Real-time data interaction and supervision traffic transportation safety management integrated platform

Abstract

The invention, which relates to the technical field of traffic transportation management, discloses a real-time data interaction and supervision traffic transportation safety management integrated platform comprising a sensing module, an edge calculation module, a safety evaluation module, a decision module and a linkage intervention module. The sensing module comprises a flexible piezoresistive sensor array; the edge calculation module is used for carrying out real-time cleaning, feature extraction and redundancy elimination processing on the data; the safety evaluation module evaluates the state of the driver in real time through a driver monitoring algorithm fused by multiple collected data; the decision-making module fuses data from the sensing module and the edge calculation module to perform dynamic risk assessment, and generates an intelligent decision-making chain; the linkage intervention module obtains a physiological signal of a driver and a vehicle running state based on the safety evaluation module and the decision module, and triggers electronic horizon line early warning, automatic speed limiting and dynamic navigation; according to the invention, more accurate and comprehensive safety assessment can be provided for vehicle safety, and the system can make more intelligent intervention decisions.

Description

Technical Field

[0001] The present invention relates to the technical field of transportation management, and specifically to an integrated platform for transportation safety management with real-time data interaction and supervision. Background Art

[0002] Transportation is an important part of the national economic development, and the road transportation industry plays a crucial role among them. It not only has a wide service scope, large transportation volume, and a large number of employees, but also has the advantages of flexibility, convenience, wide applicability, openness, etc. In addition, the road transportation industry directly promotes the prosperity and development of related industries such as the automobile industry. With the rapid development of the social economy, transportation has become increasingly important. All walks of life widely use vehicle transportation, and ensuring transportation safety is particularly important during the transportation process.

[0003] However, the current transportation management system still faces the following key technical bottlenecks:

[0004] Currently, most transportation management platforms still operate as independent modules. The dispatching platform and the supervision system require manual switching of multiple interfaces, lacking an integrated design. This decentralized operation process is not only inefficient but also prone to information gaps or response delays.

[0005] Traditional transportation safety management relies on manual experience judgment, lacking dynamic risk portraits and adaptive decision-making models, and it is difficult to effectively respond to complex or sudden road conditions, especially in driver fatigue monitoring and early warning of dangerous situations, there are obvious deficiencies. Summary of the Invention

[0006] To solve the defects existing in the prior art, the present invention provides an integrated platform for transportation safety management with real-time data interaction and supervision.

[0007] To solve the above technical problems, the present invention provides the following technical solutions:

[0008] The present invention provides an integrated platform for transportation safety management with real-time data interaction and supervision, including a sensing module, an edge computing module, a safety assessment module, a decision-making module, and a linkage intervention module;

[0009] The sensing module includes an inertial measurement unit, a lidar, an on-board diagnostic system for collecting traffic data by the vehicle terminal, and a flexible piezoresistive sensor array embedded in the driver's seat for collecting the driver's physiological characteristic data;

[0010] The edge computing module is used for real-time cleaning, feature extraction, and redundancy removal processing of the data;

[0011] The safety assessment module uses a driver monitoring algorithm that fuses multiple sets of collected data to evaluate the driver's state in real time and trigger intervention actions in case of anomalies;

[0012] The decision-making module fuses data from the perception module and the edge computing module to conduct dynamic risk assessment and generate an intelligent decision-making chain;

[0013] The process of generating the decision-making chain is as follows:

[0014] S1, The dynamic Bayesian network unifies the modeling of data collected by the inertial measurement unit, lidar, on-board diagnostic system, and flexible piezoresistive sensor array, and establishes the evolution process of the system state over time. The state of the system at the current time is determined by the state at the previous moment and the current external input data, and is used to evaluate dynamic risks;

[0015] S2, The three-dimensional convolutional neural network performs convolutional operations on the temporal and spatial dimensions of the input collected data and extracts key features for detecting obstacles and path risks;

[0016] S3, By using the attention mechanism to combine different sets of collected data, the states of the vehicle and the driver are analyzed uniformly, and the important parts of different sets of collected data are automatically focused on;

[0017] S4, Use reinforcement learning to dynamically optimize the risk response strategy;

[0018] The linkage intervention module obtains the driver's physiological signals and the vehicle's operating state based on the safety assessment module and the decision-making module. When the driver's physiological signals or the vehicle's operating state show anomalies, it triggers electronic horizon warnings, automatic speed limits, and dynamic navigation.

[0019] As a preferred technical solution of the present invention, in the safety assessment module, the risk score includes the inertial measurement unit risk score R IMU , the lidar risk score R LiDAR , the on-board diagnostic system risk score R OBD , and the driver risk score R Driver .

[0020] As a preferred technical solution of the present invention, the sources of each risk score are quantitatively evaluated through the following specific algorithm formulas:

[0021] The inertial measurement unit is mainly used to monitor the dynamic behavior of the vehicle, and R IMU is represented by the following formula:

[0022] R IMU =α 1 ·|a x |+α 2 ·|a y |+α 3|ω z |;

[0023] The lidar is mainly used to monitor the distance and distribution of obstacles in the surrounding environment, R LiDAR which is expressed by the following formula:

[0024]

[0025] The on-vehicle diagnostic system is used to collect the operating state of the vehicle, R OBD which is expressed by the following formula:

[0026]

[0027] The driver risk score detects the driver's posture and state through a flexible piezoresistive sensor array, and evaluates their attention and fatigue level, R Driver which is expressed by the following formula:

[0028] R Driver =δ 1 ·F fatigue +δ 2 ·A attention +δ 3 ·P pose .

[0029] As a preferred technical solution of the present invention, the total risk score is calculated by the following formula:

[0030] R t =μ 1 ·R IMU +μ 2 ·R LiDAR +μ 3 ·R OBD +μ 4 ·R Driver .

[0031] As a preferred technical solution of the present invention, the dynamic Bayesian network in step S1 is expressed by the following formula:

[0032]

[0033] As a preferred technical solution of the present invention, the three-dimensional convolutional neural network in step S2 is expressed by the following formula:

[0034]

[0035] As a preferred technical solution of the present invention, the self-attention mechanism in step S3 is expressed by the following formula:

[0036]

[0037] As a preferred technical solution of the present invention, the reinforcement learning in step S4 is represented by the following formula:

[0038]

[0039] As a preferred technical solution of the present invention, in the sensing module, the flexible piezoresistive sensor array estimates the driver's blood pressure value based on the pulse wave transit time algorithm.

[0040] As a preferred technical solution of the present invention, the edge module uses the MEC gateway to assign QoS levels to multiple types of data and optimize network transmission performance using the 5G-TSN protocol. The MEC gateway, based on the time-sensitive network protocol, assigns data streams to high, medium, and low QoS levels according to time priority.

[0041] The beneficial effects of the present invention are:

[0042] 1. Through the interaction and coordinated work of the sensing module, edge computing module, security assessment module, decision-making module, and linkage intervention module, this integrated design greatly reduces the need for manual operation, avoiding the cumbersome operation of switching between multiple interfaces and modules. The system generates a dynamic risk assessment and intelligent decision-making chain by integrating data from the sensing module and edge computing module, automatically makes decisions and triggers responses. In this way, managers do not need to operate each module separately, and the platform can automatically make decisions and interventions based on real-time data.

[0043] 2. In the present invention, by combining the inertial measurement unit, lidar, on-board diagnostic system, and driver risk score, the overall risk status of the vehicle and the driver is comprehensively evaluated, and various potential risks are comprehensively considered, which can provide a more accurate and comprehensive safety assessment for vehicle safety, and the system can make more intelligent intervention decisions.

[0044] 3. In the present invention, through the dynamic Bayesian network and three-dimensional convolutional neural network, the driving environment and driver state are comprehensively modeled, a risk portrait is generated in real time, and the attention mechanism is used to accurately capture key risk factors. At the same time, the setting of the linkage intervention module enables the linkage intervention module to trigger the warning system when the driver or vehicle state is abnormal. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] The drawings are used to provide a further understanding of the present invention and constitute a part of the specification. They are used together with the embodiments of the present invention to explain the present invention and do not constitute a limitation to the present invention. In the drawings:

[0046] Figure 1 Schematic diagrams of each module of the present invention.

[0047] Figure 2 Schematic diagram of the decision-making chain generation process.

[0048] Figure 3 It is a structural schematic diagram of each risk score. Specific implementation manners

[0049] The preferred embodiments of the present invention will be described below with reference to the accompanying drawings. It should be understood that the preferred embodiments described herein are only for the purpose of illustrating and explaining the present invention, and are not used to limit the present invention.

[0050] Embodiment 1

[0051] As Figure 1 - Figure 2 shown, an integrated platform for transportation safety management with real-time data interaction and supervision includes a sensing module, an edge computing module, a safety assessment module, a decision-making module, and a linkage intervention module;

[0052] The sensing module includes an inertial measurement unit for collecting traffic data by an in-vehicle terminal, a lidar, an on-board diagnostic system, and a flexible piezoresistive sensor array embedded in the driver's seat for collecting the driver's physiological characteristic data. The inertial measurement unit is used to detect the motion state of the vehicle in real time through devices such as an accelerometer and a gyroscope, such as acceleration, speed, angular velocity, etc., and obtain the attitude change information of the vehicle. These data are helpful for analyzing the motion behavior of the vehicle and potential risk situations. The lidar is used to scan the surrounding environment in real time to obtain the distance, shape, and relative position of obstacles. This data is crucial for analyzing the surrounding obstacles, traffic conditions, and route planning. Especially in complex environments, the lidar can provide high-precision real-time maps and spatial information. The on-board diagnostic system collects the health status information of the vehicle, including the engine, braking system, tire pressure, fuel consumption, etc. These information are helpful for monitoring the mechanical state of the vehicle and timely discovering potential fault problems, providing safety warnings for the driver. The flexible piezoresistive sensor array is used to collect the driver's physiological characteristic data (such as heart rate, breathing rate, body movement, etc.). By monitoring the driver's physiological state, it helps to judge his fatigued or dangerous driving state. This module is crucial for driver health monitoring and fatigue driving warning;

[0053] The edge computing module is used to perform real-time cleaning, feature extraction, and redundancy removal processing on the data to ensure that the data can be transmitted to the next-level module in a timely and accurate manner for further analysis. The advantage of edge computing lies in its low latency and high performance, which can quickly respond to various emergencies, reduce the burden of data transmission to the cloud, and reduce latency;

[0054] The safety assessment module uses a driver monitoring algorithm that fuses multiple pieces of collected data to evaluate the driver's state in real time and trigger intervention behaviors in case of abnormalities. Based on the fusion and processing of multiple pieces of collected data, the safety assessment module uses the driver monitoring algorithm to evaluate the driver's state in real time, analyze whether they are in a dangerous state such as fatigue, stress, or physiological abnormalities. When it detects that the driver is abnormal (such as drowsy driving or physiological abnormalities), the safety assessment module can automatically trigger intervention measures, such as reminding the driver and activating safety mechanisms. This module combines the driver's physiological data, driving behavior, and environmental data to comprehensively evaluate the driver's current risk state and provides real-time feedback to the decision-making module for further decision-making;

[0055] The decision-making module fuses data from the perception module and the edge computing module for dynamic risk assessment and generates an intelligent decision-making chain;

[0056] The process of generating the decision-making chain is as follows:

[0057] S1, The dynamic Bayesian network unifies the data collected by the inertial measurement unit, lidar, on-board diagnostic system, and flexible piezoresistive sensor array for modeling and establishes the evolution process of the system state over time. The state of the system at the current time is determined by the state at the previous moment and the current external input data, which is used to evaluate dynamic risks and predict potential risk evolution paths;

[0058] S2, The three-dimensional convolutional neural network performs convolutional operations on the temporal and spatial dimensions of the input collected data and extracts key features for detecting obstacles and path risks and responding to real-time changes in the environment. The three-dimensional convolutional neural network can consider both time series and spatial data simultaneously, providing efficient support for the vehicle's path planning and dynamic adjustment;

[0059] S3, By combining different pieces of collected data through the attention mechanism, the state of the vehicle and the driver is analyzed uniformly, and the important parts of different pieces of collected data are automatically focused on. For example: the weights of different modal data are dynamically adjusted through the attention mechanism to ensure that more attention is paid to the data of the flexible sensor array when the driver is fatigued;

[0060] S4, Using reinforcement learning, the risk response strategy is dynamically optimized so that the system can make the best decisions according to different scenarios and achieve adaptive risk management, such as adjusting the strategy in real time according to the driving scenario (such as highway driving, rainy environment);

[0061] The linkage intervention module obtains the driver's physiological signals and the vehicle's operating state based on the safety assessment module and the decision-making module. When the driver's physiological signals or the vehicle's operating state are abnormal, it triggers electronic horizon warnings, automatic speed limits, and dynamic navigation. The linkage intervention module works closely with the safety assessment module and the decision-making module and triggers a series of intervention measures based on real-time risk assessment. This module mainly includes:

[0062] Electronic horizon warning: When potential risks are detected, early warnings are issued through electronic horizon technology to alert the driver of potential dangers or emergencies;

[0063] Automatic speed limit: Based on real-time risk assessment, the system can dynamically adjust the speed limit of the vehicle and implement automated control to prevent accidents caused by speeding;

[0064] Dynamic navigation: When emergencies occur on the road, the system can re-plan the route through the dynamic navigation function, guiding the driver to choose a safe path and avoid entering dangerous areas.

[0065] The core purpose of the above integrated platform is to improve the safety and management efficiency in the transportation process. By integrating multiple advanced technology modules, a more comprehensive and intelligent risk warning and intervention mechanism can be achieved.

[0066] In addition, the multi-source heterogeneous data collection and fusion of traditional in-vehicle and roadside equipment are inefficient. Due to the variety of sensors, different data protocols and formats, it often takes a large delay to complete information processing, and it is impossible to give early warnings and intervene in sudden risks in real time. The combination of the sensing module and the edge computing module in the present invention can effectively solve the above problems. The sensing module includes multiple sensors, and the real-time data collection of these sensors provides rich inputs for the traffic management platform. The edge computing module performs real-time cleaning, feature extraction and redundancy removal on these data, and can quickly process data from different sensors to ensure the efficient fusion and real-time update of information. By performing data preprocessing at the edge, the delay caused by large-scale data transmission is avoided, thereby improving the system response speed.

[0067] Furthermore, as Figure 3 shown, in the safety assessment module, the risk score includes the inertial measurement unit risk score R IMU , the lidar risk score R LiDAR , the on-board diagnostic system risk score R OBD , and the driver risk score R Driver . By real-time evaluating the driver's state and triggering intervention behaviors in abnormal situations, the safety of road traffic can be effectively improved.

[0068] Specifically, the sources of each risk score are quantitatively evaluated through the following specific algorithm formulas:

[0069] The inertial measurement unit is mainly used to monitor the dynamic behavior of the vehicle (such as sudden acceleration, sudden braking, tilt angle, etc.), and can monitor the driver's driving behavior in real time and identify dangerous driving behaviors. For example, sudden braking or sudden acceleration may represent the driver's control error or slow reaction, which may cause an accident. R IMUIt is expressed by the following formula:

[0070] R IMU = α 1 ·|a x | + α 2 ·|a y | + α 3 ·|ω z |

[0071] Wherein, α x represents the acceleration of the vehicle in the front - rear (x - axis) direction, α y represents the acceleration of the vehicle in the lateral (y - axis) direction, ω z represents the angular velocity of the vehicle along the vertical axis (for detecting rotation or skidding), and α 1 , α 2 , α 3 are weight parameters respectively, reflecting the influence degree of each dynamic index on the overall risk;

[0072] By accurately capturing information such as acceleration and angular velocity, the system can timely identify abnormal operations of the driver, avoid traffic accidents caused by sudden control mistakes. In the case of increased risk, the system can automatically issue warnings or take intervention measures, such as deceleration or automatic control, to reduce the probability of danger occurrence.

[0073] LiDAR is mainly used to monitor the distance and distribution of obstacles in the surrounding environment, and can effectively help identify potential dangers in the surrounding environment, especially in complex driving environments (such as urban streets or congested areas), R LiDAR is expressed by the following formula:

[0074]

[0075] Wherein, d min represents the distance to the nearest obstacle to the vehicle, N safe represents the number of obstacles within the safe - distance range, and β 1 , β 2 are weight parameters respectively, indicating the influence of the nearest obstacle and the overall environmental density on the risk;

[0076] The system can identify obstacles around the vehicle that may affect driving safety, give early warnings, and help the driver avoid collisions. In some blind spots or environments with poor visibility, the high - precision detection of LiDAR can make up for the driver's visual blind spots and improve overall driving safety.

[0077] The on - vehicle diagnostic system is used to collect the operating state of the vehicle (such as vehicle speed, engine speed, etc.), and R OBD is expressed by the following formula:

[0078]

[0079] Among them, v represents the current vehicle speed, v max represents the road speed limit or the safe vehicle speed, N rpm represents the engine speed, N rpm，max represents the maximum safe speed of the engine, E error represents the fault code detected by the on-vehicle diagnostic system (a binary variable indicating whether there is a fault), γ 1 and γ 2 and γ 3 are weight parameters respectively corresponding to the impacts of speed, rotational speed, and fault code on the risk;

[0080] By detecting the vehicle's operating status and fault information in a timely manner, safety accidents caused by equipment failures (such as speeding, engine problems, etc.) can be avoided. If the on-vehicle diagnostic system detects a fault, it can prompt the driver to repair the vehicle in a timely manner, thereby reducing the sudden risks caused by vehicle failures.

[0081] The driver risk score detects the driver's posture and status through a flexible piezoresistive sensor array, and evaluates their attention and fatigue level, R Driver is represented by the following formula:

[0082] R Driver = δ 1 ·F fatigue + δ 2 ·A attention + δ 3 ·P pose

[0083] Among them, F fatigue represents the driver fatigue score, calculated based on sensor data (such as unchanged long-term posture, abnormal pressure distribution), A attention represents the driver attention score, calculated based on the action response time or the frequency of abnormal behaviors, P pose represents the driver posture score, used to reflect the sitting posture deviation or the state unsuitable for driving, δ 1 and δ 2 and δ 3 are weight parameters respectively indicating the impacts of fatigue, attention, and posture on the risk;

[0084] By analyzing the driver's posture and behavior patterns, it is possible to accurately determine whether the driver is in a state of fatigue driving, and trigger an alarm or provide intervention measures in a timely manner. The system can evaluate the driver's attention level, and when it detects that the driver's attention is distracted, it can remind the driver to stay focused in a timely manner to avoid traffic accidents caused by insufficient attention.

[0085] The total risk score is calculated by the following formula:

[0086] R t = μ 1 ·R IMU + μ 2 ·R LiDAR + μ 3 ·R OBD + μ 4 ·R Driver

[0087] Among them, R t represents the comprehensive risk score, and μ 1 , μ 2 , μ 3 , μ 4 are weight parameters, indicating the contributions of different risk scores to the overall risk. By comprehensively evaluating the overall risk status of the vehicle and the driver, through multi-dimensional risk assessment and considering various potential risks, a more accurate and comprehensive safety assessment can be provided for vehicle safety. According to the comprehensive risk score, the system can make more intelligent intervention decisions. For example, in high-risk situations, the system can combine multiple risk points and take measures such as automatic deceleration, emergency braking, or warning, greatly reducing the probability of accidents.

[0088] In summary, through multi-data fusion and precise algorithm evaluation, this safety assessment module can comprehensively and real-time monitor the driver, vehicle, and environment. Through dynamic monitoring and timely intervention, it can prevent accidents caused by driver errors or external environment changes. By integrating multiple sensor data and making intelligent interventions, it can reduce safety hazards caused by human negligence, can give early warnings before potential risks occur, and take corresponding intervention measures to reduce accident risks.

[0089] Furthermore, the dynamic Bayesian network in step S1 is represented by the following formula:

[0090]

[0091] Among them, X t represents the state vector of the system at time t. For example: X t = [x t，1 , x t，2 ,..., x t，N , where each x t，i is a state variable (such as vehicle speed, acceleration, etc.). X t-1 represents the system state at time t - 1 and serves as the prior input for the state at time t. U t represents the external input of the system at time t, such as driver operations (throttle, brake) or environmental data. P(X t，i |Pa(X t，i )) represents the state Xt，i The conditional probability distribution, conditioned by the parent nodes Pa(X t，i ), where Pa(X t，i ) represents the causally relevant variables of state X t，i (such as X t-1 , U t ). For example, the change in vehicle speed may be determined by the previous speed (inertia effect) and the current acceleration. N is the total number of system state variables.

[0092] It can infer the future system state by observing the historical data of the system (such as vehicle state, driver behavior, etc.) and provide decision support. This method can adjust decisions in a timely manner when the system state changes, improving the real-time response ability.

[0093] Furthermore, the three-dimensional convolutional neural network in step S2 is represented by the following formula:

[0094]

[0095] where h (l+1) (x, y, z) represents the activation value of the neuron in the (l + 1)-th layer after the convolution operation at coordinates (x, y, z). K represents the number of convolutional kernels, and each convolutional kernel extracts different features. ω k (l) represents the weight parameter of the k-th convolutional kernel in the l-th layer, b (l) represents the bias value of the l-th layer, used to adjust the activation value, and k x , k y , k z represent the displacement amounts of the convolutional kernel in the x, y, and z directions.

[0096] The three-dimensional convolutional neural network can effectively process three-dimensional data (such as videos, sensor data, etc.), and is particularly suitable for dynamically changing environments, such as vehicle sensor data (including position, speed, acceleration, etc.). In a traffic management system, the three-dimensional convolutional neural network can be used to analyze in-vehicle camera data, lidar scan data, etc., identify the dynamic state of vehicles, obstacles, etc., and extract features from this information.

[0097] Furthermore, the self-attention mechanism in step S3 is represented by the following formula:

[0098]

[0099] Among them, Q represents the query vector, which represents the information to be concerned, such as the driver abnormal behavior signal in the flexible piezoresistive sensor data; K represents the key vector, which represents the index of the stored information, such as the data from the inertial measurement unit or the on-vehicle diagnostic system, representing the description of the vehicle motion state; V represents the value vector, which represents the content to be extracted from the stored information, such as the dynamic characteristics of the obstacles detected by the lidar, d k Represents the dimension of the key vector K, which is used for normalization scaling to avoid unstable gradients caused by too large numerical values, QK T Represents the similarity matrix between the query and the key, calculates the correlation between different modalities, and softmax is a normalization function that converts the similarity distribution into a probability distribution.

[0100] The self-attention mechanism is a technique for processing sequence data, which enables the model to consider the information of all other positions when processing the information at a certain position, so as to capture long-range dependencies.

[0101] The self-attention mechanism can dynamically extract useful information from different data sources by calculating the similarity between the query vector and the key vector. In the traffic system, it can combine the data of multiple sensors, such as cameras, radars, driver physiological data, etc., extract key information and perform fusion processing

[0102] Furthermore, the reinforcement learning in step S4 is represented by the following formula:

[0103]

[0104] Among them, Q(s,a) represents the value function of taking action a in state s, representing the expected cumulative return; α represents the learning rate, which is used to adjust the balance between the old and new values; r is the immediate reward, which measures the quality of the current action. For example, a positive reward can be obtained for avoiding collisions; a negative reward can be obtained for abnormal driving; γ is the discount factor, which represents the importance of future rewards, and its value range is [0,1]. Represents the value of the best action that may be taken in the new state s′, s represents the current state, and s′ represents the next state. For example, the current state s: the vehicle is driving at high speed; the next state s′: the vehicle decelerates after passing a sharp turn.

[0105] The application of reinforcement learning in the traffic management system can enable the system to continuously optimize decisions according to environmental feedback. For example, the system can adjust the vehicle driving strategy (such as accelerating, decelerating, changing lanes, etc.) by continuously simulating factors such as the driver's behavior and environmental changes to achieve the balance between safety and efficiency.

[0106] In summary, through the combination of dynamic Bayesian networks, 3D convolutional neural networks, self-attention mechanisms, and reinforcement learning, the system can comprehensively and real-time process and analyze various complex data in traffic management. These technologies provide powerful data processing capabilities for traffic safety systems, achieving the efficient operation of intelligent transportation systems from state prediction, feature extraction, information fusion to intelligent decision optimization.

[0107] Furthermore, in the sensing module, the flexible piezoresistive sensor array estimates the driver's blood pressure value based on the pulse wave transit time algorithm, and the formula is as follows:

[0108] BP = k 1 ·ln(PTT) + k 2 ·BMI + c

[0109] where BP is the blood pressure value, ln(PTT) represents the natural logarithm of the pulse wave transit time, PTT is the pulse wave transit time, which refers to the time required for the pulse wave to travel from the heart to the end of the blood vessel when the heart contracts, BMI is the body mass index, k 1 and k 2 are fitting coefficients, which are constants obtained through experiments or data fitting. They help adjust the influence degrees of the pulse wave transit time (PTT) and the body mass index (BMI) on blood pressure. k 1 controls the influence of PTT on blood pressure, k 2 controls the influence of BMI on blood pressure, and c is a calibration constant used to correct the prediction result of the model to make it closer to the actual blood pressure value. This value can be adjusted by comparing with the actual blood pressure measurement data. This blood pressure estimation method based on PTT and BMI has advantages such as non-invasiveness and strong real-time performance, and is very suitable for traffic monitoring and driver health management, providing a non-invasive and efficient way to monitor the driver's health status, thereby improving road safety.

[0110] Furthermore, the edge module uses the MEC gateway to assign QoS levels to multi-type data and optimize network transmission performance using the 5G-TSN protocol. The MEC gateway, based on the time-sensitive network protocol, distributes data streams into high, medium, and low three types of QoS levels according to time priorities.

[0111] Among them, in the transportation safety management system, the edge module uses the MEC gateway to achieve efficient management and scheduling of data streams. Especially in the context of a large amount of data exchange between in-vehicle devices and the backend system, it ensures that the system can transmit key information in real-time and reliably;

[0112] There are various data sources in the transportation system, such as vehicle movement data, driver physiological data, environmental data, etc. The real-time nature and importance of these data are different, so targeted optimization transmission is required;

[0113] The MEC gateway uses the 5G-TSN protocol to allocate the received data stream to different QoS (Quality of Service) levels according to its time sensitivity and importance. In this system, data is divided into three main categories:

[0114] High-priority QoS: For data with extremely high timeliness requirements, such as driver physiological data and emergency alert data, the highest-priority QoS is used for transmission to ensure that such data will not affect safety due to network latency;

[0115] Medium-priority QoS: Such data as vehicle status and road traffic conditions have relatively high real-time requirements, but can be appropriately delayed in case of emergency, with medium priority;

[0116] Low-priority QoS: Such as ordinary traffic monitoring data and environmental data, the transmission requirements of these data are relatively low, and they can be transmitted when the network load is light.

[0117] Furthermore, the linkage intervention module embeds a biological data management module based on the consortium blockchain evidence storage mechanism. The key records include the driver's heart rate, blood pressure, breathing rate, and abnormal warning events, which are written into the consortium blockchain in real time to ensure the data cannot be tampered with. The consortium blockchain evidence storage module can synchronize multi-node updates of the key records and automatically determine whether to initiate warning or dispatching measures based on smart contracts.

[0118] The characteristics of the consortium blockchain are that the data is not stored in a single node, but is distributed and stored in multiple network nodes. Each node verifies the data to ensure its immutability, which is crucial for the transparency and trust of the traffic management system. Especially when it comes to driver physiological data and key warning events, the credibility of the data is of utmost importance. One advantage of the consortium blockchain is that the data can be synchronously updated among multiple nodes to ensure the real-time and consistency of biological data. Whether the data is generated in roadside devices or in-vehicle devices, once written into the consortium blockchain, all nodes can quickly obtain and update the data, thus ensuring the integrity of the data. Since the data structure of the consortium blockchain is a chain structure, and each data record is encrypted and verified, it cannot be tampered with. Even in case of security threats or improper operations, the data cannot be forged or deleted, thus ensuring the authenticity of the driver physiological data.

[0119] In addition, as an automated execution contract system, smart contracts can automatically execute specific judgments and operations based on the evidence storage data. In the present invention, the smart contract based on the consortium blockchain can automatically determine whether to initiate warning or dispatching measures according to the driver's physiological data and abnormal events. Specifically:

[0120] The smart contract monitors the driver's health data. When an anomaly is detected (such as a too high heart rate, too high blood pressure, etc.), it automatically triggers an alarm to notify the relevant management system or directly notify the driver. In extreme cases, the smart contract can also automatically initiate dispatching measures, such as instructing the vehicle to automatically decelerate, guiding the vehicle to the nearest safe parking area, triggering automated driving control, etc. All these decisions are automatically executed by the smart contract, reducing the delay of human intervention and enhancing the response speed and safety.

[0121] Finally, it should be noted that the above are only the preferred embodiments of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments or perform equivalent replacements for some of the technical features. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principle of the present invention shall be included within the protection scope of the present invention.

Claims

1. An integrated transportation safety management platform for real-time data interaction and supervision, characterized by: It includes perception module, edge computing module, security assessment module, decision-making module and linkage intervention module; The perception module includes an inertial measurement unit for collecting traffic data from a vehicle terminal, a laser radar, an on-board diagnostic system, and a flexible piezoresistive sensor array embedded in the driver's seat for collecting physiological characteristic data of the driver; The edge computing module is used to perform real-time data cleaning, feature extraction and redundancy removal; The safety assessment module uses a driver monitoring algorithm that integrates multiple collected data to evaluate the driver's status in real time and trigger intervention actions in abnormal situations; The decision module integrates data from the perception module and the edge computing module to perform dynamic risk assessment and generate an intelligent decision chain; The decision chain generation process is as follows: S1, the dynamic Bayesian network unifies the data collected by the inertial measurement unit, lidar, on-board diagnostic system and flexible piezoresistive sensor array to model the evolution of the system state over time. The current state of the system is determined by the state at the previous moment and the current external input data, which is used to evaluate dynamic risks; S2, the 3D convolutional neural network performs convolution operations on the time and space dimensions of the input collected data and extracts key features for detecting obstacles and path risks; S3, combines different collected data through the attention mechanism, uniformly analyzes the vehicle and driver status, and automatically focuses on the important parts of different collected data; S4, using reinforcement learning to dynamically optimize risk response strategies; The linkage intervention module obtains the driver's physiological signals and the vehicle's operating status based on the safety assessment module and the decision-making module. When the driver's physiological signals or the vehicle's operating status are abnormal, the electronic horizon warning, automatic speed limit, and dynamic navigation are triggered.

2. The integrated transportation safety management platform for real-time data interaction and supervision according to claim 1 is characterized in that: In the safety assessment module, the risk score includes the inertial measurement unit risk score R IMU 、LiDAR risk score R LiDAR , risk score of on-board diagnostic system R OBD and the driver risk score R Driver .

3. The integrated transportation safety management platform for real-time data interaction and supervision according to claim 1 is characterized in that: The sources of each risk score are quantitatively evaluated through the following specific algorithm formula: Inertial measurement units are mainly used to monitor vehicle dynamic behavior. IMU It is expressed by the following formula: R IMU =α1·||a x |+α2|a y |+α3·|ω z | Among them, α x Expressed as the acceleration of the vehicle along the front and rear (x-axis) direction, α y Expressed as the vehicle's lateral acceleration (y-axis), ω z It is expressed as the angular velocity of the vehicle along the vertical axis (used to detect rotation or slippage), α1, α2, and α3 are weight parameters, reflecting the influence of each dynamic indicator on the overall risk; LiDAR is mainly used to monitor the distance and distribution of obstacles in the surrounding environment. LiDAR It is expressed by the following formula: Among them, d min Expressed as the distance to the nearest obstacle to the vehicle, N safe It is expressed as the number of obstacles within the safe distance range, β1 and β2 are weight parameters, which represent the impact of the nearest obstacle and the overall environment density on the risk; The on-board diagnostic system is used to collect vehicle operating status, R OBD It is expressed by the following formula: Where, v represents the current vehicle speed, v max Indicates the road speed limit or safe speed, N rpm Expressed as engine speed, N rpm，max Expressed as the maximum safe speed of the engine, E error It represents the fault code detected by the on-board diagnostic system, γ1, γ2, and γ3 are weight parameters, corresponding to the impact of speed, rotation speed, and fault code on risk, respectively; The driver risk score detects the driver's posture and state through a flexible piezoresistive sensor array to assess their attention and fatigue level. Driver It is expressed by the following formula: R Driver =δ1·F fatigue +δ2·A attention +δ3·P pose Among them, F fatigue Expressed as the driver fatigue score, calculated based on sensor data, A attention Expressed as a driver attention score, calculated based on action response time or abnormal behavior frequency, P pose It is expressed as a driver posture score, which is used to reflect the state of sitting posture deviation or unsuitable driving. δ1, δ2, and δ3 are weight parameters, representing the effects of fatigue, attention, and posture on risk.

4. The integrated transportation safety management platform for real-time data interaction and supervision according to claim 3 is characterized in that: The total risk score is calculated using the following formula: R t =μ1·R IMU +μ2·R LiDAR +μ3·R OBD +μ4·R Driver Among them, R t It is expressed as a comprehensive risk score, where μ1, μ2, μ3, and μ4 are weight parameters, indicating the contribution of different risk scores to the overall risk.

5. The integrated transportation safety management platform for real-time data interaction and supervision according to claim 1 is characterized in that: The dynamic Bayesian network in step S1 is expressed using the following formula: Among them, X t represents the state vector of the system at time t, X t-1 It is represented as the system state at time t-1, and as the prior input of the state at time t, U t Expressed as the external input of the system at time t, P(X t，i |Pa(X t，i )) is represented by state X t，i The conditional probability distribution of the parent node Pa(X t，i ) determines, Pa(X t，i ) is represented as state X t，i are the causally related variables, and N is the total number of system state variables.

6. The integrated transportation safety management platform for real-time data interaction and supervision according to claim 5 is characterized in that: The three-dimensional convolutional neural network in step S2 is expressed using the following formula: Among them, h (l+1) (x, y, z) represents the activation value of the l+1th layer neuron at the coordinate (x, y, z) after the convolution operation, K represents the number of convolution kernels, each convolution kernel extracts different features, ω k (l) It is represented as the weight parameter of the kth convolution kernel in the lth layer, b (l) Represented as the bias value of the lth layer, used to adjust the activation value, k x , k y , k z It is expressed as the displacement of the convolution kernel in the x, y, and z directions.

7. The integrated transportation safety management platform for real-time data interaction and supervision according to claim 6 is characterized in that: The self-attention mechanism in step S3 is expressed using the following formula: Among them, Q is the query vector, which indicates the information that needs attention, K is the key vector, which indicates the index of the stored information, and V is the value vector, which indicates the content that needs to be extracted from the stored information. k Represented as the dimension of the key vector K, used for normalization and scaling to avoid gradient instability caused by excessive values, QK T It is represented as a similarity matrix between the query and the key, and the correlation between different modalities is calculated. Softmax is a normalization function that converts the similarity distribution into a probability distribution.

8. The integrated transportation safety management platform for real-time data interaction and supervision according to claim 7 is characterized in that: The reinforcement learning in step S4 is expressed using the following formula: Among them, Q(s,a) is the value function of taking action a in state s, which represents the expected cumulative return. α is the learning rate, which is used to adjust the balance between new and old values. r is the immediate reward, which measures the pros and cons of the current action. γ is the discount factor, which represents the importance of future rewards and has a value range of [0,1]. It is expressed as the value of the best action that may be taken in the new state s′, s represents the current state, and s′ represents the next state. For example, the current state s: the vehicle is traveling at high speed; the next state s′: the vehicle slows down after a sharp turn.

9. The integrated transportation safety management platform for real-time data interaction and supervision according to claim 1 is characterized in that: In the sensing module, the flexible piezoresistive sensor array estimates the driver's blood pressure value based on the pulse wave transmission time algorithm.

10. The integrated transportation safety management platform for real-time data interaction and supervision according to claim 1 is characterized in that: The edge module uses the MEC gateway to adopt the 5G-TSN protocol to assign QoS levels to multiple types of data and optimize network transmission performance. The MEC gateway is based on the time-sensitive network protocol to assign data streams to three QoS levels: high, medium, and low according to time priority.

Images