A data feature driving-based automatic identification method for vehicle handling stability test project

Abstract

The application discloses a kind of based on data feature driving's car steering stability test project automatic identification method, to solve the scientificity and rigor problem of insufficient of existing steering stability test project identification method because of the correlation of feature physical quantity and test project cannot be quantified.Firstly, the test feature physical quantity is initially selected, and the maximum information coefficient algorithm is introduced, to filter all the feature physical quantities collected in the test, obtain the feature physical quantity subset that meets the correlation threshold requirement, then the random forest algorithm is used to build the test project identification model, to train the test project model, realize the automatic identification of test project.The application uses maximum information coefficient algorithm to quantify the correlation of feature physical quantity and test project, and then the feature physical quantity with high correlation is used as the input of random forest model for test project identification, to ensure the accuracy of identification, so that the identification process is more scientific and rigorous.

Description

Technical Field

[0001] This invention relates to an automatic identification method for vehicle handling stability test items, specifically to an automatic identification method for vehicle handling stability test items based on data feature-driven approaches. Background Technology

[0002] In the automotive product development process, real-vehicle track testing is a crucial verification method to ensure product performance and quality. Handling stability, as a core factor affecting vehicle driving safety and experience, is an important component of product competitiveness. Therefore, extensive real-vehicle track testing of handling stability must be conducted during the product development phase. However, handling stability testing involves numerous operating conditions and generates massive amounts of data, making traditional manual processing methods inefficient. To improve testing efficiency, OEMs are actively exploring the use of automation to solve the problem of processing massive amounts of data, and automatic identification technology for test items is the foundation and starting point for achieving automated data processing.

[0003] Based on literature review, there are two main methods for automatic identification of vehicle handling stability test items, illustrated below: Zhong Zhaohui et al., in their paper "A Method for Identifying Vehicle Handling Stability Test Types Based on Convolutional Neural Networks," analyzed the image features of vehicle handling stability test items and established a convolutional neural network model to classify the test items. Lu Pingping et al., in patent CN107238500A "A Method for Establishing a Rapid Evaluation System for Vehicle Handling Stability Tests," employed dynamic time warping and neural network methods to classify the test items.

[0004] Existing methods all rely on human experience, starting from the characteristics of each test item, selecting a characteristic physical quantity, and using its time-domain curve features as the feature for test item identification. However, this method heavily depends on human experience and does not quantify the correlation between the characteristic physical quantity and the test item, which makes the identification process lack sufficient scientific rigor and precision. Summary of the Invention

[0005] To address the shortcomings in scientific rigor and scientific validity of existing methods for identifying vehicle handling stability test items due to their inability to quantify the correlation between characteristic physical quantities and test items, this invention provides an automatic identification method for vehicle handling stability test items based on data feature-driven approaches. First, the test characteristic physical quantities are initially selected, and a maximum information coefficient algorithm is introduced to filter all characteristic physical quantities collected from the test, obtaining a set of characteristic physical quantum quantities that meet the correlation threshold requirements. Then, a random forest algorithm is used to establish an identification model, which is trained to achieve automatic identification of test items.

[0006] The objective of this invention is achieved through the following technical solution:

[0007] An automatic identification method for vehicle handling stability test items based on data feature-driven approach includes the following steps:

[0008] S1. Initially select characteristic physical quantities for maneuver stability testing and prepare test data;

[0009] S2. Preprocess the experimental data prepared in step S1 and define data labels;

[0010] S3. Feature selection of characteristic physical quantities in the manipulation stability test based on the maximum information coefficient algorithm;

[0011] S4. Construct an experimental project identification model based on the random forest algorithm;

[0012] S5. Train the experimental project identification model constructed in step S4.

[0013] Further, step S1 includes:

[0014] S11. Determine the characteristic physical quantities of the handling stability test, including: steering wheel angle, steering wheel torque, steering wheel angular velocity, lateral acceleration, yaw rate, roll angle, vehicle speed, and longitudinal acceleration;

[0015] S12. Prepare test data for each test in the vehicle handling stability test according to the handling stability test characteristic physical quantities determined in step S11.

[0016] Further, step S2 includes:

[0017] S21. Perform low-pass filtering and standardization on the test data prepared in step S1;

[0018] S22. Define data labels: Define different labels for experimental data under different experimental projects. The data after defining labels are used as the candidate dataset in the feature screening algorithm in step S3. Each label represents a candidate input variable.

[0019] Further, step S21 includes:

[0020] First, a one-dimensional median filtering method is used to remove outliers; assuming the original data sequence is {p1, p2, ... p...} i ,…p q The filter window length is set to L = 2k + 1, and the data within the window is {p}. i-k ,…,p i ,…p i+k} Sort the L data points within the window by size, and the filtered output data is the sorted center point data within the window; the filtered output expression is formula (1):

[0021] yi =median[{p i-k ,…,p i ,…p i+k (1)

[0022] In the formula, i∈[ik,…,i,…,i+k]; y i This is the output of a median filter; "median" indicates median filtering.

[0023] Then, an FIR filter is used for low-pass filtering, and the filtered test data is phase corrected to eliminate the time domain lag caused by filtering.

[0024] Finally, the experimental data were standardized to ensure that the characteristic physical quantities had the same order of magnitude, as expressed by:

[0025] x′ ik =(x ik -x min ) / (x max -x min (2)

[0026] In the formula, x′ ik For the standardized data, x ik For the observed value, x min x is the minimum value of the observed values. max This represents the maximum value of the observed values.

[0027] Further, step S3 includes:

[0028] S31. Construct a set of candidate variables, defining the eight candidate physical quantities as the set of candidate variables X = {X1, X2, ..., X3}. α}, i∈[1,8], define the characteristic physical quantum set S as an empty set;

[0029] S32. The maximum information coefficient algorithm is used to calculate the correlation between the characteristic physical quantities and data labels in the experimental data. The expression for the maximum correlation coefficient is as follows:

[0030]

[0031] In the formula, X represents the characteristic physical quantity; Y represents the data label; p(x,y) represents the joint probability distribution function of X and Y; p(x) represents the marginal probability distribution function of X; p(y) represents the marginal probability distribution function of Y; m and n represent the number of grids in the x and y directions of the two-dimensional space; B is a variable, and the size of B is set to 0.6 times the data volume.

[0032] S33. According to formula (3), in the variable set X={X1,X2…X… α}, calculate each variable X in α∈[1,8]. α The MIC value of the data label is used to record the candidate feature physical quantity X with the largest value. α The characteristic physical quantity is included in the sample set S, and the characteristic physical quantity X is removed from the candidate variable set X. α ;

[0033] S34. Determine the remaining candidate feature physical quantities X α The correlation with the test project type Y is determined. If the correlation discrimination condition is not met, the process returns to step S33 to continue filtering. If the correlation discrimination condition is met, it means that the remaining characteristic physical quantities of set X do not meet the set correlation threshold with the test project type, and the characteristic physical quantum set S is output.

[0034] The correlation discrimination condition expression is as follows:

[0035] MIC(Y;X) a )<η·MIC(Y;Y) (4)

[0036] In formula (4), η is the correlation threshold. If MIC(Y;X) a If ) < η·MIC(Y;Y), then the variable is considered to not contain information about category Y and is judged as an irrelevant variable;

[0037] The threshold η for correlation is selected to satisfy formula (5):

[0038] minMIC(X a ;Y) / MIC(Y;Y)≤η≤maxMIC(X a ;Y) / MIC(Y;Y) (5)

[0039] In the formula, minMIC(X) a ;Y) / MIC(Y;Y) represents the lower bound of η, maxMIC(X) a ;Y) represents the upper limit of η, where η∈[0,1].

[0040] Further, step S4 includes:

[0041] Random forest uses a bootstrap sampling method, taking a dataset containing physical quantity features such as vehicle speed, roll angle, yaw rate, steering wheel angle, and steering wheel torque as the original dataset, and extracting multiple samples from the original dataset to form multiple subset datasets;

[0042] Each subset of data is used to construct a decision tree. During the tree construction process, a subset of features is randomly selected for splitting. Each split is quantified and scored based on Gini impurity, and the feature with the lowest Gini impurity is selected as the node. The Gini impurity formula is as follows:

[0043] Gini(U)=∑p(u)(1-p(u i (6)

[0044] In formula (6), p(u i ) represents the probability that a random sample belongs to class i;

[0045] Set a specified number of decision trees until the specified number of decision trees are built;

[0046] The prediction results of all decision trees are statistically analyzed using a voting strategy. The output of the model after the statistical analysis is the category of the experimental item, which is the data label defined in step S2.

[0047] Further, step S5 includes:

[0048] The data with predefined labels in step S2 is divided into a training set and a test set;

[0049] The training set is used to develop the experimental item identification model constructed in step S4;

[0050] After the model is trained, the training effect is verified using test set data to ensure the accuracy of the model's recognition results.

[0051] The present invention can bring the following beneficial effects:

[0052] This invention provides an automatic identification method for vehicle handling stability test items based on data feature-driven approaches. First, it initially selects test feature physical quantities and introduces a maximum information coefficient algorithm to filter all collected feature physical quantities, obtaining a set of feature physical quantities that meet the correlation threshold requirements. Then, it uses a random forest algorithm to build a test item identification model, trains the model, and achieves automatic identification of test items. This invention uses the maximum information coefficient algorithm to quantify the correlation between feature physical quantities and test items. Secondly, it uses highly correlated feature physical quantities as input to the random forest model for test item identification, ensuring accuracy and making the identification process more scientific and rigorous. Attached Figure Description

[0053] Figure 1 This is a flowchart of an automatic identification method for vehicle handling stability test items based on data feature-driven methods, as described in an embodiment of the present invention.

[0054] Figure 2 This is a flowchart of the feature filtering process based on the maximum information coefficient as described in this embodiment of the invention;

[0055] Figure 3 The training error curve of the test item identification model constructed for embodiments of the present invention;

[0056] Figure 4 The confusion matrix is ​​trained for the test item identification model constructed for the embodiments of the present invention;

[0057] Figure 5 The confusion matrix for testing the identification model of the experimental items trained according to the embodiments of the present invention;

[0058] Figure 6 The results are the test results of the experimental project recognition model trained according to the embodiments of the present invention. Detailed Implementation

[0059] The method of the present invention will now be described in detail in practical implementation with reference to the accompanying drawings and embodiments.

[0060] This embodiment presents an automatic identification method for vehicle handling stability test items based on data feature-driven approaches. First, it initially selects test feature physical quantities and introduces a maximum information coefficient algorithm to filter all collected feature physical quantities, obtaining a set of feature physical quantum quantities that meet the correlation threshold requirements. Then, a random forest algorithm is used to establish a test item identification model, which is then trained and evaluated, as shown in the attached figure. Figure 1 As shown, it mainly includes the following steps:

[0061] S1. Initially select characteristic physical quantities for maneuver stability testing and prepare test data:

[0062] S11. Determine the characteristic physical quantities of the handling stability test:

[0063] The existing GB / T 6323-2014 standard, "Test Methods for Handling Stability of Automobiles," includes seven tests: slalom test, steering wheel center zone handling stability test, steering return test, steering ease test, steady-state turning test, steering wheel angle step test, and steering wheel angle pulse test. In addition, the ISO standard also includes a double lane change test, steering release open-loop test, and single sine wave test.

[0064] Based on the data collection requirements specified in national and ISO standards, the characteristic physical quantities that need to be collected for the handling stability test are determined, including: steering wheel angle, steering wheel torque, steering wheel angular velocity, lateral acceleration, yaw rate, roll angle, vehicle speed, and longitudinal acceleration.

[0065] S12. Test data preparation: Prepare test data for each test according to the physical quantities of the handling stability test characteristics determined in step S11.

[0066] S2. Preprocess the experimental data prepared in step S1 and define data labels:

[0067] S21. Experimental data preprocessing: Perform low-pass filtering and standardization on the experimental data prepared in step S1.

[0068] First, a one-dimensional median filtering method is used to remove outliers. Assume the original data sequence is {p1, p2, ... p...} i ,…p q The filter window length is set to L = 2k + 1, and the data within the window is {p}. i-k ,…,p i ,…p i+k The L data points within the window are sorted by size, and the filtered output is the center point data after sorting within the window. The filtered output can be expressed by the formula:

[0069] y i =median[{p i-k ,…,p i ,…p i+k (1)

[0070] In the formula, i∈[ik,…,i,…,i+k]; y i This is the output of the median filter; median indicates median filtering.

[0071] Then, an FIR filter is used for low-pass filtering with a cutoff frequency of 5Hz. The filtered test data is then phase-corrected to eliminate the time-domain lag caused by filtering.

[0072] Next, the experimental data are standardized to eliminate the influence of different amplitudes, mean values, etc., so that the characteristic physical quantities have the same order of magnitude. The expression is as follows:

[0073] x′ ik =(x ik -x min ) / (x max -x min (2)

[0074] In the formula, x′ ik For the standardized data, x ik For the observed value, x min x is the minimum value of the observed values. max This represents the maximum value of the observed values.

[0075] S22. Define data labels:

[0076] Different labels are defined for the test data under different test items. For example, the data label for the serpentine test is defined as "1", and the data label for the steering wheel center area handling stability test is defined as "2". The labeled data is used as the candidate dataset in the feature selection algorithm of step S3. Each label represents a candidate input variable. The correspondence between test items and data labels is shown in Table 1.

[0077] Table 1. Correspondence between test items and data labels

[0078] Experimental Project Data Tags Snake test 1 Steering wheel center zone handling stability test 2 Steering return test 3 Steering ease test 4 Steady-state rotation test 5 Double line shift test 6 Steering release open-loop test 7 Single sine test 8

[0079] S3. Feature selection of characteristic physical quantities in manipulation stability tests based on the Maximum Information Coefficient (MIC) algorithm, such as... Figure 2 As shown:

[0080] S31. Construct a set of candidate variables, defining the eight characteristic physical quantities to be selected, such as steering wheel angle, steering wheel torque, and steering wheel angular velocity, as the set of candidate variables X = {X1, X2, ..., X3}. α}, i∈[1,8]. Define the characteristic physical quantum set S as an empty set.

[0081] The correspondence between the candidate variable names and variable symbols is shown in Table 2:

[0082] Table 2. Correspondence between candidate variables and variable signs

[0083]

[0084]

[0085] In this embodiment, the set of characteristic physical quantities X = {X1, X2, ... X8} of the 8 test items in Table 1, α ∈ [1, 8], is used as the input to the maximum information coefficient algorithm, and the data labels of the corresponding items are used as the output.

[0086] S32. The correlation between the characteristic physical quantities and data labels in the experimental data is calculated using the maximum information coefficient algorithm. The expression for the maximum correlation coefficient is as follows:

[0087]

[0088] In formula (3), X is a random variable, representing a characteristic physical quantity; Y is a random variable, representing a data label; p(x,y) is the joint probability distribution function of X and Y; p(x) is the marginal probability distribution function of X; p(y) is the marginal probability distribution function of Y; m and n are the number of grids in the x and y directions of the two-dimensional space; B is a variable, and the size of B is set to the power of 0.6 of the data volume.

[0089] S33. According to formula (3), in the variable set X={X1,X2…X… α}, α∈[1,8] calculate each variable X α The MIC value of the data label is used to record the candidate feature physical quantity X with the largest value. α The characteristic physical quantity is included in the sample set S, and the characteristic physical quantity X is removed from the candidate variable set X. α .

[0090] In this embodiment, the maximum information coefficients of each physical feature quantity and the project category are calculated according to formula (3), sorted from largest to smallest, and the cumulative contribution rate of the physical feature quantity to the model classification is shown in Table 3.

[0091] Table 3. Characteristic Physical Quantity Set MIC Values

[0092]

[0093]

[0094] The MIC values ​​of each feature variable and the category label are calculated according to formula (3). As shown in Table 3:

[0095] minMIC(X a ;Y)=0.039,maxMIC(X a ;Y)=0.228. MIC(Y;Y)=1

[0096] S34. Determine the remaining candidate feature physical quantities X α The correlation with the test project type Y, when the discrimination condition of formula (4) is not met, indicates that the remaining characteristic physical quantities in set X are still related to the test project type, so return to step S33 and continue screening; when the discrimination condition is met, it means that the remaining characteristic physical quantities in set X do not meet the set correlation threshold with the test project type, and output the characteristic physical quantum set S.

[0097] The correlation discriminant formula is as follows:

[0098] MIC(Y;X) a )<η·MIC(Y;Y) (4)

[0099] In formula (4), η is the correlation threshold. If MIC(Y;X) a If ) < η·MIC(Y;Y), then the variable is considered not to contain information about category Y and is judged as an irrelevant variable.

[0100] Formula (5) can provide guidance for selecting the correlation threshold η:

[0101] minMIC(X a ;Y) / MIC(Y;Y)≤η≤maxMIC(X a ;Y) / MIC(Y;Y) (5)

[0102] In the formula, minMIC(X) a ;Y) / MIC(Y;Y) represents the lower bound of η, maxMIC(X) a ;Y) represents the upper limit of η, where η∈[0,1].

[0103] In this embodiment, according to formula (5), the specific range of the correlation threshold η is η∈[0.039,0.228].

[0104] Based on the specific range of the correlation threshold η, η is set to 0.1, and the characteristic physical quantities that conform to formula (3) are obtained as the characteristic physical quantum set S. The characteristic subset S contains the features {vehicle speed, roll angle, yaw rate, steering wheel angle, and steering wheel torque}.

[0105] In this embodiment, the calculation results of the feature subset S are combined with the following four factors:

[0106] 1. In the feature selection process, the individual contribution rate of each feature and their cumulative contribution to the overall model were mainly considered. Table 3 analyzes the individual contribution rates of all features. Three features had individual contribution rates below the correlation threshold of 0.1, indicating that they contributed little to the model's classification ability. Removing these three features simplifies the model and reduces computational cost.

[0107] 2. After removing the above features, the cumulative contribution rate of the remaining features was analyzed. The results show that after removing these three features, the cumulative contribution rate of the remaining features still reaches 80.1%, which can maintain the prediction accuracy of the model.

[0108] 3. Regarding longitudinal acceleration, since the steering tests were all conducted under constant speed conditions, the values ​​fluctuated around zero and were not strongly correlated with the test items.

[0109] 4. Lateral acceleration and yaw rate have a strong coupling relationship. According to the maximum information coefficient algorithm, yaw rate is more sensitive to capturing vehicle dynamic response characteristics. Therefore, yaw rate was given priority in feature selection, while lateral acceleration was excluded.

[0110] Ultimately, vehicle speed, roll angle, yaw rate, steering wheel angle, and steering wheel torque were selected as features for the classification model. These features not only have a high contribution rate but also comprehensively reflect the vehicle's handling stability, effectively ensuring the accuracy of the classification model.

[0111] S4. Constructing an experimental project identification model based on the random forest algorithm:

[0112] Random forests employ a bootstrap sampling method (random sampling with replacement) to extract multiple samples from the original dataset, forming multiple subsets. In this embodiment, the dataset containing physical quantity features such as vehicle speed, roll angle, yaw rate, steering wheel angle, and steering wheel torque is used as the original dataset, and multiple samples are extracted from the original dataset to form multiple subsets.

[0113] Each subset of data is used to construct a decision tree. During the tree construction process, a subset of features is randomly selected for splitting. Each split is quantified and scored based on Gini impurity, and the feature with the lowest Gini impurity is selected as the node. The Gini impurity formula is as follows:

[0114] Gini(U)=∑p(u)(1-p(u i (6)

[0115] In formula (6), p(u i ) represents the probability that a random sample belongs to class i.

[0116] Set a specified number of decision trees until the specified number of decision trees are constructed. In this embodiment, the specified number of books for the decision trees is 100, and the minimum number of leaf node samples is set to 1.

[0117] The prediction results of all decision trees are statistically analyzed using a voting strategy. The output of the model after the statistical analysis is the category of the test item, i.e., the label defined in step S2.

[0118] S5. Train the experimental project identification model constructed in step S4.

[0119] Divide the data with the labels defined in step 2 into training and testing sets in a 7:3 ratio.

[0120] The test project recognition model constructed in step S4 is trained using the training set. After the model is trained, the training effect is verified using the test set data to ensure the accuracy of the model recognition results.

[0121] In this embodiment, the error curve of model training is as follows: Figure 3 As shown, observe Figure 3 The error curve shows a rapid decrease during the training phase, gradually decreasing as the number of decision trees increases. After 50 decision trees, the error curve begins to converge, indicating that the model has stably learned the key features of the data, and additional decision trees no longer significantly improve model performance. 100 decision trees provide sufficient capacity for the model to capture the complexity and non-linear relationships in the data, while the minimum leaf node sample size of 1 allows the model to generate more detailed decision boundaries, thereby improving classification accuracy.

[0122] Visualize the training results and inversion using a confusion matrix. Figure 4 The training confusion matrix of the classification model shows that the model training set contains a total of 1305 samples, of which 1305 samples are correctly classified, and the model training accuracy is 100%.

[0123] After model training is complete, the training effect is validated using test set data to ensure the accuracy of the model's recognition results. Observation Figure 5 The test confusion matrix of the classification model and Figure 6 The classification model test results show that out of 559 samples in the test set, 555 were correctly classified and 4 were misclassified. The test set accuracy was 99.28%, indicating that the model has very high recognition accuracy and meets the usage requirements.

[0124] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A method for automatic identification of vehicle handling stability test items based on data feature-driven approach, characterized in that, Includes the following steps: S1. Initially select characteristic physical quantities for maneuver stability testing and prepare test data; S2. Preprocess the experimental data prepared in step S1 and define data labels; S3. Feature selection of characteristic physical quantities in the manipulation stability test based on the maximum information coefficient algorithm; step S3 includes: S31. Construct a set of candidate variables, defining the eight candidate physical quantities as the set of candidate variables X = {X1, X2, ..., X3}. α }, i∈[1,8], define the characteristic physical quantum set S as an empty set; S32. The correlation between the characteristic physical quantities and data labels in the experimental data is calculated using the maximum information coefficient algorithm. The expression for the maximum correlation coefficient is as follows: In the formula, X represents the characteristic physical quantity; Y represents the data label; p(x,y) represents the joint probability distribution function of X and Y; p(x) represents the marginal probability distribution function of X; p(y) represents the marginal probability distribution function of Y; m and n represent the number of grids in the x and y directions of the two-dimensional space; B is a variable, and the size of B is set to 0.6 times the data volume. S33. According to formula (3), in the variable set X={X1,X2…X… α }, calculate each variable X in α∈[1,8]. α The MIC value of the data label is used to record the candidate feature physical quantity X with the largest value. α The characteristic physical quantity is included in the sample set S, and the characteristic physical quantity X is removed from the candidate variable set X. α ; S34. Determine the remaining candidate feature physical quantities X α The correlation with the test project type Y is determined. If the correlation discrimination condition is not met, the process returns to step S33 to continue filtering. If the correlation discrimination condition is met, it means that the remaining characteristic physical quantities of set X do not meet the set correlation threshold with the test project type, and the characteristic physical quantum set S is output. The correlation discrimination condition expression is as follows: MIC(Y;X a )<η·MIC(Y;Y) (4) In formula (4), η is the correlation threshold. If MIC(Y;X) a If ) < η·MIC(Y;Y), then the variable is considered to not contain information about category Y and is judged as an irrelevant variable; The threshold η for correlation is selected to satisfy formula (5): minMIC(X a ;Y) / MIC(Y;Y)≤η≤maxMIC(X a ;Y) / MIC(Y;Y) (5) In the formula, minMIC(X) a ;Y) / MIC(Y;Y) represents the lower bound of η, maxMIC(X) a ;Y) represents the upper limit of η, where η∈[0,1]; S4. Constructing a test project identification model based on the random forest algorithm; step S4 includes: Random forest uses a bootstrap sampling method, taking a dataset containing physical quantity features such as vehicle speed, roll angle, yaw rate, steering wheel angle, and steering wheel torque as the original dataset, and extracting multiple samples from the original dataset to form multiple subset datasets; Each subset of data is used to construct a decision tree. During the tree construction process, a subset of features is randomly selected for splitting. Each split is quantified and scored based on Gini impurity, and the feature with the lowest Gini impurity is selected as the node. The Gini impurity formula is as follows: Gini(U)=∑p(u)(1-p(u i )) (6) In formula (6), p(u i ) represents the probability that a random sample belongs to class i; Set a specified number of decision trees until the specified number of decision trees are built; The prediction results of all decision trees are statistically analyzed using a voting strategy. The output of the model after the statistics are the categories of the experimental items, i.e., the data labels defined in step S2. S5. Train the experimental project identification model constructed in step S4.

2. The automatic identification method for vehicle handling stability test items based on data feature driving as described in claim 1, characterized in that, Step S1 includes: S11. Determine the characteristic physical quantities of the handling stability test, including: steering wheel angle, steering wheel torque, steering wheel angular velocity, lateral acceleration, yaw rate, roll angle, vehicle speed, and longitudinal acceleration; S12. Prepare test data for each test in the vehicle handling stability test according to the handling stability test characteristic physical quantities determined in step S11.

3. The automatic identification method for vehicle handling stability test items based on data feature driving as described in claim 1, characterized in that, Step S2 includes: S21. Perform low-pass filtering and standardization on the test data prepared in step S1; S22. Define data labels: Define different labels for experimental data under different experimental projects. The data after defining labels are used as the candidate dataset in the feature screening algorithm in step S3. Each label represents a candidate input variable.

4. The automatic identification method for vehicle handling stability test items based on data feature driving as described in claim 3, characterized in that, Step S21 includes: First, a one-dimensional median filtering method is used to remove outliers; assuming the original data sequence is {p1, p2, ... p...} i ,…p q The filter window length is set to L = 2k + 1, and the data within the window is {p}. i-k ,…,p i ,…p i+k } Sort the L data points within the window by size, and the filtered output data is the sorted center point data within the window; the filtered output expression is formula (1): y i =median[{p i-k ,…,p i ,…p i+k ] (1) In the formula, i∈[ik,…,i,…,i+k]; y i This is the output of a median filter; "median" indicates median filtering. Then, an FIR filter is used for low-pass filtering, and the filtered test data is phase corrected to eliminate the time domain lag caused by filtering. Finally, the experimental data were standardized to ensure that the characteristic physical quantities had the same order of magnitude, as expressed by: x′ ik =(x ik -x min ) / (x max -x min ) (2) In the formula, x′ik represents the standardized data, x ik For the observed value, x min x is the minimum value of the observed values. max This represents the maximum value of the observed values.

5. The automatic identification method for vehicle handling stability test items based on data feature driving as described in claim 1, characterized in that, Step S5 includes: The data with predefined labels in step S2 is divided into a training set and a test set; The training set is used to develop the experimental item identification model constructed in step S4; After the model is trained, the training effect is verified using test set data to ensure the accuracy of the model's recognition results.

Images