Motor vehicle safety performance classification method and system based on multi-factor coupling influence

Abstract

The invention relates to the technical field of data processing, and discloses a motor vehicle safety performance classification method and system based on multi-factor coupling influence. The method comprises the steps of collecting multidimensional safety data of a motor vehicle, calculating a mutual information value and transfer entropy to generate a coupling relation network diagram, constructing a three-layer safety index system to form a comprehensive score, obtaining low-dimensional features through phase-space mapping, determining a dynamic boundary value by applying density clustering, and classifying real-time states to output risk early warning information. According to the technical method for realizing accurate classification and early warning, the defects of neglect of factor coupling relation, lack of scene adaptability, insufficient dynamic prediction capability and the like in the prior art are overcome, and the accuracy and practicability of safety performance evaluation and early warning of the motor vehicle are improved.

Description

Technical Field

[0001] This application relates to the technical field of data processing, and particularly to a classification method and system for the safety performance of motor vehicles based on the coupled influence of multiple factors. Background Art

[0002] Traditional methods for evaluating the safety performance of motor vehicles mainly rely on static tests and single-parameter analysis, such as crash safety tests, braking distance tests, and skid tests. These methods usually evaluate the performance of a certain aspect of a motor vehicle in isolation under specific conditions, and it is difficult to comprehensively reflect the complex situation of the combined action of multiple factors in the actual road environment. In recent years, with the development of sensor technology and in-vehicle electronic systems, some advanced safety performance evaluation systems have begun to adopt multi-parameter fusion methods, such as electronic stability control systems (ESC), anti-lock braking systems (ABS), and adaptive cruise control systems (ACC). These systems can monitor multiple parameters of a vehicle in real time and make corresponding controls, but their evaluation and warning mechanisms are still based on preset fixed thresholds and simple linear relationships.

[0003] However, there are obvious deficiencies in the prior art in the evaluation and classification of the safety performance of motor vehicles. First, most methods ignore the coupling effect between various safety influencing factors, simply adding or averaging different parameters, and cannot accurately reflect the non-linear interaction effects between factors. Second, traditional methods mostly use fixed thresholds to judge the safety state, lacking the adaptability to different driving scenarios and environmental conditions, resulting in false alarms or missed alarms under complex conditions. Third, most existing systems focus on the evaluation of the immediate state, lacking the ability to perform temporal analysis on the dynamic evolution process of the safety state, and it is difficult to predict potential risks in advance. Finally, most methods focus on the safety performance evaluation of a single dimension, such as braking performance or lateral stability, lacking an overall evaluation framework that comprehensively considers multi-dimensional safety factors and cannot comprehensively reflect the actual safety state of a motor vehicle. Summary of the Invention

[0004] This application provides a classification method and system for the safety performance of motor vehicles based on the coupled influence of multiple factors, which is used to implement a technical method for accurate classification and warning, so as to overcome the defects in the prior art such as ignoring the coupling relationship between factors, lacking scenario adaptability, and insufficient dynamic prediction ability, and improve the accuracy and practicality of the evaluation and warning of the safety performance of motor vehicles.

[0005] In a first aspect, the present application provides a method for classifying the safety performance of a motor vehicle based on the coupled influence of multiple factors. The method for classifying the safety performance of a motor vehicle based on the coupled influence of multiple factors includes: collecting the dynamic parameters, environmental conditions, and driving behavior data of the motor vehicle through a distributed sensor network to obtain the original data of multi-dimensional safety influencing factors; calculating the mutual information value and transfer entropy for the original data of the multi-dimensional safety influencing factors to generate a non-linear factor coupling relationship network diagram; constructing a three-layer structure safety index system based on the non-linear factor coupling relationship network diagram to form a comprehensive safety performance score of the motor vehicle; converting the comprehensive safety performance score of the motor vehicle through phase space mapping and extracting key features to obtain a low-dimensional discriminant feature set; applying an adaptive density clustering method to the low-dimensional discriminant feature set to determine the boundary values of the dynamic safety level; and using the boundary values of the dynamic safety level to perform multi-scenario classification on the real-time collected motor vehicle state and output a classified safety risk warning message.

[0006] In a second aspect, the present application provides a system for classifying the safety performance of a motor vehicle based on the coupled influence of multiple factors. The system for classifying the safety performance of a motor vehicle based on the coupled influence of multiple factors includes:

[0007] A collection module, configured to collect the dynamic parameters, environmental conditions, and driving behavior data of the motor vehicle through a distributed sensor network to obtain the original data of multi-dimensional safety influencing factors;

[0008] A calculation module, configured to calculate the mutual information value and transfer entropy for the original data of the multi-dimensional safety influencing factors to generate a non-linear factor coupling relationship network diagram;

[0009] A construction module, configured to construct a three-layer structure safety index system based on the non-linear factor coupling relationship network diagram to form a comprehensive safety performance score of the motor vehicle;

[0010] A conversion module, configured to convert the comprehensive safety performance score of the motor vehicle through phase space mapping and extract key features to obtain a low-dimensional discriminant feature set;

[0011] A clustering module, configured to apply an adaptive density clustering method to the low-dimensional discriminant feature set to determine the boundary values of the dynamic safety level;

[0012] A classification module, configured to use the boundary values of the dynamic safety level to perform multi-scenario classification on the real-time collected motor vehicle state and output a classified safety risk warning message.

[0013] In a third aspect, a classification device for the safety performance of a motor vehicle based on the coupled influence of multiple factors is provided, including: a memory and at least one processor, wherein instructions are stored in the memory; the at least one processor calls the instructions in the memory so that the classification device for the safety performance of a motor vehicle based on the coupled influence of multiple factors executes the above-mentioned classification method for the safety performance of a motor vehicle based on the coupled influence of multiple factors.

[0014] In a fourth aspect, a computer-readable storage medium is provided, in which instructions are stored, and when it runs on a computer, it causes the computer to execute the above-mentioned classification method for the safety performance of a motor vehicle based on the coupled influence of multiple factors.

[0015] In the technical solution provided by this application, by collecting the dynamic parameters, environmental conditions, and driving behavior data of motor vehicles through a distributed sensor network, the original data of multi-dimensional safety influencing factors is obtained, realizing the all-round and multi-angle perception of the safety state of motor vehicles, and providing a rich basic data source for safety performance classification; by calculating the mutual information value and transfer entropy for the original data of multi-dimensional safety influencing factors, a non-linear factor coupling relationship network diagram is generated, accurately capturing and quantifying the complex non-linear interaction between different safety factors, breaking through the limitation of traditional methods that ignore factor coupling; according to the non-linear factor coupling relationship network diagram, a three-layer structure safety index system is constructed to form a comprehensive score of the safety performance of motor vehicles, establishing a complete index link from the original data to the final evaluation, making the evaluation result have a hierarchical interpretability; the comprehensive score of the safety performance of motor vehicles is converted through phase space mapping and key features are extracted to obtain a low-dimensional discriminant feature set, applying non-linear dynamics analysis methods to solve the problem of effective expression of high-dimensional data, improving the calculation efficiency and accuracy of classification; applying the adaptive density clustering method to the low-dimensional discriminant feature set to determine the boundary value of the dynamic safety level, and realizing the adaptive classification of the safety state through the artificial intelligence density clustering algorithm, overcoming the rigidity problem of traditional fixed threshold classification; using the boundary value of the dynamic safety level to classify the real-time collected motor vehicle state in multiple scenarios, and outputting hierarchical safety risk warning information, realizing the accurate assessment and timely warning of safety risks. The artificial intelligence algorithm features in this solution, especially the calculation of mutual information and transfer entropy, phase space mapping, adaptive density clustering, etc., have made key contributions to the solution: the mutual information and transfer entropy algorithms can discover and quantify non-linear association patterns from a large amount of original data, breaking through the limitation of traditional linear analysis; the phase space mapping and non-linear dynamics feature extraction algorithms can capture the dynamic evolution law of the safety state of motor vehicles, enhancing the prediction ability of the system; the adaptive density clustering algorithm realizes the dynamic adjustment of the safety state boundary, improving the adaptability and accuracy of classification in different scenarios. The organic combination of these algorithms forms a complete intelligent analysis framework, enabling the motor vehicle safety performance classification method to possess the comprehensive capabilities of multi-dimensional perception, coupling analysis, dynamic adaptation, and accurate warning, significantly improving the comprehensiveness and accuracy of the motor vehicle safety state assessment. Description of the Drawings

[0016] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following will briefly introduce the drawings required for the description of the embodiments. Obviously, the drawings in the following description are some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can also be obtained based on these drawings.

[0017] Figure 1 It is a schematic diagram of an embodiment of the motor vehicle safety performance classification method based on multi-factor coupling influence in the embodiment of this application;

[0018] Figure 2 Schematic diagram of an embodiment of a motor vehicle safety performance classification system based on multi - factor coupling influence in an embodiment of the present application;

[0019] Figure 3 Structural schematic block diagram of a motor vehicle safety performance classification device based on multi - factor coupling influence in an embodiment of the present invention. Detailed implementation manners

[0020] The embodiments of the present application provide a method and system for classifying the safety performance of motor vehicles based on multi - factor coupling influence. Terms such as "first", "second", "third", "fourth", etc. (if any) in the specification, claims and the above - mentioned drawings of the present application are used to distinguish similar objects, and do not necessarily describe a specific order or sequence. It should be understood that such data can be interchanged under appropriate circumstances so that the embodiments described here can be implemented in an order different from that shown or described here. In addition, the term "comprising" or "having" and any deformation thereof are intended to cover non - exclusive inclusion. For example, a process, method, system, product or device that includes a series of steps or units does not necessarily limit to those steps or units clearly listed, but may include other steps or units not clearly listed or inherent to these processes, methods, products or devices.

[0021] For ease of understanding, the specific process of the embodiments of the present application is described below. Please refer to Figure 1 An embodiment of the method for classifying the safety performance of motor vehicles based on multi - factor coupling influence in the embodiments of the present application includes:

[0022] Step S101: Collect motor vehicle dynamic parameters, environmental conditions and driving behavior data through a distributed sensor network to obtain raw multi - dimensional safety influence factor data;

[0023] Step S102: Calculate the mutual information value and transfer entropy for the raw multi - dimensional safety influence factor data to generate a non - linear factor coupling relationship network diagram;

[0024] Step S103: Construct a three - layer structure safety index system based on the non - linear factor coupling relationship network diagram to form a comprehensive safety performance score of the motor vehicle;

[0025] Step S104: Convert the comprehensive safety performance score of the motor vehicle through phase - space mapping and extract key features to obtain a low - dimensional discriminant feature set;

[0026] Step S105: Apply the adaptive density clustering method to the low - dimensional discriminant feature set to determine the dynamic safety level boundary value;

[0027] Step S106: Classify the real-time collected motor vehicle states under multiple scenarios using dynamic safety level boundary values, and output hierarchical safety risk warning information.

[0028] It can be understood that the execution entity of this application can be a motor vehicle safety performance classification system based on the coupled influence of multiple factors, or it can also be a terminal or a server. Specifically, no limitation is made here. In this embodiment of the application, the server is used as the execution entity for illustration.

[0029] In this embodiment of the application, the motor vehicle dynamic parameters, environmental conditions, and driving behavior data are collected through a distributed sensor network to obtain the original data of multi-dimensional safety influencing factors. Specifically, the vehicle dynamic parameter sensors collect acceleration, yaw rate, and sideslip angle, the environmental monitoring sensors collect rainfall, light intensity, and road surface friction coefficient, and the driving behavior monitoring device collects steering wheel angle, pedal pressure, and driver attention distribution. These data are transmitted to the central processing unit through the in-vehicle high-speed CAN bus network and are subjected to sensor drift compensation and system error correction through a real-time data calibration algorithm. During the sensor drift compensation process, a reference value is set according to the stability characteristics of different sensors. When the sensor readings deviate for non-physical reasons over time, the drift amount is calculated by comparing the current value with the historical reference value and automatically compensated into the measured value. At the same time, the system also adopts an adaptive sampling frequency adjustment mechanism to optimize the data acquisition according to the change characteristics of different parameters. For example, when the vehicle is in a high-speed turning state, the sampling frequency of the yaw rate will automatically increase to more than 100 Hz to capture the rapidly changing dynamic characteristics. Calculate the mutual information value and transfer entropy for the original data of multi-dimensional safety influencing factors to generate a non-linear factor coupling relationship network diagram. This step first performs outlier detection and smoothing processing to obtain a preprocessed data set. Then, for the factor pairs in the preprocessed data set, the joint probability distribution is calculated through the kernel density estimation method to obtain the mutual information value matrix between factors. The mutual information value characterizes the degree of mutual dependence between different safety factors, and the higher the value, the stronger the correlation between factors. Subsequently, the preprocessed data set is segmented based on a sliding time window, and the transfer entropy between factors is calculated through symbolic transformation to form a directional coupling strength matrix. The transfer entropy describes the directionality of information flow and reveals the causal relationship between factors. After the mutual information value matrix and the directional coupling strength matrix are fused, a multi-level factor coupling network topology structure is constructed, and the redundant connections are trimmed through a sparse representation learning method to retain the key coupling paths. Finally, the characteristics of the multi-level factor coupling network topology structure are analyzed from three scales: micro, meso, and macro to generate a non-linear factor coupling relationship network diagram.

[0030] Construct a three - layer structure safety index system according to the non - linear factor coupling relationship network diagram to form a comprehensive score of motor vehicle safety performance. This step extracts basic safety indicators from the original data of multi - dimensional safety influencing factors and constructs the bottom - layer structure of the index system. The basic safety indicators directly reflect the basic safety state of the vehicle, such as braking distance, lateral stability, etc. Subsequently, according to the coupling path strength in the non - linear factor coupling relationship network diagram, identify key coupling factor combinations and generate middle - layer coupling indicators, such as composite indicators like "braking - steering coupling stability index". Calculate the information entropy for the middle - layer coupling indicators to obtain the index weight distribution coefficient. The greater the information entropy of an indicator, the lower its weight because it contains more uncertain information. Combine the middle - layer coupling indicators with the index weight distribution coefficient and form the top - layer comprehensive safety performance index through fuzzy integral operation. Calibrate the top - layer comprehensive safety performance index according to the characteristics of different driving scenarios to obtain the scenario - adaptable safety index value, and then standardize it to form a comprehensive score of motor vehicle safety performance. Convert the comprehensive score of motor vehicle safety performance through phase - space mapping and extract key features to obtain a low - dimensional discriminant feature set. First, construct a time - series data matrix for the comprehensive score of motor vehicle safety performance and determine the optimal time - delay parameter through the mutual information method. The mutual information method calculates the statistical dependence between the sequence and its delayed version at different time - delays and selects the delay corresponding to the first local minimum of the mutual information value as the optimal value. Then use the false nearest neighbor method to analyze the time - series data matrix and determine the optimal embedding dimension. The false nearest neighbor method checks the change of nearest neighbors under increasing embedding dimensions, and the dimension when the false nearest neighbor ratio drops below the threshold is the optimal embedding dimension. According to the optimal time - delay parameter and the optimal embedding dimension, reconstruct the comprehensive score of motor vehicle safety performance into a high - dimensional phase - space trajectory. Subsequently, calculate the non - linear dynamic characteristic parameters of the high - dimensional phase - space trajectory, including the maximum Lyapunov exponent, correlation dimension, and approximate entropy value, extract complex network features from the high - dimensional phase - space trajectory to obtain network topology feature quantization indicators. Combine the non - linear dynamic characteristic parameters and the network topology feature quantization indicators and obtain a low - dimensional discriminant feature set through non - linear mapping transformation.

[0031] Apply the adaptive density clustering method to the low - dimensional discriminant feature set to determine the dynamic safety - level boundary value. The specific process includes calculating the local density value and distance factor of each feature point in the low - dimensional discriminant feature set, constructing a density - distance decision graph, and identifying the initial clustering center points from the density - distance decision graph through the density peak detection method. Conduct clustering division according to the distribution characteristics of the initial clustering center points to obtain the initial safety - level clusters. Evaluate the effectiveness of the initial safety - level clusters and determine the optimal number of clusters through silhouette coefficient calculation. Construct a closed decision surface around the boundary of the safety - level cluster corresponding to the optimal number of clusters to form a static safety - level boundary. Analyze the temporal variation characteristics of the safety - level clusters based on historical data, establish boundary dynamic update rules, and generate the dynamic safety - level boundary value.

[0032] Classify the real-time collected motor vehicle status in multiple scenarios using dynamic safety level boundary values, and output hierarchical safety risk warning information. This step first receives the motor vehicle dynamic parameters, environmental conditions, and driving behavior data collected by the distributed sensor network through the real-time data acquisition interface, and constructs a real-time status data matrix. Apply a fast feature extraction algorithm to the real-time status data matrix to obtain a real-time feature vector. Project the real-time feature vector into the feature space where the low-dimensional discriminant feature set is located through the phase space mapping transformation method to generate a real-time low-dimensional discriminant point. Calculate the positional relationship between the real-time low-dimensional discriminant point and the dynamic safety level boundary value to determine the current motor vehicle safety status classification category. Based on the current motor vehicle safety status classification category and the dynamic change trend of the real-time low-dimensional discriminant point, construct a safety risk level. Generate hierarchical safety risk warning information according to the safety risk level and the preset risk threshold comparison table.

[0033] For example, in a turning scenario under wet road surface conditions on a highway, the distributed sensor network collects that the yaw rate of the vehicle is 0.3 rad / s, the sideslip angle is 4 degrees, the road surface friction coefficient is 0.6, and the steering wheel angle is 30 degrees. After processing the recognition of the coupling relationship of these original data, it is found that the mutual information value between the yaw rate and the sideslip angle reaches 0.85, indicating a high correlation between the two. The comprehensive safety performance score calculated through the three-layer index system is 75 points. After this score is mapped through the phase space, the obtained low-dimensional discriminant feature is located in a certain area of the feature space, and the distance relationship calculated with the dynamic safety level boundary value shows that the current state is close to the boundary of the "medium risk" area. The system generates a yellow warning signal accordingly, reminding the driver that the dynamic stability of the vehicle has begun to decline, and it is recommended to reduce the speed and avoid sharp steering operations.

[0034] In a specific embodiment, the process of executing step S101 may specifically include the following steps:

[0035] Collect vehicle motion state data including acceleration, yaw rate, and sideslip angle through vehicle dynamic parameter sensors;

[0036] Collect environmental state data including rainfall, light intensity, and road surface friction coefficient through environmental monitoring sensors;

[0037] Collect driving operation data including steering wheel angle, pedal pressure, and driver attention distribution through driving behavior monitoring devices;

[0038] Transmit the vehicle motion state data, environmental state data, and driving operation data to the central processing unit through the in-vehicle high-speed CAN bus network;

[0039] The collected data is compensated for sensor drift and corrected for system errors using a real-time data calibration algorithm;

[0040] According to the varying characteristics of different parameters, an adaptive sampling frequency adjustment mechanism is adopted to optimize the acquisition of vehicle motion state data, environmental state data, and driving operation data, obtaining raw data of multi-dimensional safety influencing factors.

[0041] Specifically, vehicle motion state data including acceleration, yaw rate, and sideslip angle is collected by vehicle dynamic parameter sensors. The vehicle dynamic parameter sensors consist of multiple acceleration sensors, angular velocity sensors, and sideslip angle measuring devices. Among them, the acceleration sensors are installed near the vehicle's center of gravity to collect acceleration information in three axial directions, with the unit of m / s 2 , and the data range is usually ±10g; the yaw rate sensor is installed at the center of the vehicle chassis to measure the rotation rate of the vehicle around the vertical axis, with the unit of rad / s and a typical measurement range of ±100° / s; the sideslip angle is measured indirectly by combining optical or inertial measurement devices with vehicle speed sensor data, representing the angle between the actual motion direction of the vehicle and the vehicle head orientation, with the unit of degree and an effective range usually of ±15°. The raw signals collected by the sensors are preliminarily filtered to generate data packets in a standard format, including measurement values, timestamps, and sensor status information. At the same time, environmental monitoring sensors collect environmental state data including rainfall, light intensity, and road surface friction coefficient. The environmental monitoring sensor network consists of a rainfall sensor, a light intensity sensor, and a road surface friction coefficient detector. The rainfall sensor is usually installed in the front windshield area to quantify the rainfall intensity by detecting the number and distribution density of water droplets on the glass surface, and the data is represented in discrete levels (0 - 5), where 0 indicates no rain and 5 indicates heavy rain; the light intensity sensor is located at the top of the vehicle or the rearview mirror position to measure the ambient illuminance, with the unit of lux and a measurement range from 0 (complete darkness) to 100,000 lux (strong sunlight); the road surface friction coefficient detector estimates the road adhesion by analyzing the contact characteristics between the tires and the road surface or using a dedicated detector, and the numerical range is usually from 0.1 (extremely slippery) to 1.0 (dry asphalt road surface). The data collected by these sensors undergoes preliminary digital processing to form environmental state data packets.

[0042] The driving behavior monitoring device collects driving operation data including steering wheel angle, pedal pressure, and driver attention distribution. The steering wheel angle sensor is installed on the steering column to measure the rotation angle of the steering wheel, with a data range of ±720° and a typical accuracy of 0.1°. The pedal pressure sensors are respectively installed in the mechanisms of the accelerator pedal, brake pedal, and clutch pedal (if any) to measure the pressure or displacement applied by the driver to the pedal, and the data is usually normalized to a pedal stroke of 0 - 100%. The driver attention monitoring system consists of an in-vehicle camera and an eye movement tracking device, which capture the driver's facial expressions, eye movements, and head postures, and analyze the attention distribution through image processing algorithms. The output data includes parameters such as fixation point coordinates, fixation duration, and pupil diameter changes. These different types of data are transmitted to the central processing unit through the in-vehicle high-speed CAN bus network. The CAN bus is a multi-master serial communication network protocol with a working frequency of 500 kbps to 1 Mbps and a strong anti-interference ability using differential signal transmission. During data transmission, each sensor node packs the measured data into a standard CAN frame format. Each CAN frame contains a frame ID (indicating data priority and type), data length, data field, and check bit. After receiving the CAN frame, the central processing unit parses different types of sensor data according to the frame ID and reorganizes them into a complete data stream in chronological order. For example, the yaw rate data may be transmitted as a CAN message with a frame ID of 0x220 and sent once every 10 ms, while the driver attention data may be transmitted as a message with a frame ID of 0x380 and updated once every 100 ms.

[0043] Next, a real-time data calibration algorithm is used to perform sensor drift compensation and system error correction on the collected data. Sensor drift refers to the phenomenon that the sensor output slowly changes over time and deviates from the true value, mainly caused by factors such as temperature changes, vibration accumulation, and component aging. The real-time data calibration algorithm first identifies the drift baseline through zero-point detection under static conditions. For example, when the vehicle is stationary, the horizontal component of the acceleration sensor should be 0, and the yaw rate should be 0. The detected deviation value is recorded as the drift compensation value. Under dynamic conditions, the algorithm performs cross-check through multi-sensor data fusion. For example, by comparing the GPS speed data and the wheel speed sensor data, the system error of the wheel speed sensor is identified. The calibration process uses Kalman filtering or adaptive filtering techniques, combined with the historical performance characteristics of the sensor, to dynamically adjust the compensation coefficient to make the calibrated data closer to the true physical quantity. For example, when the yaw rate sensor detects an offset value of 0.02 rad / s in the stationary state, the system will automatically subtract this offset from the subsequent measurement values.

[0044] According to the varying characteristics of different parameters, an adaptive sampling frequency adjustment mechanism is adopted to optimize the acquisition of vehicle motion state data, environmental state data, and driving operation data, obtaining the original data of multi-dimensional safety influencing factors. The adaptive sampling frequency adjustment mechanism dynamically adjusts the sensor sampling frequency based on the parameter change rate and safety criticality. The algorithm first calculates the change rate of the data, that is, the difference between consecutive sampling points divided by the time interval. When the change rate exceeds the preset threshold, the system automatically increases the sampling frequency of this parameter; when the parameter remains stable for a period of time, the system reduces the sampling frequency to save computing resources. At the same time, different parameters are assigned different base sampling frequencies and adjustment ranges according to their safety criticality. For example, during normal driving, the sampling frequency of the yaw rate may be 20Hz, but when a sharp turn operation is detected, the sampling frequency will automatically increase to 100Hz to capture the rapidly changing vehicle body dynamics; while the environmental light intensity, as a parameter with slower changes, may have a base sampling frequency of only 1Hz and will only increase to 5Hz even in the case of relatively rapid light changes. In this way, the system optimizes the efficiency and accuracy of data acquisition, forming an original database of multi-dimensional safety influencing factors containing timestamps, numerical values, and quality markers.

[0045] In a specific embodiment, the process of executing step S102 may specifically include the following steps:

[0046] Perform outlier detection and smoothing processing on the original data of multi-dimensional safety influencing factors to obtain a preprocessed data set;

[0047] For the factor pairs in the preprocessed data set, calculate the joint probability distribution through kernel density estimation to obtain the mutual information value matrix between factors;

[0048] Segment the preprocessed data set based on a sliding time window, calculate the transfer entropy between factors through symbolic transformation, and form a directional coupling strength matrix;

[0049] Fuse the mutual information value matrix and the directional coupling strength matrix to construct a multi-level factor coupling network topology structure;

[0050] Use the sparse representation learning method to prune the redundant connections of the multi-level factor coupling network topology structure and retain the key coupling paths;

[0051] Analyze the characteristics of the multi-level factor coupling network topology structure from three scales: micro, meso, and macro, and generate a non-linear factor coupling relationship network diagram.

[0052] Specifically, the outlier detection uses an improved Z-Score method, that is, the mean and standard deviation within the moving window are calculated for each type of sensor data respectively. When a data point deviates from the mean by more than 3 times the standard deviation, it is marked as an outlier. In the specific operation, for each sensor data sequence, the data is first arranged in chronological order, and then a sliding window with a size of 50 data points is set, and the statistical features are calculated within the window. For the detected outliers, a local interpolation replacement strategy is adopted, that is, the weighted average of the normal data points before and after the outlier is used to replace the outlier. The data smoothing process uses the Savitzky-Golay filter, which is a smoothing technique based on local polynomial fitting and can effectively retain the peak characteristics of the data while filtering out high-frequency noise. For sensors with slow responses such as environmental rainfall data, the Savitzky-Golay filter with a window length of 7 is used; for fast-changing dynamic parameters such as yaw rate, a filter with a window length of 3 is used to retain the fast-changing characteristics. Through these processes, the burrs, jumps, and white noise in the original data are effectively removed, forming a continuous and smooth preprocessed data set.

[0053] For the factor pairs in the preprocessed data set, the joint probability distribution is calculated by the kernel density estimation method to obtain the mutual information value matrix between factors. Mutual information is an index in information theory used to measure the degree of mutual dependence between two random variables, indicating the degree of reduction in the uncertainty of another variable given one variable. Calculating mutual information requires first estimating the marginal probability distribution and joint probability distribution of the variables. In this method, the Gaussian kernel density estimation method is used to achieve non-parametric probability density estimation. For any two safety influencing factors X and Y, first normalize their data to the [0,1] interval, and then use the Gaussian kernel function to smooth the data points to estimate the joint probability density function. In the specific calculation process, for each pair of factors in the data set, a two-dimensional plane grid is constructed, the data points are mapped onto the grid, and then the density value of each grid point is calculated through the kernel function. Based on the obtained probability density function, the mutual information value is calculated:

[0054]

[0055] Among them, MI(X,Y) represents the mutual information value between factors X and Y, p(x,y) is the joint probability density function of X and Y, and p(x) and p(y) are the marginal probability density functions of X and Y respectively. Repeat the above calculation for all factor pairs to obtain an N×N mutual information value matrix, where N is the total number of safety influencing factors. Segment the preprocessed data set based on a sliding time window, and calculate the transfer entropy between factors through symbolic transformation to form a directional coupling strength matrix. Transfer entropy is an asymmetric measure that can identify the directionality of information flow and helps to reveal the causal relationship between factors. The first step in calculating transfer entropy is to perform symbolic transformation on the data, that is, convert the continuous time series data into a discrete symbol sequence. This method uses the uniform binning technique to evenly divide the numerical range of each factor into 8 intervals, and each interval corresponds to a symbol. For factor X, its time series {x1,x2,...,xT} is converted into a symbol sequence {s1,s2,...,sT}.

[0056] The calculation formula of transfer entropy is as follows:

[0057]

[0058] Among them, TE Y→X represents the transfer entropy from factor Y to factor X, which measures the influence degree of factor Y on the future state of factor X; x t+h represents the state of factor X at time t+h, where h is the prediction step size, usually set to 1; represents the historical sequence of k states of factor X before time t; represents the historical sequence of l states of factor Y before time t; is the joint probability distribution; and are the conditional probability distributions respectively. In actual calculation, the historical lengths k and l usually take values of 2-3 to balance the computational complexity and information capture ability. Calculate the transfer entropy for all factor pairs to obtain a directional coupling strength matrix, and the element TE ab in the matrix represents the information flow strength from factor b to factor a.

[0059] Fuse the mutual information value matrix and the directional coupling strength matrix to construct a multi-level factor coupling network topology structure. The fusion process uses a weighted combination method to define the comprehensive coupling strength:

[0060] CE ab =α·MI ab +(1-α)·TE ab

[0061] Among them, CE ab$CE_{ab}$ represents the comprehensive coupling strength between factor a and factor b. $\alpha$ is the weight coefficient, with a value range of [0, 1], which is adjusted according to the specific application scenario and generally takes 0.5. Based on the comprehensive coupling strength matrix, a multi-level network is constructed, where the network nodes represent security impact factors, and the weights of the connections between nodes are determined by $CE_{ab}$. To reflect the hierarchy of factors, the factors are classified into bottom-layer sensor data nodes, middle-layer state feature nodes, and top-layer security index nodes according to their functional characteristics, forming a three-layer nested network structure.

[0062] The redundant connections of the multi-level factor coupling network topology are pruned through the sparse representation learning method, and the key coupling paths are retained. A large number of weak connections in the network will increase the computational complexity and introduce noise interference. Sparse representation learning realizes the sparsity of the network by minimizing the reconstruction error while forcing most connection weights to approach zero. In the specific implementation, first, a coupling strength threshold $\theta$ is set, and the connections with $CE$ ab $<\theta$ are deleted. The threshold $\theta$ is determined by an adaptive method so that the number of retained connections accounts for about 20% of the total possible connections. Then, the retained connections are sorted by importance, and the L1 regularization method is used to further reduce the weights of unimportant connections. In this way, the complex fully connected network is simplified into a sparse network containing key coupling paths, which not only reduces the computational complexity but also improves the interpretability of the network.

[0063] Finally, the characteristics of the multi-level factor coupling network topology are analyzed from three scales: micro, meso, and macro, and a non-linear factor coupling relationship network diagram is generated. The micro-scale analysis focuses on the local structure of a single node and its directly connected nodes, and calculates the degree, clustering coefficient, and local influence index of the node. The meso-scale analysis focuses on the community structure of the network, uses the Louvain algorithm to identify tightly coupled factor groups, and calculates the coupling strength within and between groups. The macro-scale analysis examines the global characteristics of the entire network, including indicators such as the average path length, network diameter, and global efficiency. The analysis results of these three scales are integrated into the same non-linear factor coupling relationship network diagram, and the coupling characteristics of different scales are intuitively represented by the node size, color, and line thickness.

[0064] For example, in the multi-factor coupling analysis under an emergency braking scenario, first, multi-dimensional raw data including vehicle speed, brake pedal pressure, tire slip ratio, road surface friction coefficient, etc. is collected. Outlier detection is performed on this data, and several obvious outliers are found in the tire slip ratio data, which may be caused by sensor jitter. These outliers are replaced with reasonable values through local interpolation. After data smoothing, the mutual information values between factors are calculated using the kernel density estimation method. The results show that the mutual information value between the brake pedal pressure and the tire slip ratio is as high as 0.82, indicating a high correlation between the two; while the mutual information value between the vehicle speed and the ambient light intensity is only 0.08, showing a very low correlation. Further, the causal relationship between factors is determined through transfer entropy calculation. For example, the transfer entropy from the brake pedal pressure to the tire slip ratio is 0.75, while the transfer entropy in the reverse direction is only 0.12, clearly revealing the one-way influence relationship of the braking operation on tire slip. After fusing the mutual information matrix and the transfer entropy matrix, a multi-level network containing all safety factors is constructed. The factors related to the braking system in the network form an obvious tightly connected sub-network. By setting a coupling strength threshold of 0.3, a large number of weak connections are deleted, such as the weak association between the vehicle speed and the wiper speed. The final non-linear factor coupling relationship network diagram clearly shows the key factor chain under the emergency braking scenario: driver attention → brake pedal pressure → wheel pressure distribution → tire slip ratio → vehicle body attitude. The coupling strength of the factors on this chain is much higher than that of other paths, indicating that this is the core influence chain that should be focused on in safety analysis.

[0065] In a specific embodiment, the process of executing step S103 may specifically include the following steps:

[0066] Extract basic safety indicators from the multi-dimensional raw data of safety influencing factors and construct the underlying structure of the indicator system;

[0067] Identify key coupling factor combinations based on the coupling path strength in the non-linear factor coupling relationship network diagram and generate intermediate-level coupling indicators;

[0068] Calculate the information entropy for the intermediate-level coupling indicators to obtain the indicator weight distribution coefficients;

[0069] Combine the intermediate-level coupling indicators with the indicator weight distribution coefficients and form the top-level comprehensive safety performance indicator through fuzzy integral operation;

[0070] Perform scenario calibration on the top-level comprehensive safety performance indicator according to the characteristics of different driving scenarios to obtain the scenario-adaptive safety indicator value;

[0071] Standardize the scenario-adaptive safety indicator value to form the comprehensive score of the motor vehicle safety performance.

[0072] Specifically, basic safety indicators are extracted from the original multi-dimensional safety impact factor data to construct the underlying structure of the indicator system. Basic safety indicators are numerical indicators that can reflect the basic safety characteristics of motor vehicles and are directly calculated from the original sensor data. During the extraction process, specialized processing methods are used for different types of original data: for vehicle dynamic parameter data, the braking distance indicator (theoretical braking distance calculated based on the initial speed and deceleration), the lateral stability indicator (calculated based on the change rate of yaw angular velocity and sideslip angle), and the longitudinal stability indicator (calculated based on the acceleration change and speed control deviation) are calculated; for environmental state data, the road surface adhesion index (calculated based on the friction coefficient and road surface type) and the environmental visibility index (calculated comprehensively based on light intensity and rainfall) are calculated; for driving behavior data, the operation smoothness index (calculated based on the change rate of steering wheel angle and pedal pressure) and the driver attention index (calculated based on eye movement tracking data) are calculated. These basic indicators are independent of each other and directly reflect the safety characteristics of motor vehicles in a single dimension, jointly constituting the underlying structure of the indicator system. According to the coupling path strength in the non-linear factor coupling relationship network diagram, key coupling factor combinations are identified to generate intermediate-level coupling indicators. Intermediate-level coupling indicators are different from the underlying basic indicators, and they reflect the composite safety characteristics under the interaction of multiple safety factors. The identification process first conducts a path strength analysis on the non-linear factor coupling relationship network diagram to extract the factor connection paths with a coupling strength in the top 30%. Then, based on these key paths, factor subsets forming a closed-loop or star structure are searched for, and these subsets usually represent tightly coupled functional units. For example, the "braking-road-tire" subset includes three factors: braking pedal pressure, road surface friction coefficient, and tire grip, which jointly affect the braking performance of motor vehicles. For each identified factor subset, a corresponding coupling indicator calculation method is designed. Common intermediate-level coupling indicators include: "braking-steering coupling stability index" (reflecting the stability of motor vehicles when braking and steering simultaneously), "road surface-tire-suspension coupling adaptability index" (reflecting the adaptability of the vehicle suspension system to different road surface conditions), "driver-vehicle response coordination index" (measuring the matching degree between the driver's operation intention and the actual response of the vehicle), etc. These intermediate-level coupling indicators are calculated by combining multiple basic indicators and can more comprehensively reflect the safety performance of motor vehicles under complex conditions. Information entropy calculation is performed on the intermediate-level coupling indicators to obtain the indicator weight distribution coefficient. Information entropy is an indicator that measures the uncertainty of data. The higher the entropy value, the greater the uncertainty contained. In determining the indicator weights, adopting the information entropy principle means giving lower weights to indicators with high information content (i.e., large dispersion) and higher weights to indicators with low information content (i.e., concentrated distribution). The calculation process first constructs a historical data frequency distribution table for each intermediate-level coupling indicator, divides the indicator value range into several equal-interval intervals, and counts the sample frequencies in each interval. Then, based on the frequency distribution, the information entropy of each indicator is calculated:

[0073]

[0074] where H(C r ) represents the information entropy of the r-th middle-level coupling index C r , G is the number of divided intervals, and f rg is the sample frequency of the r-th index in the g-th interval. Next, calculate the coefficient of variation of each index:

[0075]

[0076] where D(C r ) is the coefficient of variation of the r-th middle-level coupling index, and log(G) is the theoretical maximum entropy value. Finally, normalize the coefficient of variation to obtain the weight distribution coefficient:

[0077]

[0078] where W(C r ) is the weight distribution coefficient of the r-th middle-level coupling index, and R is the total number of middle-level coupling indices. Through this method, the indices with higher dispersion obtain lower weights, and the indices with lower dispersion obtain higher weights, reflecting the maximum information principle in information theory.

[0079] Combine the middle-level coupling indices with the index weight distribution coefficients to form the top-level comprehensive safety performance index through fuzzy integral operation. Fuzzy integral is a non-linear integration method that can handle the interaction between indices and is particularly suitable for dealing with the situation where there are overlaps or synergistic effects between indices. This method uses the improved Choquet fuzzy integral, and its calculation process is as follows: First, standardize all middle-level coupling indices so that their value ranges are unified in the interval [0,1]:

[0080]

[0081] where is the value of the r-th middle-level coupling index after standardization, and Cr,min and C r,max are the historical minimum and maximum values of this index respectively. Then, sort the standardized indices from large to small by value:

[0082]

[0083] where represents the index at the s-th position after sorting. Next, calculate the fuzzy measure, and the fuzzy measure reflects the importance of the index subset:

[0084]

[0085] where γ(A s) is the subset A that contains the top s indicators before sorting s is the fuzzy measure, β is the interaction parameter between indicators, determined according to the indicator correlation analysis, and usually ranges from [-1, 1]. Finally, based on the sorting and fuzzy measure, calculate the Choquet fuzzy integral value as the top-level comprehensive safety performance indicator:

[0086]

[0087] where TSI is the top-level comprehensive safety performance indicator and γ(A0) = 0. The Choquet fuzzy integral can effectively capture the synergistic effect and conflict effect between indicators, and can better reflect the overall performance of complex systems than simple weighted average.

[0088] Calibrate the top-level comprehensive safety performance indicator according to the characteristics of different driving scenarios to obtain the scenario-adaptive safety indicator value. The purpose of scenario calibration is to make the safety performance evaluation more in line with the actual driving environment, because the importance and threshold standards of safety indicators are different in different scenarios. The calibration process first uses the clustering method to divide the historical driving data into several typical scenarios, such as highway cruising, urban congestion, mountain road curves, bad weather, etc. For each scenario, a special calibration model is constructed, and the calibration formula is:

[0089]

[0090] where CSI q is the scenario-adaptive safety indicator value in scenario q, η q is the scenario importance coefficient, reflecting the danger level of scenario q. The larger the value, the higher the safety requirements in this scenario; is the scenario environment response function, which depends on the current environment feature vector E q , and is used to adjust the sensitivity of safety indicators under different environmental conditions. For example, in the scenario of a rainy and slippery road surface, the weight of the indicator related to tire grip will be automatically increased, while in the scenario of highway cruising, the weight of the indicator related to lane keeping will be enhanced. Through scenario calibration, the safety assessment results are more environmentally adaptable and practical.

[0091] Finally, standardize the scenario-adaptive safety indicator value to form the comprehensive score of motor vehicle safety performance. The purpose of standardization is to convert the safety indicator values in various scenarios into a unified dimension and a comprehensive score that is easy to understand and compare. The standardization process first sets the scoring range from 0 to 100 points, and then establishes a non-linear mapping relationship:

[0092]

[0093] where MCSS is the comprehensive score of motor vehicle safety performance, α is the slope parameter, controlling the steepness of score change, βq It is the midpoint parameter of scenario q, representing the safety index threshold for obtaining 50 points in this scenario. This Sigmoid-type mapping can provide a high resolution in the middle region while having a compression effect on extreme values, making the score distribution more reasonable. The finally obtained comprehensive score intuitively reflects the safety performance level of the motor vehicle in the current driving scenario, and the higher the score, the better the safety performance.

[0094] In a specific embodiment, the process of executing step S104 may specifically include the following steps:

[0095] Construct a time series data matrix for the comprehensive score of motor vehicle safety performance, and determine the optimal time delay parameter through the mutual information method;

[0096] Use the false nearest neighbor method to analyze the time series data matrix and determine the optimal embedding dimension;

[0097] According to the optimal time delay parameter and the optimal embedding dimension, reconstruct the comprehensive score of motor vehicle safety performance into a high-dimensional phase space trajectory;

[0098] Calculate the nonlinear dynamic characteristic parameters of the high-dimensional phase space trajectory, including the maximum Lyapunov exponent, correlation dimension, and approximate entropy value;

[0099] Extract complex network features from the high-dimensional phase space trajectory to obtain network topology feature quantization indicators;

[0100] Combine the nonlinear dynamic characteristic parameters and the network topology feature quantization indicators, and through nonlinear mapping transformation, obtain a low-dimensional discriminant feature set.

[0101] Specifically, a time series data matrix is constructed for the comprehensive safety performance score of motor vehicles, and the optimal time delay parameter is determined by the mutual information method. During the construction of the time series data matrix, the continuously collected comprehensive safety performance scores of motor vehicles are arranged in chronological order to form a one-dimensional vector, where each element represents the comprehensive safety performance score at a certain time point. The mutual information method is based on the principle of information theory and is used to determine the best delay value during the reconstruction of the time series. It is achieved by calculating the statistical dependence degree between the time series and its delayed version. In specific operations, for each candidate delay value, the mutual information value between the original sequence and the delayed sequence is calculated. First, the original data and the delayed data are divided into several intervals, the joint probability distribution and the marginal probability distribution are statistically analyzed, and then the mutual information value is calculated. When the delay increases, the mutual information value usually shows a trend of first decreasing and then stabilizing. The delay corresponding to the first time the mutual information value drops to the local minimum is selected as the optimal time delay parameter. This parameter represents the minimum time interval at which adjacent points in the time series begin to exhibit relative independence. Analyzing the time series data matrix using the false nearest neighbor method to determine the optimal embedding dimension is another key step in phase space reconstruction. The core idea of the false nearest neighbor method is that when the embedding dimension is insufficient, points that are far apart in the high-dimensional space may become nearest neighbors in the low-dimensional projection, which is called a "false nearest neighbor". As the embedding dimension increases, the number of false nearest neighbors will decrease. In specific implementation, a series of candidate embedding dimensions are first set. For each dimension, the corresponding delayed vectors are constructed and the nearest neighbors of each point are identified. Then, it is checked whether these nearest neighbor points still maintain the nearest neighbor relationship after the dimension increases. If the ratio of the distance between two points in the higher-dimensional space to the distance in the current dimension exceeds a preset threshold, it is considered a false nearest neighbor. Calculate the proportion of false nearest neighbor points in the total number of points for different dimensions. When this proportion first drops below a specific threshold, the corresponding dimension is the optimal embedding dimension. This dimension is sufficient to unfold the phase space trajectory and avoid trajectory self-intersection.

[0102] Based on the optimal time delay parameter and the optimal embedding dimension, the comprehensive safety performance score of motor vehicles is reconstructed into a high-dimensional phase space trajectory. The reconstruction is based on the Takens embedding theorem, which states that the phase space topologically equivalent to the original dynamic system can be reconstructed through the delayed coordinates of a univariate time series. During the reconstruction process, for each time point, a vector composed of the current point and its subsequent delayed points is constructed. For example, if the optimal time delay is 5 and the optimal embedding dimension is 4, the state vector at time point t is [MCSS(t), MCSS(t + 5), MCSS(t + 10), MCSS(t + 15)]. State vectors are constructed for all possible t, and the distribution of a series of points in the high-dimensional space is obtained, forming a high-dimensional phase space trajectory. This trajectory retains the dynamic characteristics of the original time series and can reflect the evolution law of the safety performance of motor vehicles over time.

[0103] Calculating the non - linear dynamic characteristic parameters of high - dimensional phase - space trajectories, including the largest Lyapunov exponent, correlation dimension, and approximate entropy value, is an important step in extracting key features from the trajectories. The largest Lyapunov exponent measures the sensitivity of the system to initial conditions, that is, the degree of chaos. When calculating, two close points on the trajectory are selected, and their separation over time is tracked, and the average value of the separation rate is taken. A positive Lyapunov exponent indicates that the system has chaotic characteristics, and the larger the value, the higher the uncertainty. The correlation dimension characterizes the geometric complexity of the trajectory. During the calculation process, the correlation between trajectory point pairs under different distance thresholds is statistically analyzed, and the dimension value is determined through the power - law behavior of the correlation integral. The higher the dimension, the greater the degree of freedom of the system and the more complex the dynamic behavior. The approximate entropy measures the regularity and predictability of the time series and is calculated by comparing the matching of patterns of different lengths. The higher the entropy value, the more irregular the sequence and the lower the predictability. These three parameters jointly describe the complex dynamic characteristics of the motor vehicle safety performance over time. Extracting complex network features from high - dimensional phase - space trajectories and obtaining quantitative indicators of network topological features is an important method for further characterizing the trajectory characteristics. The core idea of complex network feature extraction is to transform the phase - space trajectory into a network structure and then analyze the topological characteristics of the network. In specific operations, first, the phase - space is evenly divided into multiple small regions, and each region corresponds to a node in the network. Then, network connections are established according to the transfer situation of trajectory points between regions. If the phase - space trajectory moves from region A to region B, a directed edge from node A to node B is established in the network, and the weight of the edge can be the frequency of the transfer occurrence. After constructing the network, calculate the topological feature indicators of the network, including node degree - distribution characteristics (such as average degree, degree - distribution entropy), clustering coefficient (reflecting the local compactness of the network), centrality indicators (the importance of nodes in the network), small - world characteristics (comparison of the average path length of the network with that of a random network), etc. These network features can characterize the patterns and laws of safety - performance changes from a topological perspective.

[0104] Combining the nonlinear dynamics characteristic parameters and the network topology characteristic quantification indicators, and through nonlinear mapping transformation, obtaining a low-dimensional discriminant feature set is the step of compressing high-dimensional features into a low-dimensional representation convenient for classification. The methods adopted for nonlinear mapping transformation mainly include t-SNE and the improved Isomap algorithm. The t-SNE algorithm is particularly good at preserving the local structure of data. Its core is to construct the conditional probability distributions between point pairs in high-dimensional and low-dimensional spaces respectively, and then minimize the KL divergence between the two. The Isomap algorithm reduces the dimension by preserving the geodesic distance of data on the manifold and can better preserve the global geometric structure of data. In practical applications, the advantages of the two algorithms are often combined. First, the Isomap is used to obtain the initial mapping, and then the t-SNE is used for fine adjustment. In addition, to enhance the discriminant ability of features, the Fisher discriminant criterion and the maximum mutual information criterion are introduced for feature selection to eliminate redundant and noisy features. The finally obtained low-dimensional discriminant feature set usually has a dimension between 3 and 5, and each dimension is a nonlinear combination of the original high-dimensional features, which can effectively distinguish different security state categories.

[0105] In a specific embodiment, the process of executing step S105 may specifically include the following steps:

[0106] Calculate the local density values and distance factors of each feature point in the low-dimensional discriminant feature set, and construct a density-distance decision graph;

[0107] Identify the initial clustering center points from the density-distance decision graph through the density peak detection method;

[0108] Perform clustering division according to the distribution characteristics of the initial clustering center points to obtain the initial security level clusters;

[0109] Evaluate the effectiveness of the initial security level clusters, and determine the optimal number of clusters by calculating the silhouette coefficient;

[0110] Construct a closed decision surface around the boundary of the security level cluster corresponding to the optimal number of clusters to form a static security level boundary;

[0111] Analyze the temporal variation characteristics of the security level clusters based on historical data, establish a boundary dynamic update rule, and generate dynamic security level boundary values.

[0112] Specifically, in the classification method of motor vehicle safety performance based on multi-factor coupling effects, adaptive density clustering and dynamic boundary determination are the key steps. In this process, the local density values and distance factors of each feature point in the low-dimensional discriminant feature set are first calculated to construct a density-distance decision graph. The local density value represents the degree of aggregation of data points around the feature point, while the distance factor represents the minimum distance between this point and other high-density points. When calculating the local density value, for each feature point, first count the number of points whose distance from this point is less than the cut-off distance, and then divide this number by the total number of points to obtain the normalized local density. The cut-off distance generally takes the 2%-3% quantile of the distances between all point pairs. The distance factor is, for each point, the distance to the point with the minimum distance among all other points whose density is greater than this point. If a point has the maximum local density, its distance factor is defined as the maximum distance between all points. In this way, each point has a pair of local density values and distance factors, and these two values of all points are plotted on a two-dimensional plane to form a density-distance decision graph. In this graph, the points with both high local density and large distance factors are usually located in the upper right region of the graph, and these points are ideal candidate points for cluster centers.

[0113] Identifying the initial cluster center points from the density-distance decision graph through the density peak detection method is the first step of clustering. The core idea of density peak detection is that the cluster center should have a relatively high local density and maintain a large distance from other high-density points. In the specific implementation, first calculate the decision value of each point, which is the product of the local density and the distance factor. Then, sort all points in descending order of the decision value, and select several points with the largest decision values as the initial cluster centers. During the selection process, the minimum distance constraint between cluster centers also needs to be considered to prevent the center points from being too concentrated. In the analysis of motor vehicle safety performance, the initial cluster centers usually correspond to typical safety state modes, such as "high stability driving state", "critical stability state", "slight instability state", and "severe instability state", etc. These center points are distributed in different regions of the feature space and can well represent different safety level categories.

[0114] Clustering and partitioning based on the distribution characteristics of the initial clustering center points to obtain the initial safety level clusters is a key step in forming safety classification. The partitioning process adopts a density-based assignment strategy, that is, each non-center point is assigned to the cluster where the clustering center with the strongest density association is located. In specific operations, first, all non-center points are sorted from high to low according to local density. Then, for each non-center point, find the points with assigned cluster labels among the points in its neighborhood (usually defined within the cut-off distance) whose local density is greater than that of this point. Assign this non-center point to the cluster corresponding to the cluster label with the highest frequency of occurrence among them. If there are no points with assigned cluster labels in the neighborhood, this point is regarded as an outlier or noise point, and it can be temporarily not assigned or assigned to a special "noise cluster". This top-down assignment method can ensure that data points flow along the density gradient towards the clustering center, forming a natural clustering boundary. In the classification of motor vehicle safety performance, this process assigns low-dimensional discriminant feature points to different safety level clusters, and each cluster represents a type of safety state.

[0115] Evaluating the effectiveness of the initial safety level clusters and determining the optimal number of clusters through silhouette coefficient calculation is an important link to ensure the clustering quality. The silhouette coefficient is an internal metric for evaluating clustering quality, which comprehensively considers the intra-cluster compactness and inter-cluster separation. For each data point, calculate its silhouette value, which reflects the degree to which the point is correctly clustered. The calculation includes three steps: First, for point i, calculate the average distance a(i) between it and other points in the same cluster, which represents the dissimilarity of point i in its belonging cluster; then, calculate the average distance between point i and all points in each other cluster, and take the minimum value b(i), which represents the similarity of point i to its nearest non-belonging cluster; finally, calculate the silhouette value s(i)=(b(i)-a(i)) / max{a(i),b(i)}. The silhouette value is between -1 and 1, and the larger the value, the better the clustering effect. Calculate the average value of the silhouette values of all points, which is the silhouette coefficient of the clustering. By trying different numbers of clusters (starting from 2 and increasing), calculate the corresponding silhouette coefficients, and select the number of clusters with the largest silhouette coefficient as the optimal number of clusters. In the classification of motor vehicle safety performance, the optimal number of clusters is usually between 3 and 5, corresponding to different levels of safety states.

[0116] Constructing a closed decision surface around the boundary of the safety level cluster corresponding to the optimal number of clusters to form a static safety level boundary is the basis for achieving safety state classification. The closed decision surface refers to the boundary surface that divides different safety level clusters in the feature space, which defines the dividing line of different safety states. In the construction process, an improved Support Vector Data Description (SVDD) algorithm is adopted. This algorithm can generate a compact closed boundary for each cluster. For each safety level cluster, first, all the points in this cluster are regarded as positive samples, and the points in other clusters are regarded as negative samples, and then the SVDD model is trained. The core of SVDD is to find a hypersphere with the minimum radius, so that positive samples are as much as possible inside the sphere and negative samples are as much as possible outside the sphere. By introducing a kernel function, SVDD can handle non-spherical data distributions. The optimization goal is to minimize the volume of the hypersphere while maximizing the inclusion rate of positive samples. After solving the SVDD problem to obtain the support vectors and boundary parameters, these parameters are used to define the decision function, which can judge whether a new point is inside the cluster. Combining the decision functions of each cluster forms a complete classification boundary, that is, the static safety level boundary.

[0117] Analyzing the temporal variation characteristics of the safety level cluster based on historical data and establishing rules for dynamic boundary update to generate dynamic safety level boundary values is a key measure to adapt to the dynamic changes of safety states. The static safety level boundary cannot cope with the dynamic change characteristics of the vehicle safety state, so a mechanism for dynamic boundary update needs to be established. First, a large amount of historical data is collected, and the change trajectory of the safety state is recorded in time series to analyze the evolution law of the safety level cluster under different driving scenarios. Through analysis, key factors affecting the change of the safety level are identified, such as environmental conditions, vehicle speed changes, driving operations, etc. Then, boundary adjustment rules based on time series patterns are established, including expansion rules (when a trend of increasing safety risk is detected, the boundary shrinks to give an early warning) and contraction rules (when the safety state is stable, the boundary is appropriately relaxed). At the same time, an adaptive learning mechanism is introduced to enable the boundary to be continuously optimized according to new data. In the dynamic update process, the sliding time window technology is adopted to recalculate key parameters, such as local density threshold, distance factor threshold, etc. in the recent time window, and the boundary parameters are adjusted accordingly. This dynamic boundary technology enables the system to adapt to the safety state changes under different driving conditions and provide more accurate safety classification.

[0118] For example, in this process, assume that safety performance data of a certain vehicle model under various driving conditions are collected. After being processed through the foregoing steps, a three-dimensional low-dimensional discriminant feature set is obtained, which contains a total of 2,000 feature points. First, calculate the local density value of each feature point. The truncation distance is selected as the 2.5% quantile of the point-to-point distance, which is approximately 0.15. For feature point A, the number of points with a distance less than 0.15 from it is counted as 85, so its normalized local density is 85 / 2,000 = 0.0425. Then calculate the distance factor. It is found that among the points with a local density greater than 0.0425 of point A, the nearest one is point B, and the distance between the two points is 0.58. Therefore, the distance factor of point A is 0.58. Similarly, calculate the local density values and distance factors of all points, and construct a density-distance decision graph. Four obvious peak points are identified from this graph, that is, the points with relatively large local density values and distance factors. These points are located in different regions of the feature space, corresponding to different safety state modes: the local density of peak point 1 is 0.085 and the distance factor is 0.92, representing the "highly stable driving state"; the local density of peak point 2 is 0.072 and the distance factor is 0.85, representing the "normal driving state"; the local density of peak point 3 is 0.063 and the distance factor is 0.78, representing the "slightly unstable state"; the local density of peak point 4 is 0.055 and the distance factor is 0.81, representing the "obviously unstable state". Next, take these 4 peak points as the initial clustering centers, and perform cluster partitioning on all feature points according to the density gradient allocation strategy to obtain 4 initial safety level clusters. Then calculate the silhouette coefficients under different numbers of clusters (2 to 6 clusters), and it is found that the silhouette coefficient is the highest when there are 4 clusters, which is 0.68, confirming that the optimal number of clusters is 4. Construct SVDD decision boundaries around these 4 clusters respectively to form static safety level boundaries. Finally, analyze the temporal variation characteristics of the safety state in historical data and find that under wet road surface conditions, the transition from the "normal state" to the "slightly unstable state" often occurs faster than on dry road surfaces. Accordingly, set the corresponding expansion coefficient in the boundary dynamic update rule so that the system can identify potential unstable states earlier under wet road surface conditions and give early warnings. Through this dynamic boundary technology, the safety performance classification system can adapt to the changes in safety states under different driving environments and provide more accurate grading warning information.

[0119] In a specific embodiment, the process of executing step S106 may specifically include the following steps:

[0120] Receive the motor vehicle dynamic parameters, environmental conditions, and driving behavior data collected by the distributed sensor network through the real-time data acquisition interface, and construct a real-time state data matrix;

[0121] Apply a fast feature extraction algorithm to the real-time state data matrix to obtain a real-time feature vector;

[0122] Project the real-time feature vector into the feature space where the low-dimensional discriminant feature set is located through the phase space mapping transformation method to generate real-time low-dimensional discriminant points;

[0123] Calculate the positional relationship between the real-time low-dimensional discriminant points and the dynamic safety level boundary values to determine the category to which the current motor vehicle safety state belongs;

[0124] Construct a safety risk level based on the category to which the current motor vehicle safety state belongs and the dynamic change trend of the real-time low-dimensional discriminant points;

[0125] Generate hierarchical safety risk warning information according to the comparison table of safety risk levels and preset risk thresholds.

[0126] Specifically, after the motor vehicle safety performance classification method based on multi-factor coupling influence completes the determination of the safety level boundary, it is necessary to classify the real-time collected motor vehicle states and output warning information. First, receive the motor vehicle dynamic parameters, environmental conditions, and driving behavior data collected by the distributed sensor network through the real-time data acquisition interface to construct a real-time state data matrix. The real-time data acquisition interface is a bridge connecting the sensor network and the central processing unit, and data transmission is realized through the high-speed CAN bus protocol. The acquired data is organized in a unified format, including four basic fields: timestamp, sensor identifier, value, and quality flag. After the data is received, it is rearranged according to the data type and time sequence to form a real-time state data matrix. Each row of this matrix corresponds to a time point, and each column corresponds to a sensor parameter. To handle the sampling frequency differences of different sensors, an interpolation alignment method is used to align all data to a unified time point. For example, when the sampling frequency of the yaw rate sensor is 100Hz and the sampling frequency of the environmental temperature sensor is only 1Hz, linear interpolation is performed on the environmental temperature data to estimate the values at intermediate time points to ensure that each time point in the matrix has a complete set of parameters. In addition, the real-time state data matrix also includes a sliding time window mechanism that only retains the data for the most recent period (usually 10 - 30 seconds), which not only ensures the real-time nature of the analysis but also provides sufficient historical information for situation analysis.

[0127] Applying a fast feature extraction algorithm to the real-time state data matrix to obtain the real-time feature vector is a key step in data dimensionality reduction and feature representation. The fast feature extraction algorithm is a feature calculation method optimized for real-time application scenarios, which needs to meet the computational efficiency requirements while ensuring the feature representation ability. The algorithm first preprocesses the real-time state data matrix, including normalization, detrending, and smoothing. Then, it calculates time-domain features (such as mean, standard deviation, peak value, valley value, peak-valley difference, etc.), frequency-domain features (spectral characteristics obtained through fast Fourier transform), and statistical features (such as skewness, kurtosis, entropy, etc.). For vehicle safety performance evaluation, features related to dynamic stability are mainly extracted, such as the fluctuation range of yaw rate, the change rate of braking distance, and the spectral distribution of lateral acceleration. These original features are preliminarily reduced in dimension through principal component analysis (PCA), and the principal components that retain more than 90% of the explained variance are retained. During the PCA process, a pre-computed eigenvector matrix is used to perform a linear transformation on the real-time features, avoiding re-computing the eigenvalue decomposition each time and improving the computational efficiency. The dimension of the finally obtained real-time feature vector is usually between 10 and 20, which contains the main information of the original data.

[0128] Projecting the real-time feature vector into the feature space where the low-dimensional discriminant feature set is located through a phase space mapping transformation method to generate the real-time low-dimensional discriminant point is the bridge connecting the training model and real-time monitoring. The phase space mapping transformation method refers to transforming the extracted feature vector into the same feature space as the training model to ensure that the real-time data is compared with the historical model in the same reference system. In the specific implementation, first, the mapping parameters saved in the training stage, including the parameter matrix and reference point set of the t-SNE or Isomap algorithm, are used to project the real-time feature vector into the low-dimensional feature space. For the t-SNE mapping, a parameter inheritance strategy is adopted, using the embedding coordinates of the pre-trained model as the initial value, and only performing a limited number of optimization iterations on the new data points to avoid completely re-computing the embedding coordinates and improving the real-time response ability. For the Isomap mapping, the connection relationship between the real-time feature vector and the k nearest neighbor points in the training dataset is constructed, and the real-time feature is projected into the low-dimensional space based on the calculation of geodesic distance. In this way, a low-dimensional discriminant point corresponding to the real-time state data is generated, and the position of this point in the low-dimensional feature space intuitively reflects the current safety state characteristics of the vehicle.

[0129] Calculating the positional relationship between the real-time low-dimensional discriminant points and the dynamic safety level boundary values to determine the category to which the current motor vehicle safety state belongs is the core step of classification decision-making. The positional relationship calculation is based on the distance metric from the point to the boundary, and the improved nearest point projection distance method is adopted. First, calculate the distances between the real-time low-dimensional discriminant points and the boundaries of each safety level cluster, and find the nearest boundary and its direction. The distance calculation is based on the function value of the point to the SVDD decision boundary, which is positive inside the boundary and negative outside the boundary, and the numerical size represents the relative distance to the boundary. Then, determine the safety level category to which the real-time discriminant point belongs based on the distance value. If the discriminant point is inside the boundary of a certain safety level cluster, it belongs to that category; if it is inside the boundaries of multiple clusters at the same time, select the cluster with the largest distance as the belonging category; if it is outside the boundaries of all clusters, infer its most likely belonging category based on the distance and direction to the nearest boundary, combined with historical trajectory information. For discriminant points near the boundary, the concept of fuzzy belonging is introduced, and the belonging degrees to various categories are calculated simultaneously for subsequent risk level assessment.

[0130] Constructing the safety risk level based on the category to which the current motor vehicle safety state belongs and the dynamic change trend of the real-time low-dimensional discriminant points is an important part of risk assessment. The safety risk level not only considers the category belonging of the current state but also needs to consider the change trend of the state, especially the possibility of evolving into an unsafe state. The dynamic change trend analysis is based on the movement trajectory of the real-time low-dimensional discriminant points in the feature space, and the velocity vector (the position difference between adjacent time points) and the acceleration vector (the change rate of the velocity vector) of the discriminant points are calculated. Then, evaluate the relative directions of these vectors and the safety level boundary. If the velocity vector points to a more unsafe area and has a large amplitude, it indicates that the safety state is deteriorating rapidly, and the risk level should be increased accordingly. In addition, the distance between the real-time discriminant point and the safety level boundary is also combined, and the smaller the distance, the higher the risk. By comprehensively considering factors such as the category belonging of the discriminant point, the boundary distance, the movement speed, and the movement direction, calculate the comprehensive risk score. This score is divided into different safety risk levels according to the set thresholds, such as five levels: "Safe", "Attention", "Warning", "Danger", and "Emergency".

[0131] Generating hierarchical safety risk warning information according to the comparison table of safety risk levels and preset risk thresholds is the last link of warning output. The preset risk threshold comparison table is a mapping relationship between risk levels and warning content preset based on historical data and expert knowledge. Different risk levels correspond to different warning methods, warning content, and warning urgency. When generating warning information, first query the preset comparison table according to the current safety risk level to obtain the corresponding warning template. Then, combined with the current specific safety state characteristics, such as the type of unstable factors (such as sideslip, tail swing, insufficient braking, etc.), environmental conditions (such as slippery road surface, sharp turn, etc.), and driving operations (such as sharp steering, sudden braking, etc.), parameterize and fill the warning template to form specific warning content. The warning content includes the risk level, risk description, possible consequences, and recommended countermeasures. Finally, select an appropriate warning method according to the warning urgency, such as dashboard prompt, sound warning, seat vibration, or combined warning, to ensure that the driver can notice the risk in time and take corresponding measures.

[0132] For example, in a scenario of high-speed driving, through the real-time data acquisition interface, the yaw rate is received as 0.32 rad / s, the lateral acceleration is 4.2 m / s 2 , the steering wheel angle is 45 degrees, and the road surface friction coefficient is 0.6 (slippery road surface) and other data, which are constructed into a real-time state data matrix. Apply the fast feature extraction algorithm to this matrix, and calculate the time-domain features such as the standard deviation of the yaw rate 0.08 rad / s, the frequency-domain features such as the main frequency of the yaw rate 2.3 Hz, and the statistical features such as the skewness of the lateral acceleration 1.2, etc., to form a 15-dimensional real-time feature vector in total. Project this feature vector into a 3D feature space through the phase space mapping transformation method to generate a real-time low-dimensional discriminant point (0.68, -0.42, 0.15). Calculate the positional relationship between this discriminant point and the dynamic safety level boundary, and find that it is located within the "slightly unstable state" cluster, close to the boundary, and the distance from the boundary of the "obviously unstable state" cluster is shrinking. Analyze the movement trajectory of the discriminant point, and find that its velocity vector (0.04, -0.03, 0.01) points to the "obviously unstable state" cluster, indicating that the safety state is deteriorating. Based on the attribution category "slightly unstable state", the short distance to the boundary, and the trend of moving towards a more unsafe area, construct a safety risk level of "warning". Query the preset risk threshold comparison table according to the risk level "warning", and combined with the current specific conditions of the slippery road surface and the large steering angle, generate a warning message: "Warning: The lateral stability of the vehicle has decreased, and there is a risk of sideslip under the current slippery road surface conditions. It is recommended to decelerate and avoid sharp steering operations", and at the same time trigger a yellow visual warning and a short sound prompt to guide the driver to take appropriate measures to avoid the deterioration of the risk.

[0133] The above describes the classification method for motor vehicle safety performance based on multi-factor coupling effects in the embodiments of the present application. Next, the classification system for motor vehicle safety performance based on multi-factor coupling effects in the embodiments of the present application will be described. Please refer to Figure 2 , an embodiment of the classification system for motor vehicle safety performance based on multi-factor coupling effects in the embodiments of the present application includes:

[0134] An acquisition module, configured to collect original data of multi-dimensional safety influencing factors by collecting motor vehicle dynamic parameters, environmental conditions, and driving behavior data through a distributed sensor network;

[0135] A calculation module, configured to calculate mutual information values and transfer entropy for the original data of multi-dimensional safety influencing factors, and generate a non-linear factor coupling relationship network diagram;

[0136] A construction module, configured to construct a three-layer structure safety index system according to the non-linear factor coupling relationship network diagram to form a comprehensive score of motor vehicle safety performance;

[0137] A conversion module, configured to convert and extract key features from the comprehensive score of motor vehicle safety performance through phase space mapping to obtain a low-dimensional discriminant feature set;

[0138] A clustering module, configured to apply an adaptive density clustering method to the low-dimensional discriminant feature set to determine dynamic safety level boundary values;

[0139] A classification module, configured to perform multi-scenario classification on the real-time collected motor vehicle state by using the dynamic safety level boundary values, and output hierarchical safety risk warning information.

[0140] Through the collaborative cooperation of the above-mentioned various components, dynamic parameters, environmental conditions, and driving behavior data of motor vehicles are collected through a distributed sensor network to obtain raw data of multi-dimensional safety influencing factors, realizing an all-round and multi-angle perception of the safety state of motor vehicles and providing a rich basic data source for safety performance classification; by calculating the mutual information value and transfer entropy for the raw data of multi-dimensional safety influencing factors, a non-linear factor coupling relationship network diagram is generated, accurately capturing and quantifying the complex non-linear interaction between different safety factors, breaking through the limitation of traditional methods that ignore factor coupling; a three-layer structure safety index system is constructed based on the non-linear factor coupling relationship network diagram to form a comprehensive score of the safety performance of motor vehicles, establishing a complete index link from raw data to the final evaluation, making the evaluation result have a hierarchical interpretability; the comprehensive score of the safety performance of motor vehicles is transformed through phase space mapping and key features are extracted to obtain a low-dimensional discriminant feature set, applying non-linear dynamics analysis methods to solve the problem of effective expression of high-dimensional data, improving the calculation efficiency and accuracy of classification; the adaptive density clustering method is applied to the low-dimensional discriminant feature set to determine the boundary value of the dynamic safety level, realizing the adaptive classification of the safety state through the artificial intelligence density clustering algorithm, overcoming the rigidity problem of traditional fixed threshold classification; the dynamic safety level boundary value is used to classify the real-time collected motor vehicle states in multiple scenarios, and hierarchical safety risk warning information is output, realizing the accurate assessment and timely warning of safety risks. The artificial intelligence algorithm features in this solution, especially the calculation of mutual information and transfer entropy, phase space mapping, adaptive density clustering, etc., have made key contributions to the solution: the mutual information and transfer entropy algorithms can discover and quantify non-linear association patterns from a large amount of raw data, breaking through the limitation of traditional linear analysis; the phase space mapping and non-linear dynamics feature extraction algorithms can capture the dynamic evolution law of the safety state of motor vehicles, enhancing the prediction ability of the system; the adaptive density clustering algorithm realizes the dynamic adjustment of the safety state boundary, improving the adaptability and accuracy of classification in different scenarios. The organic combination of these algorithms forms a complete intelligent analysis framework, enabling the motor vehicle safety performance classification method to have comprehensive capabilities of multi-dimensional perception, coupling analysis, dynamic adaptation, and accurate warning, significantly improving the comprehensiveness and accuracy of motor vehicle safety state assessment.

[0141] Above Figure 2 The motor vehicle safety performance classification system based on multi-factor coupling influence in the embodiments of the present invention is described in detail from the perspective of modular functional entities. Next, the motor vehicle safety performance classification device based on multi-factor coupling influence in the embodiments of the present invention is described in detail from the perspective of hardware processing.

[0142] Figure 3FIG. 0 is a schematic structural diagram of a motor vehicle safety performance classification device based on multi-factor coupling influence provided by an embodiment of the present invention. The motor vehicle safety performance classification device 300 based on multi-factor coupling influence may vary greatly due to different configurations or performances, and may include one or more processors (central processing units, CPUs) 310 (for example, one or more processors) and a memory 320, and one or more storage media 330 for storing application programs 333 or data 332 (for example, one or more mass storage device terminals). Among them, the memory 320 and the storage media 330 may be transient storage or persistent storage. The program stored in the storage media 330 may include one or more modules (not shown in the figure), and each module may include a series of instruction operations on the motor vehicle safety performance classification device 300 based on multi-factor coupling influence. Further, the processor 310 may be configured to communicate with the storage media 330 and execute a series of instruction operations in the storage media 330 on the motor vehicle safety performance classification device 300 to implement the steps of the above-mentioned motor vehicle safety performance classification method based on multi-factor coupling influence.

[0143] The motor vehicle safety performance classification device 300 based on multi-factor coupling influence may further include one or more power supplies 340, one or more wired or wireless network interfaces 350, one or more input / output interfaces 360, and / or, one or more operating systems 331, such as Windows Serve, Mac OS X, Unix, Linux, FreeBSD, etc. Those skilled in the art can understand that Figure 3 The shown structural diagram of the motor vehicle safety performance classification device based on multi-factor coupling influence does not limit the motor vehicle safety performance classification device provided by the present invention, and may include more or fewer components than shown in the figure, or combine some components, or have different component arrangements.

[0144] The present invention also provides a computer-readable storage medium, which may be a non-volatile computer-readable storage medium or a volatile computer-readable storage medium. Instructions are stored in the computer-readable storage medium. When the instructions are run on a computer, the computer is caused to execute the steps of the motor vehicle safety performance classification method based on multi-factor coupling influence.

[0145] Those skilled in the art can clearly understand that for the convenience and brevity of description, the specific working processes of the above-described systems, systems, and units may refer to the corresponding processes in the foregoing method embodiments and will not be described herein again.

[0146] When the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes several instructions to enable a motor vehicle safety performance classification device based on multi-factor coupling influence (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The foregoing storage medium includes: various media that can store program codes such as USB flash drives, mobile hard disks, read-only memories (ROM), random access memories (RAM), magnetic disks, or optical discs.

[0147] As described above, the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that they can still modify the technical solutions described in the foregoing embodiments, or perform equivalent replacements on some of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the various embodiments of the present invention.

Claims

1. A classification method for the safety performance of motor vehicles based on the coupled influence of multiple factors, characterized in that, The method includes: Collecting motor vehicle dynamic parameters, environmental conditions, and driving behavior data through a distributed sensor network to obtain raw data of multi-dimensional safety influencing factors; Calculating mutual information values and transfer entropy for the raw data of the multi-dimensional safety influencing factors to generate a non-linear factor coupling relationship network diagram; Constructing a three-layer structure safety index system based on the non-linear factor coupling relationship network diagram to form a comprehensive score of motor vehicle safety performance; Converting and extracting key features of the comprehensive score of motor vehicle safety performance through phase space mapping to obtain a low-dimensional discriminant feature set; Applying an adaptive density clustering method to the low-dimensional discriminant feature set to determine the boundary values of dynamic safety levels; Using the boundary values of the dynamic safety levels to perform multi-scenario classification on the real-time collected motor vehicle states and outputting hierarchical safety risk warning information.

2. The classification method for the safety performance of a motor vehicle based on the influence of multi-factor coupling according to claim 1, wherein The collecting motor vehicle dynamic parameters, environmental conditions, and driving behavior data through a distributed sensor network to obtain raw data of multi-dimensional safety influencing factors includes: Collecting vehicle motion state data including acceleration, yaw rate, and sideslip angle through vehicle dynamic parameter sensors; Collecting environmental state data including rainfall, light intensity, and road surface friction coefficient through environmental monitoring sensors; Collecting driving operation data including steering wheel angle, pedal pressure, and driver attention distribution through a driving behavior monitoring device; Transmitting the vehicle motion state data, the environmental state data, and the driving operation data to a central processing unit through an in-vehicle high-speed CAN bus network; Using a real-time data calibration algorithm to perform sensor drift compensation and system error correction on the collected data; According to the change characteristics of different parameters, adopting an adaptive sampling frequency adjustment mechanism to optimize the collection of the vehicle motion state data, the environmental state data, and the driving operation data to obtain the raw data of the multi-dimensional safety influencing factors.

3. The classification method for the safety performance of a motor vehicle based on the combined influence of multiple factors according to claim 1, characterized in that, The calculating mutual information values and transfer entropy for the raw data of the multi-dimensional safety influencing factors to generate a non-linear factor coupling relationship network diagram includes: Performing outlier detection and smoothing processing on the raw data of the multi-dimensional safety influencing factors to obtain a preprocessed data set; Calculating the joint probability distribution for factor pairs in the preprocessed data set through kernel density estimation to obtain a mutual information value matrix between factors; Segmenting the preprocessed data set based on a sliding time window and calculating the transfer entropy between factors through symbolic transformation to form a directional coupling strength matrix; Fusing the mutual information value matrix and the directional coupling strength matrix to construct a multi-level factor coupling network topology structure; Performing redundant connection pruning on the multi-level factor coupling network topology structure through a sparse representation learning method to retain key coupling paths; Analyzing the characteristics of the multi-level factor coupling network topology structure from three scales of micro, meso, and macro to generate the non-linear factor coupling relationship network diagram.

4. The method for classifying the safety performance of a motor vehicle based on the combined influence of multiple factors according to claim 1, characterized in that The constructing a three-layer structure safety index system based on the non-linear factor coupling relationship network diagram to form a comprehensive score of motor vehicle safety performance includes: Extracting basic safety indicators from the raw data of the multi-dimensional safety influencing factors to construct the bottom layer structure of the index system; Identify key coupling factor combinations based on the coupling path strength in the non - linear factor coupling relationship network diagram, and generate intermediate - level coupling indicators; Calculate the information entropy for the intermediate - level coupling indicators to obtain the index weight distribution coefficients; Combine the intermediate - level coupling indicators with the index weight distribution coefficients, and form a top - level comprehensive safety performance indicator through fuzzy integral operation; Perform scenario calibration on the top - level comprehensive safety performance indicator according to different driving scenario characteristics to obtain the scenario - adaptable safety indicator value; Standardize the scenario - adaptable safety indicator value to form the comprehensive score of the motor vehicle safety performance.

5. The classification method for the safety performance of motor vehicles based on the combined influence of multiple factors according to claim 1, wherein, The above - mentioned comprehensive score of the motor vehicle safety performance is transformed through phase - space mapping and key features are extracted to obtain a low - dimensional discriminant feature set, including: Construct a time - series data matrix for the comprehensive score of the motor vehicle safety performance, and determine the optimal time - delay parameter through the mutual information method; Analyze the time - series data matrix using the false nearest - neighbor method to determine the optimal embedding dimension; According to the optimal time - delay parameter and the optimal embedding dimension, reconstruct the comprehensive score of the motor vehicle safety performance into a high - dimensional phase - space trajectory; Calculate the non - linear dynamic characteristic parameters of the high - dimensional phase - space trajectory, including the maximum Lyapunov exponent, correlation dimension, and approximate entropy value; Extract complex network features from the high - dimensional phase - space trajectory to obtain network topology feature quantization indicators; Combine the non - linear dynamic characteristic parameters and the network topology feature quantization indicators, and obtain the low - dimensional discriminant feature set through non - linear mapping transformation.

6. The classification method for the safety performance of a motor vehicle based on the influence of multi-factor coupling according to claim 1, characterized in that Apply the adaptive density clustering method to the low - dimensional discriminant feature set to determine the dynamic safety - level boundary values, including: Calculate the local density value and distance factor of each feature point in the low - dimensional discriminant feature set, and construct a density - distance decision graph; Identify the initial clustering center points from the density - distance decision graph through the density - peak detection method; Perform clustering division according to the distribution characteristics of the initial clustering center points to obtain the initial safety - level clusters; Evaluate the effectiveness of the initial safety - level clusters, and determine the optimal number of clusters through silhouette coefficient calculation; Construct a closed decision surface around the boundary of the safety - level cluster corresponding to the optimal number of clusters to form a static safety - level boundary; Analyze the temporal variation characteristics of the safety - level clusters based on historical data, establish boundary dynamic update rules, and generate the dynamic safety - level boundary values.

7. The classification method for the safety performance of a motor vehicle based on the influence of multi-factor coupling according to claim 1, characterized in that, Use the dynamic safety - level boundary values to perform multi - scenario classification on the real - time collected motor vehicle states, and output hierarchical safety risk warning information, including: Receive the motor vehicle dynamic parameters, environmental conditions, and driving behavior data collected by the distributed sensor network through the real - time data acquisition interface, and construct a real - time state data matrix; Apply a fast feature extraction algorithm to the real - time state data matrix to obtain a real - time feature vector; Project the real - time feature vector into the feature space where the low - dimensional discriminant feature set is located through the phase - space mapping transformation method to generate real - time low - dimensional discriminant points; Calculate the positional relationship between the real - time low - dimensional discriminant points and the dynamic safety - level boundary values to determine the category to which the current motor vehicle safety state belongs; Construct a safety risk level based on the current motor vehicle safety status attribution category and the dynamic change trend of the real-time low-dimensional discriminant points; Generate the classified safety risk warning information according to the comparison table of the safety risk level and the preset risk threshold; 8. A motor vehicle safety performance classification system based on the coupled influence of multiple factors, characterized in that, For implementing the motor vehicle safety performance classification method based on multi-factor coupling influence as described in any one of claims 1-7, the motor vehicle safety performance classification system based on multi-factor coupling influence includes: An acquisition module, configured to acquire the original data of multi-dimensional safety influence factors by collecting the dynamic parameters, environmental conditions and driving behavior data of the motor vehicle through a distributed sensor network; A calculation module, configured to calculate the mutual information value and transfer entropy for the original data of the multi-dimensional safety influence factors, and generate a non-linear factor coupling relationship network diagram; A construction module, configured to construct a three-layer structure safety index system according to the non-linear factor coupling relationship network diagram, and form a comprehensive score of the motor vehicle safety performance; A conversion module, configured to convert and extract key features of the comprehensive score of the motor vehicle safety performance through phase space mapping, and obtain a low-dimensional discriminant feature set; A clustering module, configured to apply an adaptive density clustering method to the low-dimensional discriminant feature set to determine the dynamic safety level boundary value; A classification module, configured to perform multi-scenario classification on the real-time collected motor vehicle status by using the dynamic safety level boundary value, and output the classified safety risk warning information; 9. A classification device for the safety performance of motor vehicles based on the coupled influence of multiple factors, characterized in that, It includes a memory and a processor, the memory stores a computer program that can run on the processor, and when the processor executes the computer program, it implements the motor vehicle safety performance classification method based on multi-factor coupling influence as described in any one of claims 1 to 7; 10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is run by the processor, the processor is caused to execute the motor vehicle safety performance classification method based on multi-factor coupling influence as described in any one of claims 1 to 7.

Images