station timetable and actual arrival time, and passenger order data; if the operating vehicle type is a delivery vehicle, then the logistics and distribution system is used to obtain the delivery task list, planned route, and planned delivery time, and parcel entry and exit data are collected simultaneously from the vehicle's barcode scanning device or cargo box sensor.

[0018] Furthermore, the operational tasks include one or more buses or delivery vehicles.

[0019] Furthermore, the data analysis and data mining include:

[0020] S41. Obtain a basic vehicle operation safety score based on the frequency of system takeover requests, the number of times of rapid acceleration and deceleration, and the number of traffic rule violations.

[0021] S42. Calculate the business efficiency score based on the vehicle operation type.

[0022] S421. If the vehicle operation type is a bus, the first score for business efficiency is calculated based on the real-time passenger volume, the planned trip completion rate, and the station punctuality rate.

[0023] S422. If the vehicle operation type is a courier delivery vehicle, the real-time task completion rate, the number of delayed packages, and the average delivery time per package are used to calculate the second score of business efficiency.

[0024] S43. Calculate the service quality score based on the service complaint rate and the number of abnormal service interaction events as statistically analyzed for vehicles or operational tasks.

[0025] S44. For any vehicle, a comprehensive score analysis is performed on the vehicle's basic operational safety score, business efficiency score, and service quality score; if the comprehensive score is lower than the first threshold, a multi-condition capability judgment of core business is performed.

[0026] S45. Analyze early warning information based on multi-scale digital twins of vehicles for different services;

[0027] S46. Verify the first or second warning information based on the vehicle-level digital twin, including:

[0028] S461. Obtain the real-time vehicle parameters corresponding to the first or second early warning information and inject them into the vehicle-scale digital twin. Compare the real-time values ​​of the early warning parameters, such as the braking acceleration of a bus, with the predicted values ​​of the vehicle-scale digital twin. If the residual between the two exceeds the fifth threshold, the verification passes and the verification early warning is triggered.

[0029] S462. Generate active vehicle situation control information based on the vehicle-level digital twin and verification and early warning information;

[0030] S463. Generate road network guidance information based on the road network-level digital twin and verification and early warning information.

[0031] Furthermore, the assessment of core business capabilities under multiple conditions includes:

[0032] S4411. If the vehicle's operating type is a delivery vehicle, then check whether the vehicle's task completion rate is consistently higher than the preset second threshold, and then continue monitoring.

[0033] S4412. When the vehicle task completion rate is not higher than the preset second threshold, check whether the system takeover frequency is abnormal; if the system takeover frequency in the current statistical period is less than 1.5 times the average frequency of the fleet during the same period, continue to monitor.

[0034] S4413. If the system takeover frequency in the current statistical period is not less than 1.5 times the average frequency of the fleet during the same period, then check whether the number of package delay complaints has increased; if the number of valid complaints about vehicles in the current statistical period is lower than the preset absolute number threshold, then continue monitoring.

[0035] S4414. If the number of valid complaints against vehicles is not lower than the preset absolute number threshold during the current statistical period, then the multi-condition capability warning for the core business of express delivery vehicles will be triggered, forming the first warning information.

[0036] Furthermore, the assessment of core business multi-condition capabilities also includes:

[0037] S4421. If the vehicle's operating type is a bus, then check the difference between the vehicle's real-time passenger load and the average historical passenger load at the current moment. If the absolute value of the difference is less than the third threshold, then continue monitoring.

[0038] S4422. If the absolute value of the difference between the real-time passenger volume and the historical average passenger volume at the current time is not lower than the third threshold, then the vehicle station accuracy rate is determined. If the vehicle station accuracy rate is higher than the fourth threshold, then monitoring continues.

[0039] S4423. If the vehicle station accuracy rate is not higher than the fourth threshold, check whether the number of complaints about the operation task has increased. If the number of valid complaints about vehicles is lower than the preset absolute number threshold in the current statistical period, continue to monitor.

[0040] S4424. If the number of valid complaints about vehicles is not lower than the preset absolute number threshold during the current statistical period, then the multi-condition capability warning for the core business of public buses will be triggered, forming a second warning message.

[0041] An active safety monitoring system for intelligent connected commercial vehicles, used to implement an active safety monitoring method for intelligent connected commercial vehicles, the system comprising:

[0042] The vehicle data acquisition module is used to acquire vehicle dynamic data; it is acquired through an on-board terminal or data interface, including vehicle CAN bus data, GPS location information, autonomous driving system control mode, and system fault codes.

[0043] It is also used to obtain vehicle operation business data, and after determining the vehicle type, it collects data based on the vehicle route coverage and target area range.

[0044] It is also used to obtain vehicle service data and external data; to obtain vehicle service data based on the enterprise customer service system, including valid complaint records and handling information for vehicles or operational tasks; and to obtain vehicle traffic violation data based on external data interfaces.

[0045] The intelligent analysis and early warning engine is used to perform data analysis and data mining on vehicle dynamic data, vehicle operation business data, vehicle service data and external data, and obtain analysis results based on business scenarios.

[0046] The data display module is used to input the analysis results into the regulatory visualization interface to obtain vehicle situation maps and traffic heat maps. The situation maps include buses and delivery vehicles distinguished by different colors and icons on the map, and vehicles in the warning and alarm states are highlighted. The traffic heat map shows the traffic status of the road network within the target area.

[0047] The risk alert module is used to determine the risk level of a vehicle based on the vehicle situation map and traffic heat map, and push warning information and a list of optimal handling methods to vehicles whose risk level exceeds the threshold.

[0048] A computer-readable storage medium storing a computer program, characterized in that a processor executes the computer program to implement a proactive safety monitoring method for intelligent connected vehicles.

[0049] A terminal device includes a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that the processor executes the computer program to implement an active safety monitoring method for intelligent connected vehicles.

[0050] The beneficial effects of this invention are as follows:

[0051] 1) This solution adopts a multi-level early warning mechanism that combines comprehensive scoring screening with multi-scale twin verification. It initially screens high-risk vehicles through basic safety, business efficiency, and service quality scores, and designs differentiated verification rules for the business characteristics of buses and delivery vehicles. Finally, it uses vehicle-level, fleet-level, and road network-level twins to verify the authenticity of the early warning, effectively reducing false alarms and missed alarms. After the early warning, control and guidance information can be directly output. At the same time, it realizes a fundamental change in the regulatory model. Through a configurable intelligent rule engine, it performs real-time correlation analysis of multi-source data, which can proactively issue early warnings in the early stages of problems, significantly improving the foresight and effectiveness of supervision.

[0052] 2) This solution differentiates between two types of operating vehicles, namely buses and express delivery vehicles, and collects corresponding business data such as bus dispatch data, express delivery lists, and parcel entry and exit data in a differentiated manner. At the same time, it integrates vehicle dynamics, service, and external violation data to achieve full coverage of multi-dimensional data of "vehicle-business-environment". This avoids redundancy and omissions caused by generalized collection and lays the foundation for subsequent accurate analysis. It can also effectively discover deep-seated operational safety issues that cannot be revealed by single-dimensional data.

[0053] 3) This solution constructs a three-level digital twin of vehicles, fleets, and road networks to achieve cross-scale linkage of micro-level vehicle state inversion, meso-level fleet situation control, and macro-level road network anomaly detection. At the same time, through reinforcement learning, RLHF fine-tuning, and OTA distribution, it can adapt to dynamic changes such as vehicle aging and road network evolution, and improve the adaptability and effectiveness of supervision in the long term.

[0054] The above description is merely an overview of the technical solution of the present invention. In order to better understand the technical means of the present invention, it can be implemented according to the contents of the specification. In order to make the above description and other objects, features and advantages of the present invention more easily understood, preferred embodiments are provided and described in detail below. Attached Figure Description

[0055] Various other advantages and benefits will become apparent to those skilled in the art upon reading the following detailed description of preferred embodiments. The accompanying drawings are for illustrative purposes only and are not intended to limit the invention. Furthermore, the same reference numerals denote the same parts throughout the drawings.

[0056] Figure 1 A flowchart for a proactive safety supervision method for intelligent connected commercial vehicles.

[0057] Figure 2 This is a structural diagram of an active safety monitoring system for intelligent connected vehicles. Detailed Implementation

[0058] Exemplary embodiments of the present disclosure will now be described in more detail with reference to the accompanying drawings. While exemplary embodiments of the present disclosure are shown in the drawings, it should be understood that the present disclosure may be implemented in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided so that this disclosure will be thorough and complete, and will fully convey the scope of the disclosure to those skilled in the art.

[0059] In the description of this invention, unless otherwise explicitly specified and limited, the terms "installation," "connection," "joining," "fixing," etc., should be interpreted broadly. For example, they can refer to a connection, a detachable connection, or an integral part; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; they can refer to the internal communication of two components or the interaction between two components. Those skilled in the art can understand the specific meaning of the above terms in this invention according to the specific circumstances. Example

[0060] like Figure 1 As shown, a proactive safety supervision method for intelligent connected commercial vehicles includes the following steps:

[0061] S1. Obtain vehicle dynamic data;

[0062] Data is collected through the vehicle terminal or data interface, including vehicle CAN bus data, GPS location information, autonomous driving system control mode, and system fault codes; among which, CAN bus data includes vehicle speed, acceleration, steering angle, braking status, and door or cargo box door opening and closing status signals.

[0063] S2. Obtain vehicle operation business data, determine the vehicle type, and collect data based on the vehicle route coverage and target area.

[0064] Specifically: if the operating vehicle type is a bus, the system obtains the scheduling plan, the planned route, the station timetable and the actual arrival time, and passenger order data from the bus dispatch system; if the operating vehicle type is a delivery vehicle, the system obtains the delivery task list, the planned route and the planned delivery time from the logistics and delivery system, and simultaneously collects package entry and exit data from the on-board scanning device or cargo box sensor.

[0065] S3. Obtain vehicle service data and external data; based on the enterprise's customer service system, obtain vehicle service data including valid complaint records and handling process information for vehicles or operational tasks; where operational tasks include one or more buses or delivery vehicles; obtain vehicle traffic violation data through external data interfaces;

[0066] S4. Based on the intelligent analysis and early warning engine, perform data analysis and data mining on vehicle dynamic data, vehicle operation business data, vehicle service data and external data, and obtain analysis results based on business scenarios.

[0067] The data analysis and data mining include:

[0068] S41. Obtain a basic vehicle operation safety score based on the frequency of system takeover requests, the number of times of rapid acceleration and deceleration, and the number of traffic rule violations.

[0069] S42. Calculate the business efficiency score based on the vehicle operation type.

[0070] S421. If the vehicle operation type is a bus, the first score for business efficiency is calculated based on the real-time passenger volume, the planned trip completion rate, and the station punctuality rate.

[0071] S422. If the vehicle operation type is a courier delivery vehicle, the real-time task completion rate, the number of delayed packages, and the average delivery time per package are used to calculate the second score of business efficiency.

[0072] S43. Calculate the service quality score based on the service complaint rate and the number of abnormal service interaction events as statistically analyzed for vehicles or operational tasks.

[0073] The vehicle basic operation safety score, business efficiency score, and service quality score are included in the comprehensive score calculation, which includes but is not limited to weighted summation.

[0074] S44. Conduct a comprehensive scoring analysis of the vehicle's basic operational safety score, business efficiency score, and service quality score for any given vehicle.

[0075] If the overall score is lower than the first threshold, such as 60 points, then a multi-condition capability assessment of core business will be conducted.

[0076] According to the industry standard for safe operation of intelligent connected vehicles, if the comprehensive score is less than 60 points, at least one of the vehicle operation safety, business efficiency and service quality is at a substandard level, and multi-condition verification needs to be initiated.

[0077] For example:

[0078] In S41, a standardized deduction system is adopted, first setting a basic safety maximum score (denoted as ). (The default score is 100 points). Points are then deducted based on the degree of abnormality of each indicator. The final score is the vehicle's basic operational safety score. The calculation formula is as follows:

[0079]

[0080] in:

[0081] : System takeover request frequency (times / hour) within the statistical period (default 1 operating shift). This is a deduction factor for the frequency of takeovers (default 5 points / time / hour). The higher the frequency of takeovers, the more points are deducted. ≥5 times / hour, points will be deducted up to 5 times / hour.

[0082] This indicates the total number of rapid accelerations and decelerations within the statistical period (rapid acceleration is defined as acceleration ≥ 2 m / s²; rapid deceleration is defined as deceleration ≤ -2 m / s², which can be adjusted according to vehicle type in actual calculations). The penalty coefficient for rapid acceleration and deceleration (default 0.5 points / time);

[0083] This indicates the number of traffic rule violations (such as running red lights, speeding, and illegal lane changes) within the statistical period. The penalty points are calculated based on the violation score (default 10 points / time). Serious violations (such as drunk driving or causing an accident) will result in a penalty of 0 points.

[0084] The final value of S1 is between 0 and 100 points. If the calculation result is negative, it will be counted as 0 points.

[0085] S42. Calculate the business efficiency score S2 based on the vehicle operation type;

[0086] S421. If the vehicle operation type is public bus, the first score for business efficiency (i.e., public bus business efficiency score) is calculated based on real-time passenger volume, planned trip completion rate, and stop punctuality rate. The weighted summation algorithm is used, and the formula is as follows:

[0087]

[0088] in:

[0089] This represents the real-time passenger load factor (real-time passenger volume / vehicle rated passenger volume × 100), with a value range of 0-100 points. A passenger load factor between 30% and 70% receives a full score of 100 points, while a load factor below 30% or above 70% is deducted in a linearly decreasing manner.

[0090] This represents the planned trip completion rate (actual completed trips / planned trips × 100), with a value range of 0-100. A completion rate of 100% earns full marks, and 10 points are deducted for every 10% decrease.

[0091] The on-time rate of the station is represented by (number of on-time arrivals / total number of arrivals × 100), with a value range of 0-100. A 100% on-time rate earns full marks, and 10 points are deducted for every 5% decrease in on-time rate.

[0092] Let be the weighting coefficient, satisfying Default value: .

[0093] S422. If the vehicle's operating type is a delivery vehicle, then the second business efficiency score (i.e., the delivery vehicle business efficiency score) is calculated based on the real-time task completion rate, the number of delayed packages, and the average delivery time per package. The weighted summation algorithm is used, and the formula is as follows:

[0094]

[0095] in:

[0096] This represents the real-time task completion rate (number of completed tasks / total number of tasks × 100), with a value range of 0-100 points. A completion rate of 100% earns full marks, and 10 points are deducted for every 5% decrease in completion rate.

[0097] This indicates the number of delayed packages (pieces) within the statistical period. The delay penalty coefficient (default 2 points / item);

[0098] This indicates the average delivery time per order (minutes / order). This is the time-based deduction coefficient (default 0.1 points / minute). If the average time is lower than the industry benchmark (default 30 minutes / ticket), no points will be deducted.

[0099] These are the weighting coefficients; the default values ​​are: ,final The value range is 0-100 points. If the calculation result is negative, it will be counted as 0 points.

[0100] S43. Calculate the service quality score S3 based on the service complaint rate and the number of abnormal service interaction events, according to the statistics of vehicles or operational tasks.

[0101] Specific calculation method: A standardized deduction system is adopted, with a maximum score set for service quality (referred to as...). (Default score is 100 points). Points are deducted based on complaints and abnormal events, using the following formula:

[0102]

[0103] in:

[0104] This represents the service complaint rate (number of valid complaints / total number of orders handled × 100%). This is the penalty coefficient for the complaint rate (default 500 points / %), meaning that 50 points are deducted for every 0.1% increase in the complaint rate.

[0105] This indicates the number of abnormal service interaction events (such as signal loss, location error, and no response) within the statistical period. This is the deduction coefficient for abnormal events (default 5 points / event);

[0106] The final value of S3 is between 0 and 100 points. If the calculation result is negative, it will be counted as 0 points. A full score of 100 points will be awarded if there are no complaints or abnormal events.

[0107] Among them, the vehicle basic operation safety score ( ), business efficiency score ( Bus pick-up Courier delivery vehicle pickup ), service quality rating ( The calculation of the comprehensive score includes, but is not limited to, weighted summation.

[0108] S44. For any vehicle, conduct a comprehensive scoring analysis by combining the vehicle's basic operational safety score, business efficiency score, and service quality score.

[0109] The weighted summation algorithm is used to calculate the comprehensive score S, and the scoring criteria are clearly defined as follows:

[0110] Comprehensive score calculation formula

[0111]

[0112] in:

[0113] a, b, and c are the weighting coefficients for the three ratings, satisfying... It can be dynamically adjusted according to the regulatory focus; the default value is:

[0114] the bus: (Focus on safety);

[0115] Delivery vehicles: (Efficiency is the primary focus).

[0116] The value of S ranges from 0 to 100, and is rounded to one decimal place.

[0117] S4411. If the vehicle's operating type is a delivery vehicle, then check whether the vehicle's task completion rate is consistently higher than the preset second threshold, and then continue monitoring.

[0118] S4412. When the vehicle task completion rate is not higher than the preset second threshold (85%), check whether the system takeover frequency is abnormal; if the system takeover frequency in the current statistical period is less than 1.5 times the average frequency of the fleet during the same period, continue to monitor.

[0119] The second threshold is set at 80%-85%. The industry benchmark for daily task completion rate in the express delivery industry is above 95%, and 80% is the low-risk warning threshold. If it is lower than this value, it indicates that the vehicle delivery efficiency is significantly abnormal.

[0120] S4413. If the system takeover frequency in the current statistical period is not less than 1.5 times the average frequency of the fleet during the same period, then check whether the number of package delay complaints has increased; if the number of valid complaints about vehicles in the current statistical period is lower than the preset absolute number threshold, then continue monitoring.

[0121] S4414. If the number of valid complaints against vehicles is not lower than the preset absolute number threshold during the current statistical period, then the multi-condition capability warning for the core business of express delivery vehicles will be triggered, forming the first warning information.

[0122] S4421. If the vehicle's operating type is a bus, then check the difference between the vehicle's real-time passenger load and the average historical passenger load at the current moment. If the absolute value of the difference is less than the third threshold (30%), then continue monitoring.

[0123] The third threshold is set at 30%-35%, based on historical passenger flow data of urban public transport. Under normal operating conditions, the relative deviation between the real-time passenger volume and the historical average for the same period should not exceed 20%, and 30% is the critical value for abnormal passenger flow (such as missed stops, sudden increase or decrease in passenger flow).

[0124] S4422. If the absolute value of the difference between the real-time passenger volume and the historical average passenger volume at the current time is not lower than the third threshold, then the vehicle station accuracy rate is judged. If the vehicle station accuracy rate is higher than the fourth threshold (85%), then monitoring continues.

[0125] The fourth threshold is set at 85%. The urban public transportation operation service standard requires that the on-time rate of bus stops be ≥90%. 85% is the warning threshold. If it is lower than this value, it indicates that the vehicle operation and scheduling is significantly abnormal.

[0126] S4423. If the vehicle station accuracy rate is not higher than the fourth threshold, check whether the number of complaints about the operation task has increased. If the number of valid complaints about vehicles is lower than the preset absolute number threshold in the current statistical period, continue to monitor.

[0127] S4424. If the number of valid complaints about vehicles is not lower than the preset absolute number threshold during the current statistical period, then the multi-condition capability warning for the core business of public buses will be triggered, forming a second warning message.

[0128] S45. Analyze early warning information based on multi-scale digital twins of vehicles for different services;

[0129] The table below shows the classification and explanation of the indicators.

[0130]

[0131] The steps for generating the multi-scale digital twin of the vehicle include:

[0132] S451. Construct a vehicle-scale digital twin to achieve mapping from the physical vehicle entity to the virtual space; including:

[0133] S4511: Real-time acquisition of high-frequency CAN bus data (sampling rate ≥30Hz) via vehicle gateway, including battery cell voltage / temperature, motor speed / torque, brake line pressure, and steering angular velocity signal. Simultaneously, combining the vehicle's factory BOM data and CAD 3D model, a parametric geometric skeleton of the vehicle is constructed. Material thermal properties such as battery pack thermal conductivity, brake disc specific heat capacity, mechanical wear curves such as brake pad thickness decay function with mileage, and sensor noise characteristic curves, including Gaussian distribution parameters based on historical data statistics, are assigned to the corresponding mesh nodes of the geometric model.

[0134] S4512. A vehicle-scale digital twin based on a deep fully connected neural network model is realized using pure data-driven methods. The deep fully connected neural network model includes an input layer, a hidden layer, and an output layer. The input layer receives sparse observations from real vehicle sensors, such as surface temperature and total voltage, as well as control commands such as throttle and brake opening. The hidden layer contains multiple nonlinear activation layers to fit complex nonlinear physical relationships. The output layer predicts the full-field state variables inside the vehicle, such as the three-dimensional temperature distribution inside the battery and the stress field of the brake disc contact surface.

[0135] The deep fully connected neural network model structure includes:

[0136] Number of input layer neurons: 64, corresponding to 64-dimensional real vehicle sensor observations and control commands;

[0137] Hidden layers: 3 layers, with 128, 256, and 128 neurons respectively;

[0138] Activation functions: ReLU hidden layer, Sigmoid output layer;

[0139] Optimizer: Adam, Learning rate: 0.001, Batch size: 32, Number of training epochs: 100;

[0140] Embedded physical equations: heat conduction equation (3D), vehicle dynamics equations (longitudinal / lateral)

[0141] Training dataset: Real vehicle historical operating data (≥100,000 km) + high-fidelity simulation data (generated by CarSim / SUMO, covering normal / fault / extreme operating conditions), dataset split: training set 80%, validation set 15%, test set 5%;

[0142] S4513. Construct the loss function for the deep fully connected neural network model. The loss function is:

[0143] L=L data + L physics + L boundary

[0144] Among them, L data L represents the mean square error between the network prediction and the actual measured values ​​from the sparse sensors on the vehicle. physics Representing the residual terms of the physical equations, the heat conduction equation is discretized and embedded into a network. This allows the twin to deduce its internal state based on physical laws even in the event of sensor loss or failure. boundary These represent boundary condition constraints to ensure that the model conforms to the physical boundaries of the vehicle's actual operation.

[0145] S4514. First, offline pre-training is performed in the cloud using massive historical operational data and high-fidelity simulation data. When the residual between the actual vehicle state and the predicted value of the vehicle-scale digital twin exceeds the dynamic threshold, backpropagation is triggered to update the network weights and adaptively correct the model deviation caused by vehicle aging, such as increased battery internal resistance and increased mechanical clearance.

[0146] S4515. Using the trained deep fully connected neural network model PINN, extreme fault scenarios from reality are injected into the virtual space to complete the construction of a vehicle-scale digital twin.

[0147] S452. Construct a fleet-scale digital twin to aggregate discrete vehicle states into a continuous task flow and fleet formation dynamic map; including:

[0148] S4521, Multi-source business data fusion: Access TMS (Transportation Management System), OMS (Order Management System), and GPS / BeiDou positioning data to complete data cleaning and alignment;

[0149] S4522. Construct a spatiotemporal task hypergraph, define hypergraph nodes and hyperedges, and assign high-dimensional state vectors to nodes; where nodes represent vehicles, and edges represent the relationships between vehicles; for example, two vehicles in front of and behind each other on the same road or transfer vehicles sharing the same batch of goods, multiple vehicles arriving at a station within the same time period can be linked by edges; the high-dimensional state vector of a node has 32 dimensions (including vehicle location, task progress, running status, fault information, etc.); the hyperedge association threshold is: vehicles on the same route / delivery area and within the same operating period (within 40 minutes) are associated.

[0150] S4523. Simulate the dynamic evolution of the task flow. Based on the discrete event DES simulation engine, simulate the node state update and associated effects after the event is triggered.

[0151] S4524 Transformer Model Training: Constructing input sequences, designing spatiotemporal self-attention modules, and adopting multi-task joint and course learning training strategies;

[0152] The Transformer model structure includes:

[0153] Encoder layers: 6 layers; Decoder layers: 6 layers;

[0154] Multi-head attention head count: 8; word embedding dimension: 128;

[0155] Spatiotemporal self-attention module: The temporal attention window size is 1 hour, and the spatial attention neighborhood range is 3km;

[0156] Training hyperparameters: optimizer AdamW, learning rate 0.0005, number of training epochs 50, and early stopping mechanism: training stops if the validation set loss does not decrease for 5 consecutive epochs.

[0157] The multi-task joint training involves designing the model's output layer as a multi-task head, simultaneously predicting key metrics for the next K time steps:

[0158] Performance risks: Probability distribution of predicted task completion rate deviation and delay duration;

[0159] Safety Trends: Predicting the probability of takeover, the frequency of sudden acceleration and deceleration, and the probability of triggering complaints in the future;

[0160] Resource bottlenecks: predicting station backlog and charging pile queue length;

[0161] The course learning and training strategy includes first allowing the fleet-scale digital twin to learn the patterns under normal operating conditions, and then gradually adding samples containing sudden events such as rainstorms and large-scale events to improve the robustness of the model.

[0162] S4525, Predicting and Controlling Trends: The fleet-scale digital twin model outputs predictions of performance risks or safety trends, triggering intervention strategies.

[0163] For example, inputting a multidimensional time series X into a fleet-scale digital twin. t This includes historical operational data from the past T hours, real-time fleet status snapshots, fleet vehicle warning points, and external context, including road condition information.

[0164] The fleet-scale digital twin model outputs a probability distribution with confidence intervals. When the predicted takeover probability or delay risk exceeds a dynamic threshold, an intervention strategy is generated, and the prediction result is fed back to the macro scale as a basis for regional risk assessment. Through a continuous "prediction-execution-feedback" closed loop, the meso-level twin achieves advanced perception and proactive control of the fleet's operational status.

[0165] S453. Construct a road network-scale digital twin to achieve dynamic, updated holographic digital foundation simulation of the "transportation-environment-infrastructure" within the target area; including:

[0166] S4531 will aggregate multi-source data from high-precision map static layers, real-time floating car data FCD, roadside unit RSU perception data, meteorological grid data such as rainfall, visibility, and road surface temperature, and infrastructure IoT data such as traffic light status, manhole cover displacement, and water accumulation sensor data, and achieve spatiotemporal alignment through timestamp synchronization and coordinate transformation.

[0167] S4532. On the basis of traditional static maps, a dynamic semantic layer is overlaid; each grid cell contains geographic information and carries dynamic attributes in real time: real-time traffic flow density, average vehicle speed, estimated road friction coefficient, inversion based on vehicle ABS trigger frequency, environmental hazards such as fog concentration; inject risk feature vectors at the road network level, including: distribution of historical accident black spots, accident rate statistics under specific weather conditions, impact range of road construction, and pedestrian flow heat map.

[0168] S4533, Spatiotemporal Graph Convolutional ST-GNN model training, constructing a road network-scale dynamic graph structure, adopting a graph convolutional GCN fusion temporal convolutional TCN hybrid architecture, and performing unsupervised training to obtain a road network-scale digital twin;

[0169] The urban road network is abstracted as a dynamic graph G=(V,E); nodes V represent road segments and intersections; edges E represent the connectivity between nodes and the traffic flow influence relationship, such as directed weighted edges, with weights based on real-time travel time or traffic volume. The node feature matrix contains real-time traffic indicators, environmental indicators, and the aggregation status of operating vehicles in the area, such as average takeover rate and average number of emergency brakings.

[0170] The ST-GNN spatiotemporal graph convolutional neural network model structure includes:

[0171] Graph Convolutional Network (GCN) number of layers: 3, Temporal Convolutional Network (TCN) kernel size: 3, convolution stride: 1;

[0172] Node feature matrix dimensions: 64 dimensions (including traffic flow, environment, and aggregation status of operating vehicles, etc.);

[0173] Activation function: GELU, optimizer: Adam, learning rate: 0.001, number of training epochs: 80.

[0174] S4534, Anomaly Detection: Anomalies are determined based on reconstruction errors, the root cause is located, and the anomaly information is sent back to the vehicle-level and fleet-level twins; wherein, the reconstruction error threshold is 0.2; the root cause location threshold is ≥0.3 for node contribution.

[0175] Based on GCN, neighborhood node information is aggregated to capture the spatial propagation effect of risks, such as the accident risk caused by frequent lane changes by downstream vehicles due to upstream intersection congestion. Based on TCN, the temporal evolution of traffic flow and risk indicators is captured to identify the difference between periodic fluctuations and sudden anomalies. The loss function is Reconstruction Error. Under normal conditions, the reconstruction error should be very small; when regional anomalies occur, such as sudden fog leading to multi-vehicle pile-up risks or traffic light malfunctions causing intersection chaos, the actual data distribution deviates from the learned normal manifold, causing the reconstruction error to increase sharply. When the error of a node in a certain region exceeds a threshold, it is judged as an anomaly. Further gradient backtracking or node contribution analysis is used to locate the key subgraph or key feature causing the anomaly, such as whether "rainfall" or "sudden traffic flow change" is dominant.

[0176] Once the vehicle-level twin verification warning passes (residual > fifth threshold), the fleet-level + road network-level twin linkage analysis is immediately triggered;

[0177] S454. Based on reinforcement learning, the optimal policy is explored, and the network is fine-tuned using RLHF. The model is updated at each scale through incremental learning. The updated model parameters are packaged and distributed over-the-air (OTA). After verification in shadow mode, the model is officially switched, forming a closed loop.

[0178] After the new model is launched, it continues to run in "shadow mode" within the digital twin—that is, it runs in parallel but is not actually controlled. The performance of the new and old models is compared, and the official switch is only made after the performance improvement is confirmed. This forms a complete positive cycle of "real data driving the evolution of the digital twin, training and optimizing the model in the digital twin environment, then empowering real-world supervision with the optimized model, and finally the return of new data," ensuring that the regulatory system has continuous self-evolution and adaptability.

[0179] S46. Verify the first or second warning information based on the vehicle-level digital twin, including:

[0180] S461. Obtain the real-time vehicle parameters corresponding to the first or second early warning information and inject them into the vehicle-scale digital twin. Compare the real-time values ​​of the early warning parameters, such as the braking acceleration of a bus, with the predicted values ​​of the vehicle-scale digital twin. If the residual between the two exceeds the fifth threshold (15%), the verification passes, triggering a verification early warning and performing vehicle anomaly root cause analysis (such as braking system failure or sensor anomaly). At the same time, immediately push the verification results and vehicle anomaly root causes to the fleet-scale digital twin.

[0181] The fifth threshold is set at 15%. The industry allowable range for the prediction error of intelligent connected vehicle sensors and digital twins is within 10%. 15% is the critical value for passing the verification. Exceeding this value indicates that the deviation between the actual vehicle parameters and the virtual model prediction is significant, and the warning information is true and effective.

[0182] S462. Generate proactive fleet status control information based on the fleet-level digital twin and verification and early warning information; input vehicle-level verification results, abnormal vehicle information, and real-time fleet operation data (such as vehicle location, task progress, and remaining capacity) into the Transformer model, predict fleet performance risks or safety trends through the Transformer model, and generate fleet control strategies, including proactive fleet status control information (such as issuing additional spare vehicles, adjusting departure intervals / delivery routes, and allocating surrounding vehicles for support); if the control strategy involves capacity allocation in the road network route adjustment area, immediately push the fleet control requirements and abnormal vehicle locations to the road network-level twin;

[0183] S463. Generate road network guidance information based on the road network-level digital twin and verification and early warning information; input fleet-level control requirements, abnormal vehicle locations, and real-time road network traffic data (traffic flow, congestion status, weather, road construction) into the ST-GNN model, analyze road network anomalies through the ST-GNN model, locate the risk propagation range, and generate road network-level guidance strategies; including road network guidance information (such as alternative vehicle detour routes, traffic restriction suggestions for congested sections, and regional vehicle departure control instructions); synchronously distribute fleet control information and road network guidance information to the monitoring platform, vehicle on-board terminals, and fleet dispatch center;

[0184] or

[0185] S464. Inject the real-time parameters of the vehicle corresponding to the first or second warning information into the multi-scale digital twin to perform automated fault analysis, such as during evening rush hour heavy rain, sudden braking of the vehicle in front, and abnormal acceleration and deceleration frequency of the current vehicle. Obtain the current road network traffic analysis and road planning information, and send additional vehicle information and current vehicle abnormal information to the fleet. If the current vehicle abnormal information output by the multi-scale digital twin is consistent with the warning information, then send the current road network traffic analysis and road planning information to the vehicles in the fleet, and send additional vehicle information to the fleet.

[0186] The consistency between vehicle abnormality information and warning information indicates that the vehicle abnormality information (such as system takeover request frequency or vehicle speed information) and the vehicle-related parameters (system takeover request frequency or vehicle speed information) in the warning information both exceed the preset safety value, and the difference between the two is less than the set value (10%).

[0187] The multi-scale interconnected simulation environment integrates the vehicle-level, fleet-level, and road network-level twins to construct an end-to-end simulation environment. The vehicle-level twins provide vehicle dynamics response, the fleet-level twins provide task flow logic, and the road network-level twins provide environmental context. Visualization rendering and physics solving are performed using game engines (such as Unity / Unreal) or professional simulation software (such as CarSim / SUMO), supporting Hardware-in-the-Loop (HIL) and Software-in-the-Loop (SIL) testing. Once the vehicle-level twin verification warning passes (residual > fifth threshold), the fleet-level + road network-level twin interconnected analysis is immediately triggered.

[0188] For example, based on the verification and warning information, relevant information about the warning vehicles is obtained, and a multi-dimensional time series X is input into the fleet-level digital twin. t This includes historical operational data from the past T hours, such as passenger pick-up and drop-off or freight throughput at each station, road travel time, accident records, real-time fleet status snapshots (e.g., speed reduction, stopping), and external context such as weather forecasts, holiday tags, road condition information, fleet warning vehicle numbers, and fleet warning vehicle status.

[0189] The fleet-scale digital twin model outputs a probability distribution with confidence intervals. For example, if the verification warning information includes abnormal vehicle speed, and the "delay risk" predicted by the fleet-scale digital twin model exceeds a dynamic threshold, intervention strategies are generated, such as suggesting adjusting departure intervals or switching to backup vehicles.

[0190] The road network-scale digital twin outputs road network guidance information, such as verification and warning information including abnormal vehicle speed. At the same time, when the error of the road network-scale digital twin in identifying the node of the area where the vehicle is located exceeds the threshold, it is judged as abnormal. If the abnormal vehicle speed is determined to be caused by weather, such as heavy rain, then a suspension of departure suggestion is generated.

[0191] Furthermore, the judgment of abnormal vehicle conditions also includes:

[0192] When a vehicle remains stationary in an unplanned area for more than a preset time without reporting a system malfunction, the system automatically triggers an "abnormal operational standby" warning process, which includes:

[0193] S491. Data Acquisition and Status Monitoring: Real-time acquisition of high-precision vehicle positioning data and fault status information reported by the vehicle control system.

[0194] S492, Multi-condition collaborative judgment: System parallel execution area deviation judgment, system status judgment, and time threshold judgment.

[0195] Area deviation judgment: Comparing the real-time location of the bus with the preset bus operation route (e.g., a certain bus route, starting point A station, ending point B station, passing through 12 stops), it was found that the current location of the vehicle deviated from the preset route by 500m, and there was no reason for the deviation manually reported by the dispatcher.

[0196] System status assessment: By analyzing the fault information reported by the vehicle control system, a fault was found in the braking system (fault code P0571, abnormal brake switch circuit), and the vehicle was determined to be in a fault state that prevented it from driving normally.

[0197] Duration threshold judgment: Statistical analysis of the vehicle's stationary time (speed ≤ 0km / h). Starting from deviation from the route or occurrence of a malfunction, if the vehicle remains stationary for 6 consecutive minutes, it exceeds the preset 5-minute duration threshold for buses.

[0198] Furthermore, deviation thresholds are set according to the type of business;

[0199] Bus route deviation threshold: 500m (urban roads) / 1000m (suburban roads);

[0200] Vehicle stationary duration thresholds: 5 minutes for buses, 8 minutes for delivery vehicles (delivery vehicles have many and widespread delivery points, so short stops are normal operating conditions).

[0201] Fault reporting timeout threshold: 3 minutes (After a vehicle detects a fault, it must upload the fault report to the monitoring platform within 3 minutes; failure to upload will be considered an abnormal fault reporting).

[0202] S493. Warning generation and classification: When the above judgment conditions are met simultaneously, an "abnormal operation stoppage" warning is automatically generated and classified.

[0203] Because the above three conditions—regional deviation, abnormal system status, and exceeding the time threshold—were met simultaneously, the system automatically generated a Level 2 "Operational Abnormality and Stagnation" warning. Simultaneously, based on comprehensive scoring analysis, the bus's comprehensive score was 58 points (below the first threshold of 60 points). Furthermore, after multi-condition verification (the absolute value of the difference between real-time passenger volume and the historical average for the same period ≥ the third threshold, the station punctuality rate ≤ the fourth threshold, and the number of valid complaints ≥ the absolute number threshold), a second warning message (multi-condition capability warning for core bus operations) was triggered. The warning parameters were specified as "system takeover request frequency 8 times / hour (exceeding the preset safety value of 5 times / hour) and abnormal vehicle speed (stationary for 6 minutes)."

[0204] S494, Automated Fault Analysis and Verification of Multi-Scale Digital Twins;

[0205] The second early warning information and the real-time vehicle parameters corresponding to the "abnormal operation stoppage" warning (system takeover request frequency of 8 times / hour, vehicle speed of 0km / h, braking system fault code, real-time location, passenger capacity, etc.) are simultaneously injected into the vehicle-level, fleet-level, and road network-level digital twins to perform automated fault analysis and simulate the current scenario (evening rush hour, localized light rain, and the bus's braking failure after the preceding vehicle brakes suddenly):

[0206] Vehicle-level digital twin: Reverses the internal state of the vehicle, confirms that a braking system failure caused the vehicle to be unable to move, and outputs the vehicle abnormality information as "system takeover request frequency 7.8 times / hour, vehicle speed 0km / h"; Fleet-level digital twin: Analyzes the operational status of the bus fleet to which this bus belongs, predicts the backlog of subsequent stops, and outputs the fleet's additional vehicle dispatch information (one additional backup bus needs to be dispatched to connect with subsequent trips from station A).

[0207] Road network-level digital twin: Analyzes the current traffic status of the regional road network and finds that the road segment is experiencing localized congestion due to the evening rush hour and vehicle stagnation. Outputs road network traffic analysis (congestion range 1km, estimated relief time 15 minutes) and road planning information (alternate buses detour via adjacent branch roads to shorten connection time).

[0208] S495. Visualization and Alarm Push: Warning information is pushed to the monitoring platform in real time and visualized by changing the color of vehicle icons on the map and generating structured records in the alarm center.

[0209] On the monitoring map, the bus icon is marked in yellow (Level 2 warning), highlighting its stationary location and the surrounding congestion area; a structured warning record is generated, which records in detail the vehicle number, warning type, fault information, stationary duration, digital twin verification results, fleet dispatch suggestions, etc.

[0210] S5. Input the analysis results into the regulatory visualization interface to obtain the vehicle situation diagram and traffic heat map;

[0211] The situational map includes buses and delivery vehicles distinguished by different colors and icons on the map, and vehicles in warning and alarm states are highlighted; the traffic heat map shows the traffic status of the road network within the target area.

[0212] S6. Determine the vehicle risk level based on the vehicle situation map and traffic heat map, and push early warning information and a list of optimal handling methods to vehicles whose risk level exceeds the threshold.

[0213] Furthermore, the vehicle push notification information includes the warning level and the root cause analysis results of the multi-scale digital twin;

[0214] Furthermore, the list of optimal treatment methods includes differentiated optimal treatment methods, such as those categorized by vehicle type and risk level.

[0215] For example, multi-level early warning systems for buses include:

[0216] Level 1 Warning: Adjust departure intervals, strengthen driver reminders, and monitor passenger flow in real time;

[0217] Level 2 alert: Increase the number of backup vehicles, temporarily adjust routes, and dispatch personnel to guide passenger flow on-site;

[0218] Level 3 warning: Immediately stop the operation of the disabled vehicle, arrange for rescue vehicles to tow it away, activate the temporary bus shuttle plan, and notify the traffic management department to control the road section.

[0219] For example, multi-level early warning systems for express delivery vehicles include:

[0220] Level 1 Warning: Adjust delivery routes, allocate nearby couriers for support, and extend delivery time per order;

[0221] Level 2 warning: Replace delivery vehicles, suspend some non-emergency delivery tasks in the area, and notify customers of the reason for the delay;

[0222] Level 3 warning: Immediately stop the malfunctioning vehicle, activate the parcel transfer plan, and arrange other delivery vehicles to take over the delivery task.

[0223] Example 2

[0224] like Figure 2 As shown, an active safety monitoring system for intelligent connected commercial vehicles is used to implement an active safety monitoring method for intelligent connected commercial vehicles. The system includes:

[0225] The vehicle data acquisition module is used to acquire vehicle dynamic data; it is acquired through an on-board terminal or data interface, including vehicle CAN bus data, GPS location information, autonomous driving system control mode, and system fault codes.

[0226] It is also used to obtain vehicle operation business data, and after determining the vehicle type, it collects data based on the vehicle route coverage and target area range.

[0227] It is also used to obtain vehicle service data and external data; to obtain vehicle service data based on the enterprise customer service system, including valid complaint records and handling information for vehicles or operational tasks; and to obtain vehicle traffic violation data based on external data interfaces.

[0228] The intelligent analysis and early warning engine is used to perform data analysis and data mining on vehicle dynamic data, vehicle operation business data, vehicle service data and external data, and obtain analysis results based on business scenarios.

[0229] The data display module is used to input the analysis results into the regulatory visualization interface to obtain vehicle situation maps and traffic heat maps. The situation maps include buses and delivery vehicles distinguished by different colors and icons on the map, and vehicles in the warning and alarm states are highlighted. The traffic heat map shows the traffic status of the road network within the target area.

[0230] The risk alert module is used to determine the risk level of a vehicle based on the vehicle situation map and traffic heat map, and push warning information and a list of optimal handling methods to vehicles whose risk level exceeds the threshold.

[0231] Example 3

[0232] A computer-readable storage medium storing a computer program, characterized in that a processor executes the computer program to implement a proactive safety monitoring method for intelligent connected vehicles.

[0233] Example 4

[0234] A terminal device includes a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that the processor executes the computer program to implement an active safety monitoring method for intelligent connected vehicles.

[0235] The beneficial effects of this invention are as follows:

[0236] 1) This solution adopts a multi-level early warning mechanism that combines comprehensive scoring screening with multi-scale twin verification. It initially screens high-risk vehicles through basic safety, business efficiency, and service quality scores, and designs differentiated verification rules for the business characteristics of buses and delivery vehicles. Finally, it uses vehicle-level, fleet-level, and road network-level twins to verify the authenticity of the early warning, effectively reducing false alarms and missed alarms. After the early warning, control and guidance information can be directly output. At the same time, it realizes a fundamental change in the regulatory model. Through a configurable intelligent rule engine, it performs real-time correlation analysis of multi-source data, which can proactively issue early warnings in the early stages of problems, significantly improving the foresight and effectiveness of supervision.

[0237] 2) This solution differentiates between two types of operating vehicles, namely buses and express delivery vehicles, and collects corresponding business data such as bus dispatch data, express delivery lists, and parcel entry and exit data in a differentiated manner. At the same time, it integrates vehicle dynamics, service, and external violation data to achieve full coverage of multi-dimensional data of "vehicle-business-environment". This avoids redundancy and omissions caused by generalized collection and lays the foundation for subsequent accurate analysis. It can also effectively discover deep-seated operational safety issues that cannot be revealed by single-dimensional data.

[0238] 3) This solution constructs a three-level digital twin of vehicles, fleets, and road networks to achieve cross-scale linkage of micro-level vehicle state inversion, meso-level fleet situation control, and macro-level road network anomaly detection. At the same time, the multi-level digital twins can adapt to dynamic changes such as vehicle aging and road network evolution through reinforcement learning, RLHF fine-tuning, and OTA distribution, thereby improving the adaptability and effectiveness of supervision in the long term.

[0239] The above description is merely a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A proactive safety supervision method for intelligent connected commercial vehicles, characterized in that, The method includes the following steps: S1. Obtain vehicle dynamic data; Data collected via vehicle-mounted terminals or data interfaces includes vehicle CAN bus data, GPS location information, autonomous driving system control modes, and system fault codes. S2. Obtain vehicle operation business data, determine the vehicle type, and collect data based on the vehicle route coverage and target area. S3. Obtain vehicle service data and external data; based on the enterprise customer service system, obtain vehicle service data including valid complaint records and handling process information for vehicles or operational tasks; obtain vehicle traffic violation data through external data interfaces. S4. Based on the intelligent analysis and early warning engine, perform data analysis and data mining on vehicle dynamic data, vehicle operation business data, vehicle service data and external data, and obtain analysis results based on business scenarios. S5. Input the analysis results into the regulatory visualization interface to obtain the vehicle situation diagram and traffic heat map; The situational map includes buses and delivery vehicles distinguished by different colors and icons on the map, and vehicles in warning and alarm states are highlighted; the traffic heat map shows the traffic status of the road network within the target area. S6. Determine the vehicle risk level based on the vehicle situation map and traffic heat map, and push early warning information and a list of optimal handling methods to vehicles whose risk level exceeds the threshold. The data analysis and data mining include: S41. Obtain a basic vehicle operation safety score based on the frequency of system takeover requests, the number of times of rapid acceleration and deceleration, and the number of traffic rule violations. S42. Calculate the business efficiency score based on the vehicle operation type. S421. If the vehicle operation type is a bus, the first score for business efficiency is calculated based on the real-time passenger volume, the planned trip completion rate, and the station punctuality rate. S422. If the vehicle operation type is a courier delivery vehicle, the real-time task completion rate, the number of delayed packages, and the average delivery time per package are used to calculate the second score of business efficiency. S43. Calculate the service quality score based on the service complaint rate and the number of abnormal service interaction events as statistically analyzed for vehicles or operational tasks. S44. For any vehicle, a comprehensive score analysis is performed on the vehicle's basic operational safety score, business efficiency score, and service quality score; if the comprehensive score is lower than the first threshold, a multi-condition capability judgment of core business is performed. S45. Analyze early warning information based on multi-scale digital twins of vehicles for different services; S46. Verify the first or second warning information based on the vehicle-level digital twin, including: S461. Obtain the real-time vehicle parameters corresponding to the first or second early warning information and inject them into the vehicle-scale digital twin. Compare the real-time values ​​of the early warning parameters with the predicted values ​​of the vehicle-scale digital twin. If the residual between the two exceeds the fifth threshold, the verification passes and the verification early warning is triggered. S462. Generate active vehicle situation control information based on the vehicle-level digital twin and verification and early warning information; S463. Generate road network guidance information based on the road network-level digital twin and verification and early warning information.

2. The proactive safety supervision method for intelligent connected operating vehicles according to claim 1, characterized in that, CAN bus data includes vehicle speed, acceleration, steering angle, braking status, and door or cargo box door opening / closing status signals.

3. The proactive safety supervision method for intelligent connected operating vehicles according to claim 1, characterized in that, The process involves acquiring vehicle operation data, determining the vehicle type, and then collecting data based on the vehicle's route coverage and target area. For example, if the vehicle is a bus, the system obtains the bus dispatching system's schedule, planned routes, station timetables, actual arrival times, and passenger order data. If the vehicle is a delivery vehicle, the system obtains the delivery task list, planned routes, and planned delivery times from the logistics and delivery system, and simultaneously collects parcel entry and exit data from onboard scanning devices or cargo box sensors.

4. The proactive safety supervision method for intelligent connected operating vehicles according to claim 1, characterized in that: Operational tasks include one or more buses or delivery vehicles.

5. The proactive safety supervision method for intelligent connected operating vehicles according to claim 1, characterized in that: The core business multi-condition capability assessment includes: S4411. If the vehicle's operating type is a delivery vehicle, then check whether the vehicle's task completion rate is consistently higher than the preset second threshold, and then continue monitoring. S4412. When the vehicle task completion rate is not higher than the preset second threshold, check whether the system takeover frequency is abnormal; if the system takeover frequency in the current statistical period is less than 1.5 times the average frequency of the fleet during the same period, continue to monitor. S4413. If the system takeover frequency in the current statistical period is not less than 1.5 times the average frequency of the fleet during the same period, then check whether the number of package delay complaints has increased; if the number of valid complaints about vehicles in the current statistical period is lower than the preset absolute number threshold, then continue monitoring. S4414. If the number of valid complaints against vehicles is not lower than the preset absolute number threshold during the current statistical period, then the multi-condition capability warning for the core business of express delivery vehicles will be triggered, forming the first warning information.

6. The proactive safety supervision method for intelligent connected operating vehicles according to claim 1, characterized in that: The multi-condition capability assessment of core business also includes: S4421. If the vehicle's operating type is a bus, then check the difference between the vehicle's real-time passenger load and the average historical passenger load at the current moment. If the absolute value of the difference is less than the third threshold, then continue monitoring. S4422. If the absolute value of the difference between the real-time passenger volume and the historical average passenger volume at the current time is not lower than the third threshold, then the vehicle station accuracy rate is determined. If the vehicle station accuracy rate is higher than the fourth threshold, then monitoring continues. S4423. If the vehicle station accuracy rate is not higher than the fourth threshold, check whether the number of complaints about the operation task has increased. If the number of valid complaints about vehicles is lower than the preset absolute number threshold in the current statistical period, continue to monitor. S4424. If the number of valid complaints about vehicles is not lower than the preset absolute number threshold during the current statistical period, then the multi-condition capability warning for the core business of public buses will be triggered, forming a second warning message.

7. A proactive safety monitoring system for intelligent connected vehicles, used to execute the method as described in any one of claims 1-6, the system comprising: The vehicle data acquisition module is used to acquire vehicle dynamic data; Data collected via vehicle-mounted terminals or data interfaces includes vehicle CAN bus data, GPS location information, autonomous driving system control modes, and system fault codes. It is also used to obtain vehicle operation business data, and after determining the vehicle type, it collects data based on the vehicle route coverage and target area range. It is also used to obtain vehicle service data and external data; to obtain vehicle service data based on the enterprise customer service system, including valid complaint records and handling information for vehicles or operational tasks; and to obtain vehicle traffic violation data based on external data interfaces. The intelligent analysis and early warning engine is used to perform data analysis and data mining on vehicle dynamic data, vehicle operation business data, vehicle service data and external data, and obtain analysis results based on business scenarios. The data display module is used to input the analysis results into the regulatory visualization interface to obtain vehicle situation maps and traffic heat maps. The situation maps include buses and delivery vehicles distinguished by different colors and icons on the map, and vehicles in the warning and alarm states are highlighted. The traffic heat map shows the traffic status of the road network within the target area. The risk alert module is used to determine the risk level of a vehicle based on the vehicle situation map and traffic heat map, and push warning information and a list of optimal handling methods to vehicles whose risk level exceeds the threshold.

8. A computer-readable storage medium storing a computer program, characterized in that, The processor executes the computer program to implement the method as described in any one of claims 1-6.

9. A terminal device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, The processor executes the computer program to implement the method as described in any one of claims 1-6.

Images