Performance evaluation method and system for automobile rear wheel active steering system

Abstract

The invention relates to the technical field of automobile electronic control and testing, and discloses a performance evaluation method and system for an automobile rear wheel active steering system. The method comprises the following steps: acquiring time sequence data of a control instruction, a rear wheel steering angle, a yaw velocity and a lateral acceleration of a system under a test working condition; performing alignment and difference calculation on an instruction and a rotation angle feedback time sequence to generate a following error sequence, and performing error distribution statistics based on a dynamic threshold value to obtain a dynamic following precision evaluation result; performing sequential conjoint analysis on the yaw velocity and the lateral acceleration to generate a response feature vector, matching the response feature vector with a preset stability reference model to obtain a response error vector, and calculating a stability deviation degree in combination with an error tolerance interval to obtain a dynamic stability evaluation result; and fusing the two to generate a comprehensive performance score. According to the method, the continuous error statistics of the dynamic process and the response quality quantification of multi-signal combination are carried out, so that the accurate and comprehensive evaluation on the influence of the tracking precision and the stability of the system is realized.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of automotive electronic control and testing technology, in particular to a performance evaluation method and system for a rear wheel active steering system of a vehicle. BACKGROUND

[0002] Currently, for the performance evaluation of the rear wheel active steering system of a vehicle, the industry generally adopts a method based on steady-state indicators or simple threshold judgments. For the evaluation of the response accuracy of the system, the existing technology usually directly compares the control instruction value of the rear wheel steering actuator at a certain moment with the feedback value of the steering angle sensor, calculates the static deviation, or observes a few time-domain characteristic parameters such as response time and overshoot. In terms of the evaluation of the influence of the system on the dynamic stability of the vehicle, the existing scheme mainly depends on whether the single numerical value of the yaw rate or the vehicle body lateral acceleration exceeds the preset safety threshold, so as to judge whether the system causes instability risk.

[0003] The above-mentioned prior art scheme has defects. The static deviation or the isolated time-domain parameter cannot accurately depict the continuous following performance of the system in the whole dynamic change process, ignores the time sequence mismatch problem between the control instruction and the sensor feedback caused by communication delay and asynchronous sampling, and makes the accuracy evaluation result one-sided and inaccurate. At the same time, the stability judgment method relying on only a single physiological signal threshold cannot consider the dynamic coupling relationship between yaw and lateral motion, and cannot distinguish between the ideal response change caused by the system intervention and the harmful instability tendency, so that the stability evaluation lacks comprehensiveness and cannot quantify the actual influence degree of the system on the lateral dynamic quality of the vehicle.

[0004] There is a need for an evaluation method that can solve the problem of inaccurate evaluation of the continuous following accuracy of the rear wheel active steering system in the dynamic process in the prior art, and the problem of incomplete evaluation and inability to quantify the influence of the system on the lateral dynamic stability of the vehicle. SUMMARY

[0005] The purpose of the present application is to provide a performance evaluation method and system for a rear wheel active steering system of a vehicle to solve the problems raised in the background art.

[0006] To achieve the above-mentioned purpose, the present application provides a performance evaluation method for a rear wheel active steering system of a vehicle, which comprises: obtaining a running data set of the rear wheel active steering system to be evaluated under a preset test condition, the running data set comprising a control instruction time sequence of a rear wheel steering actuator, a rear wheel steering angle sensor feedback time sequence, a vehicle yaw rate time sequence and a vehicle body lateral acceleration time sequence; performing time sequence alignment and difference calculation processing on the control instruction time sequence of the rear wheel steering actuator and the rear wheel steering angle sensor feedback time sequence to generate a steering angle instruction following error sequence; Based on the steering command following error sequence and the preset dynamic following accuracy threshold, the error distribution statistical processing is performed to generate the dynamic following accuracy evaluation result of the rear wheel active steering system; The vehicle yaw rate time series and the vehicle body lateral acceleration time series are jointly analyzed and processed to generate a vehicle lateral dynamic response feature vector. The vehicle's lateral dynamic response feature vector is matched with the desired response by calling a preset stability benchmark model to generate yaw and lateral dynamic response error vectors. Based on the yaw and lateral dynamic response error vectors and the preset stability error tolerance range, the stability deviation is calculated to generate the dynamic stability evaluation result of the rear wheel active steering system. By combining the dynamic following accuracy evaluation results and the dynamic stability evaluation results, a comprehensive performance score for the vehicle's rear-wheel active steering system is generated.

[0007] Preferably, the step of aligning and calculating the timing difference between the control command timing of the rear wheel steering actuator and the feedback timing of the rear wheel steering angle sensor to generate a steering angle command following error sequence includes: The timing of the control commands for the rear wheel steering actuator is remapped at sampling points to make its sampling frequency consistent with the sampling frequency of the feedback timing of the rear wheel steering angle sensor. The dynamic time warping algorithm is invoked to perform non-linear time axis alignment processing on the remapped control command timing and the feedback timing of the rear wheel steering angle sensor to eliminate the phase delay between the two. Calculate the absolute difference between the command angle value of each sampling point in the aligned control command timing and the actual angle value of the corresponding sampling point in the feedback timing of the rear wheel angle sensor; The absolute difference sequence is subjected to low-pass filtering to remove high-frequency noise interference, generating a smoothed corner command following error sequence.

[0008] Preferably, the step of performing error distribution statistical processing based on the steering command following error sequence and a preset dynamic following accuracy threshold to generate the dynamic following accuracy evaluation result of the rear wheel active steering system includes: The cornering command following error sequence is divided into multiple consecutive error analysis time windows; Within each error analysis time window, the root mean square error value, the maximum peak error value, and the proportion of duration during which the error exceeds the preset dynamic tracking accuracy threshold are calculated for the corner command following error sequence. Based on the root mean square error value, the maximum peak error value, and the duration ratio, combined with the preset graded scoring rules, the local following accuracy score for each error analysis time window is calculated. The local following accuracy scores for all error analysis time windows are processed by time-weighted averaging to generate the dynamic following accuracy evaluation results. The weight of the time-weighted averaging is positively correlated with the severity of the test conditions corresponding to the error analysis time window.

[0009] Preferably, the step of jointly analyzing the vehicle yaw rate time series and the vehicle body lateral acceleration time series to generate a vehicle lateral dynamic response feature vector includes: The yaw rate time series and the lateral acceleration time series of the vehicle body are subjected to frequency domain transformation processing to obtain the yaw rate spectrum characteristics and the lateral acceleration spectrum characteristics, respectively. Energy distribution features within a preset frequency band are extracted from the yaw rate spectral features, and energy distribution features within the same preset frequency band are extracted from the lateral acceleration spectral features. Calculate the coherence function values ​​of the yaw rate spectral characteristics and the lateral acceleration spectral characteristics within the preset frequency band to obtain the frequency domain coherence characteristics; The energy distribution features of the yaw rate spectrum, the energy distribution features of the lateral acceleration spectrum, and the frequency domain coherence features are vectorized and concatenated to generate the vehicle's lateral dynamic response feature vector.

[0010] Preferably, the step of calling a preset stability benchmark model to perform expected response matching processing on the vehicle's lateral dynamic response feature vector to generate yaw and lateral dynamic response error vectors includes: The vehicle's lateral dynamic response feature vector is input into the feature parsing layer of the stability benchmark model, which is constructed based on an ideal linear two-degree-of-freedom vehicle model. The feature parsing layer maps the vehicle's lateral dynamic response feature vector into an actual yaw rate gain vector and an actual lateral acceleration gain vector. The desired response calculation layer of the stability benchmark model is invoked to calculate the desired yaw rate gain vector and the desired lateral acceleration gain vector based on the steering wheel angle input and vehicle speed under the current test conditions. Calculate the difference between the actual yaw rate gain vector and the desired yaw rate gain vector, and the difference between the actual lateral acceleration gain vector and the desired lateral acceleration gain vector, respectively. The two difference sequences are combined to form the yaw and lateral dynamic response error vector.

[0011] Preferably, the step of calculating the stability deviation based on the yaw and lateral dynamic response error vectors and a preset stability error tolerance range to generate the dynamic stability evaluation result of the rear-wheel active steering system includes: Principal component analysis was performed on the yaw and lateral dynamic response error vectors to extract the principal component eigenvectors characterizing the main dynamic deviation modes and their corresponding variance contribution rates. The Mahalanobis distance between the yaw and lateral dynamic response error vectors in the high-dimensional space formed by the preset stability error tolerance interval boundary is calculated based on the principal component eigenvectors. The Mahalanobis distance is normalized to obtain the original stability deviation score; The original stability deviation score is weighted and corrected based on the variance contribution rate to generate a corrected stability deviation score. The corrected stability deviation score is mapped to a preset stability level range to generate the dynamic stability evaluation result.

[0012] Preferably, the step of integrating the dynamic following accuracy evaluation results and the dynamic stability evaluation results to generate a comprehensive performance score for the vehicle's rear-wheel active steering system includes: The dynamic tracking accuracy evaluation results and the dynamic stability evaluation results are quantified into tracking accuracy values ​​and stability values, respectively. Based on the type of the current test condition, the corresponding following accuracy weight coefficient and stability weight coefficient are queried from the preset weight configuration database. Under low-speed steering conditions, the following accuracy weight coefficient is higher than the stability weight coefficient, and under high-speed steering conditions, the stability weight coefficient is higher than the following accuracy weight coefficient. Based on the retrieved following accuracy weight coefficient and stability weight coefficient, the following accuracy value and the stability value are linearly weighted and summed to generate an initial comprehensive performance score; A preset performance scoring calibration model is invoked, and the initial comprehensive performance score is calibrated based on historical evaluation data to ensure that the comprehensive performance score conforms to the preset historical score distribution pattern.

[0013] Preferably, the step of calling a preset performance scoring calibration model and performing distribution calibration processing on the initial comprehensive performance score based on historical evaluation data includes: Retrieve from the historical evaluation database the initial comprehensive performance score set of historical systems that are the same as or similar to the current vehicle's rear-wheel active steering system model; Calculate the mean and standard deviation of the initial comprehensive performance score set; Based on the mean and standard deviation, the initial comprehensive performance score is standardized using Z-scores to obtain a standardized score. The standardized score is linearly mapped to the preset final output range of the comprehensive performance score to obtain the comprehensive performance score.

[0014] Preferably, before acquiring the set of operating data of the rear-wheel active steering system of the vehicle to be evaluated under preset test conditions, the following steps are included: Establish a standardized test condition protocol covering low-speed steering, medium-speed lane changing, and high-speed steady-state turning; According to the standardized test condition protocol, deploy a data acquisition system in a real vehicle test track or a high-fidelity dynamics simulation platform; The data acquisition system synchronously acquires the control command timing of the rear wheel steering actuator, the feedback timing of the rear wheel steering angle sensor, the vehicle yaw rate timing, and the vehicle body lateral acceleration timing. The collected raw time-series data undergoes preprocessing operations such as timestamp synchronization correction, outlier removal, and missing value imputation to form the running data set.

[0015] Preferably, when the processor executes the computer program, it implements the steps of the performance evaluation method for a rear-wheel active steering system of an automobile as described in any of the above-mentioned methods.

[0016] Compared with the prior art, the beneficial effects of the present invention are: By aligning the timing of the rear wheel steering actuator control commands with the timing of the angle sensor feedback, the comparison distortion problem caused by asynchronous signal transmission and sampling was resolved. Based on this, a calculated angle command following error sequence fully records all instantaneous errors of the system throughout the entire dynamic testing process. Furthermore, by statistically analyzing the error distribution of this continuous error sequence against a preset dynamic following accuracy threshold, the overall following accuracy of the system under varying commands can be accurately quantified. This represents a shift in the evaluation of the system's dynamic response accuracy from a single-point, static assessment to a full-process, statistically quantitative assessment, and the evaluation results more accurately reflect the system's tracking capability and reliability under complex real-world operating conditions.

[0017] By jointly analyzing time-series data of vehicle yaw rate and lateral acceleration, a comprehensive lateral dynamic response feature vector is generated, providing a more complete description of the vehicle's coupled motion state under system intervention. A pre-defined stability benchmark model is used to match this feature vector and generate the desired ideal response, thus obtaining yaw and lateral dynamic response error vectors. These error vectors reveal the multi-dimensional deviations between the actual and ideal responses. Combined with a pre-defined stability error tolerance range, the stability deviation is calculated. This deviation is a comprehensive quantitative indicator that determines whether the system has maintained basic vehicle stability and accurately assesses its ability to keep the vehicle's lateral dynamics near the ideal response. This elevates stability evaluation from a simple "whether it exceeds the limit" judgment to a continuous quantitative evaluation of the dynamic response quality's "closeness to the ideal state." Attached Figure Description

[0018] Figure 1 This is a schematic diagram illustrating the working principle of the performance evaluation method for the rear-wheel active steering system of an automobile as described in this invention. Figure 2 A flowchart for generating the cornering command follow-up error sequence; Figure 3 A flowchart for generating the lateral dynamic response feature vector of a vehicle; Figure 4 A radar chart showing the performance of a car's rear-wheel active steering system; Figure 5 Analysis diagram of contribution to the performance of the rear-wheel active steering system of automobiles under multiple operating conditions. Detailed Implementation

[0019] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0020] Please see Figure 1 This invention provides a performance evaluation method for a rear-wheel active steering system of an automobile. The method includes: acquiring a set of operating data of the rear-wheel active steering system of the automobile under preset test conditions, the set of operating data including the control command timing sequence of the rear-wheel steering actuator, the feedback timing sequence of the rear-wheel steering angle sensor, the vehicle yaw rate timing sequence, and the vehicle body lateral acceleration timing sequence; performing timing alignment and difference calculation processing on the control command timing sequence of the rear-wheel steering actuator and the feedback timing sequence of the rear-wheel steering angle sensor to generate a steering angle command following error sequence; and performing error distribution statistical processing based on the steering angle command following error sequence and a preset dynamic following accuracy threshold to generate a rear-wheel active steering system performance evaluation method. The system's dynamic following accuracy evaluation results are used as follows: The vehicle's yaw rate time series and lateral acceleration time series are jointly analyzed to generate a vehicle lateral dynamic response feature vector; a preset stability benchmark model is used to perform expected response matching processing on the vehicle's lateral dynamic response feature vector, generating yaw and lateral dynamic response error vectors; based on the yaw and lateral dynamic response error vectors and a preset stability error tolerance range, stability deviation is calculated to generate the dynamic stability evaluation results of the rear-wheel active steering system; the dynamic following accuracy evaluation results and the dynamic stability evaluation results are integrated to generate a comprehensive performance score for the vehicle's rear-wheel active steering system.

[0021] In one embodiment of the present invention, see [reference] Figure 2The control command timing of the rear wheel steering actuator is sampled and remapped to match the sampling frequency of the rear wheel angle sensor feedback timing. The dynamic time warping algorithm is then used to perform nonlinear time axis alignment between the remapped control command timing and the rear wheel angle sensor feedback timing to eliminate the phase delay between them. The absolute difference between the command angle value of each sample point in the aligned control command timing and the actual angle value of the corresponding sample point in the rear wheel angle sensor feedback timing is calculated. The absolute difference sequence is then low-pass filtered to remove high-frequency noise interference, generating a smoothed angle command following error sequence. The cornering command following error sequence is divided into multiple consecutive error analysis time windows. Within each error analysis time window, the root mean square error (RMSE), peak error (MPA), and the duration of errors exceeding a preset dynamic following accuracy threshold are calculated. Based on the RMS, MPA, and duration, and combined with preset grading and scoring rules, a local following accuracy score for each error analysis time window is calculated. A time-weighted average is applied to all local following accuracy scores for all error analysis time windows to generate a dynamic following accuracy evaluation result. The weight of the time-weighted average is positively correlated with the severity of the test conditions corresponding to the error analysis time window.

[0022] In practical implementation, the performance evaluation of a vehicle's rear-wheel active steering system involves the quantitative analysis of dynamic following accuracy. The implementation process uses the control command timing of the rear-wheel steering actuator and the feedback timing of the rear-wheel steering angle sensor as the basic input data. For example, in a low-speed steering test scenario, the control command timing is recorded at a frequency of 100 Hz, and the rear-wheel steering angle sensor feedback timing is recorded at a frequency of 50 Hz. Through sampling point remapping processing, the sampling frequency of the control command timing is adjusted to 50 Hz, making the sampling frequencies of the two consistent. A dynamic time warping algorithm is then used to perform non-linear time axis alignment processing on the remapped control command timing and the rear-wheel steering angle sensor feedback timing to eliminate phase delays caused by system latency or communication lag. After alignment, the absolute difference between the command angle value at each sampling point in the control command timing and the actual angle value at the corresponding sampling point in the rear-wheel steering angle sensor feedback timing is calculated point by point, generating an original absolute difference sequence. The original absolute difference sequence was subjected to low-pass filtering to remove high-frequency noise interference. The cutoff frequency of the low-pass filter was set to 10 Hz to generate a smoothed corner command following error sequence. In the data comparison, the peak fluctuation of the absolute difference sequence before filtering was 0.5 degrees, and the peak fluctuation of the corner command following error sequence after filtering was reduced to 0.1 degrees.

[0023] In some embodiments, the cornering command following error sequence is divided into multiple consecutive error analysis time windows, each lasting 1 second and covering the entire duration of the test condition. Within each error analysis time window, the root mean square error (RMSE), peak value, and duration exceeding a preset dynamic following accuracy threshold are calculated. The dynamic following accuracy threshold is set to 0.2 degrees. Based on the RMS error, peak value, and duration proportion, and combined with a preset grading scoring rule, a local following accuracy score for each error analysis time window is calculated. The grading scoring rule stipulates that a score of 100 is awarded when the RMS error is less than 0.1 degrees, the peak value is less than 0.2 degrees, and the duration proportion is less than 5%. The formula for calculating the RMS error is expressed as follows: in: This represents the root mean square error value within the error analysis time window. This represents the total number of sampling points within the error analysis time window. This indicates the first error in the turn command follow-up sequence. Error value of each sampling point.

[0024] Optionally, a time-weighted average is applied to the local following accuracy scores for all error analysis time windows to generate a dynamic following accuracy evaluation result. The weight of the time-weighted average is positively correlated with the severity of the test conditions corresponding to the error analysis time windows, and the severity is determined based on a combination of the steering angle change rate and vehicle speed. The weight of the error analysis time window is set to 1.0 in low-speed steering conditions, 1.5 in medium-speed lane-changing conditions, and 2.0 in high-speed steady-state turning conditions. Data comparison shows that the local following accuracy score is 95 points in low-speed steering conditions, 90 points in medium-speed lane-changing conditions, and 85 points in high-speed steady-state turning conditions. After time-weighted averaging, the dynamic following accuracy evaluation result is 88.75 points.

[0025] It is understandable that the duration proportion of errors exceeding the preset dynamic following accuracy threshold is obtained by comparing the number of sampling points with error values ​​greater than 0.2 degrees in the cornering command following error sequence with the total number of sampling points within the error analysis time window. In specific implementations, the maximum peak error value is directly extracted from the maximum absolute value within each error analysis time window of the cornering command following error sequence. The graded scoring rule maps the root mean square error value, the maximum peak error value, and the duration proportion to scores between 0 and 100, using linear interpolation. In some embodiments, low-pass filtering is implemented using a Butterworth filter, the distance metric for the dynamic time warping algorithm uses Euclidean distance, and the control command timing sequence after nonlinear alignment of the time axis achieves point-to-point matching with the rear wheel cornering sensor feedback timing sequence in the time dimension. It is understandable that the sampling point remapping process converts high-frequency control command timing data into a sampling point sequence consistent with the low-frequency rear wheel cornering sensor feedback timing sequence using linear interpolation, ensuring data dimension alignment.

[0026] In one embodiment of the present invention, see [reference] Figure 3 The vehicle yaw rate time series and the vehicle lateral acceleration time series are processed by frequency domain transformation to obtain yaw rate spectrum features and lateral acceleration spectrum features, respectively. Energy distribution features within a preset frequency band are extracted from the yaw rate spectrum features, and energy distribution features within the same preset frequency band are extracted from the lateral acceleration spectrum features. The coherence function values ​​of the yaw rate spectrum features and the lateral acceleration spectrum features within the preset frequency band are calculated to obtain frequency domain coherence features. The energy distribution features of the yaw rate spectrum features, the energy distribution features of the lateral acceleration spectrum features, and the frequency domain coherence features are vectorized and concatenated to generate the vehicle lateral dynamic response feature vector.

[0027] In practical implementation, the generation of the vehicle's lateral dynamic response feature vector is based on the joint analysis and processing of the vehicle yaw rate time series and the vehicle body lateral acceleration time series. Taking the medium-speed lane change test scenario as an example, the sampling frequency of both the vehicle yaw rate time series and the vehicle body lateral acceleration time series is 100 Hz, and the time series data length for a single test condition lasts for 10 seconds. Frequency domain transformation processing is performed on the vehicle yaw rate time series and the vehicle body lateral acceleration time series respectively. The frequency domain transformation processing uses the Fast Fourier Transform algorithm to convert the 1000-point time domain signal into spectral data, obtaining the yaw rate spectral features and the lateral acceleration spectral features respectively. Energy distribution features within a preset frequency band are extracted from the yaw rate spectral features. The preset frequency band range is set to 0.1 Hz to 2 Hz, which covers the main response frequencies of the vehicle's lateral dynamics. The energy distribution features are obtained by calculating the sum of the squares of the spectral amplitudes within this frequency band. Energy distribution features within the same preset frequency band were extracted from the lateral acceleration spectrum features. The extraction method was to calculate the sum of squares of the spectral amplitudes within the 0.1 Hz to 2 Hz frequency band. Data comparison showed that the energy value of the yaw rate spectrum feature within the preset frequency band was 150 (deg / s)², and the energy value of the lateral acceleration spectrum feature within the same preset frequency band was 0.8 (m / s²)².

[0028] In some embodiments, the coherence function values ​​of the yaw rate spectral characteristics and the lateral acceleration spectral characteristics within a preset frequency band are calculated to obtain frequency domain coherence characteristics. The coherence function values ​​characterize the degree of linear correlation between the two signals in the frequency domain. The formula for calculating the coherence function values ​​is expressed as follows: in: Indicates frequency The coherence function value at that point, The cross-power spectral density represents the time series of vehicle yaw rate and vehicle lateral acceleration. The self-power spectral density represents the time series of vehicle yaw rate. This represents the self-power spectral density of the vehicle body's lateral acceleration time sequence. In specific implementation, the coherent function values ​​within a preset frequency band of 0.1 Hz to 2 Hz are arithmetically averaged to generate a scalar form of frequency domain coherent feature. Data comparison shows that the average value of the frequency domain coherent feature under direct steering input is 0.92, while the average value of the frequency domain coherent feature under road interference is 0.75.

[0029] Optionally, the energy distribution features of the yaw rate spectrum, the energy distribution features of the lateral acceleration spectrum, and the frequency domain coherence features are vectorized and concatenated to generate a vehicle lateral dynamic response feature vector. The energy distribution features are represented in scalar numerical form, and the frequency domain coherence features are also represented in scalar numerical form. The vectorization and concatenation process combines these three scalars into a three-dimensional real vector in a fixed order. For example, a specific vehicle lateral dynamic response feature vector is represented as [150, 0.8, 0.92], where the first dimension element is the energy value of the yaw rate spectrum, the second dimension element is the energy value of the lateral acceleration spectrum, and the third dimension element is the average value of the frequency domain coherence features.

[0030] It is understood that the preset frequency band range can be adjusted according to the vehicle type and testing purpose. For tests emphasizing low-frequency response characteristics, the preset frequency band range can be set to 0.1 Hz to 1 Hz. In some embodiments, the extraction of energy distribution characteristics is not limited to calculating the total energy, but can also extract the energy proportion of a specified sub-band within the preset frequency band as a feature. Before frequency domain transformation processing, a Hanning window function is applied to the vehicle yaw rate time series and the vehicle body lateral acceleration time series to reduce spectral leakage.

[0031] In one embodiment of the present invention, the lateral dynamic response feature vector of the vehicle is input into the feature parsing layer of the stability benchmark model. The stability benchmark model is constructed based on an ideal linear two-degree-of-freedom vehicle model. The feature parsing layer maps the lateral dynamic response feature vector of the vehicle into the actual yaw rate gain vector and the actual lateral acceleration gain vector. The expected response calculation layer of the stability benchmark model is called. Based on the steering wheel angle input and vehicle speed under the current test conditions, the expected yaw rate gain vector and the expected lateral acceleration gain vector are calculated. The difference between the actual yaw rate gain vector and the expected yaw rate gain vector, and the difference between the actual lateral acceleration gain vector and the expected lateral acceleration gain vector are calculated respectively. The two difference sequences are combined to form the yaw and lateral dynamic response error vectors. Principal component analysis is performed on the yaw and lateral dynamic response error vectors to extract the principal component eigenvectors representing the main dynamic deviation modes and their corresponding variance contribution rates. Based on the principal component eigenvectors, the Mahalanobis distance between the yaw and lateral dynamic response error vectors in the high-dimensional space formed by the boundaries of the preset stability error tolerance interval is calculated. The Mahalanobis distance is normalized to obtain the original stability deviation score. The original stability deviation score is then weighted and corrected by combining the variance contribution rate to generate the corrected stability deviation score. The corrected stability deviation score is then mapped to the preset stability level interval to generate the dynamic stability evaluation result.

[0032] In practice, a pre-defined stability benchmark model is invoked to perform expected response matching processing on the vehicle's lateral dynamic response feature vector to generate yaw and lateral dynamic response error vectors. The stability benchmark model is constructed based on an ideal linear two-degree-of-freedom vehicle model, and its input is the vehicle's lateral dynamic response feature vector characterizing the vehicle's dynamics. The vehicle's lateral dynamic response feature vector is input into the feature parsing layer of the stability benchmark model. A specific vehicle lateral dynamic response feature vector is [150, 0.8, 0.92]. Through a pre-defined fully connected neural network in the feature parsing layer, the vehicle's lateral dynamic response feature vector is mapped into an actual yaw rate gain vector and an actual lateral acceleration gain vector. The actual yaw rate gain vector is a sequence containing gain values ​​at 10 frequency points, and the actual lateral acceleration gain vector is a corresponding sequence containing gain values ​​at 10 frequency points. The expected response calculation layer of the stability benchmark model is invoked. The expected response calculation layer has a built-in transfer function of the ideal linear two-degree-of-freedom vehicle model. Based on the steering wheel angle input timing and vehicle speed parameters under the current test conditions, the expected yaw rate gain vector and expected lateral acceleration gain vector corresponding to the same set of frequency points are calculated. The differences between the actual yaw rate gain vector and the desired yaw rate gain vector, as well as the differences between the actual lateral acceleration gain vector and the desired lateral acceleration gain vector, were calculated separately. Each difference sequence contained 10 error values. The two difference sequences were combined sequentially to form a 20-dimensional yaw and lateral dynamic response error vector. Data comparison showed that under ideal matching conditions, the values ​​of each element of the error vector were close to zero, while when there was a deviation in the system, the elements of the error vector would show significantly non-zero values.

[0033] In some embodiments, stability deviation is calculated based on the yaw and lateral dynamic response error vectors and a preset stability error tolerance interval to generate dynamic stability evaluation results. Principal component analysis is performed on the 20-dimensional yaw and lateral dynamic response error vectors. Principal component analysis extracts the principal component eigenvectors representing the main dynamic deviation modes and their corresponding variance contribution rates; for example, the variance contribution rates of the first three principal components are 65%, 20%, and 10%, respectively. The Mahalanobis distance of the yaw and lateral dynamic response error vectors in the high-dimensional space defined by the preset stability error tolerance interval is calculated based on the principal component eigenvectors. The stability error tolerance interval defines the allowed error boundary in each dimension. The formula for calculating the Mahalanobis distance is expressed as: in: Represents Mahalanobis distance, This represents the yaw and lateral dynamic response error vectors. This represents the median vector formed by the boundary values ​​of each dimension of the stability error tolerance interval. This represents the inverse matrix of the error vector covariance matrix obtained based on historical data statistics. This represents the transpose of the error vector covariance matrix obtained based on historical data statistics.

[0034] Optionally, the Mahalanobis distance is normalized to obtain the original stability deviation score. The normalization process uses the minimax scaling method to map the Mahalanobis distance to the interval of 0 to 100. The original stability deviation score is then weighted and corrected using the variance contribution rate, with the cumulative variance contribution rate of the principal components as the weighting factor, to generate the corrected stability deviation score. The corrected stability deviation score is then mapped to a preset stability level interval to generate a dynamic stability evaluation result. The stability level interval divides the score from 0 to 100 into four levels: excellent, good, satisfactory, and unsatisfactory.

[0035] In one embodiment of the present invention, the dynamic following accuracy evaluation result and the dynamic stability evaluation result are quantified into following accuracy value and stability value, respectively. Based on the type of the current test condition, the corresponding following accuracy weight coefficient and stability weight coefficient are queried from a preset weight configuration database. Under low-speed steering conditions, the following accuracy weight coefficient is higher than the stability weight coefficient, and under high-speed steering conditions, the stability weight coefficient is higher than the following accuracy weight coefficient. Based on the queried following accuracy weight coefficient and stability weight coefficient, the following accuracy value and stability value are linearly weighted and summed to generate an initial comprehensive performance score. A preset performance score calibration model is invoked, and the initial comprehensive performance score is distributed and calibrated based on historical evaluation data. The initial comprehensive performance score set of historical systems with the same or similar models as the rear-wheel active steering system of the vehicle under test is retrieved from the historical evaluation database. The mean and standard deviation of the initial comprehensive performance score set are calculated. Based on the mean and standard deviation, the initial comprehensive performance score is Z-score standardized to obtain a standardized score. The standardized score is linearly mapped to a preset final output interval for comprehensive performance scores to obtain the comprehensive performance score.

[0036] In practice, the dynamic following accuracy evaluation results and dynamic stability evaluation results are integrated to generate a comprehensive performance score for the vehicle's rear-wheel active steering system. The implementation process uses a complete evaluation encompassing multiple test conditions as an example. The dynamic following accuracy evaluation result is quantified as 88.75 points, and the dynamic stability evaluation result is quantified as 82.0 points. The dynamic following accuracy evaluation result and the dynamic stability evaluation result are quantified into following accuracy values ​​and stability values, respectively: following accuracy value is 88.75, and stability value is 82.0. Based on the type of the current test condition, the corresponding following accuracy weight coefficient and stability weight coefficient are retrieved from a preset weight configuration database. The weight configuration database stores the weight coefficients for different test condition types in tabular form. In low-speed steering conditions, the following accuracy weight coefficient is higher than the stability weight coefficient, and in high-speed steering conditions, the stability weight coefficient is higher than the following accuracy weight coefficient. This evaluation includes three sub-conditions: low-speed steering, medium-speed lane changing, and high-speed steering. Independent weight coefficients need to be retrieved for each sub-condition. See Table 1.

[0037] Table 1: Weighting Allocation Table In some embodiments, the following accuracy and stability values ​​are linearly weighted and summed based on the retrieved following accuracy weighting coefficients and stability weighting coefficients to generate an initial comprehensive performance score. For the low-speed steering sub-condition, the following accuracy value is 90.0 and the stability value is 85.0, calculated using weighting coefficients α=0.7 and β=0.3. For the medium-speed lane change sub-condition, the following accuracy value is 88.0 and the stability value is 80.0, calculated using weighting coefficients α=0.5 and β=0.5. For the high-speed steering sub-condition, the following accuracy value is 88.0 and the stability value is 81.0, calculated using weighting coefficients α=0.3 and β=0.7. The initial comprehensive performance score for the entire test cycle is obtained by arithmetically averaging the scores of each sub-condition, resulting in an initial comprehensive performance score of 86.4.

[0038] Optionally, a preset performance scoring calibration model is invoked to perform distribution calibration on the initial comprehensive performance score based on historical evaluation data. The purpose of the performance scoring calibration model is to ensure that the scoring results conform to the historical score distribution pattern. An initial comprehensive performance score set of historical systems with the same or similar models as the rear-wheel active steering system of the vehicle currently being evaluated is retrieved from the historical evaluation database, yielding 100 historical initial comprehensive performance score samples. The mean and standard deviation of the initial comprehensive performance score set are calculated, yielding a mean μ of 85.0 points and a standard deviation σ of 5.0 points.

[0039] It is understandable that the initial comprehensive performance score is based on the mean and standard deviation. The scores are standardized to obtain standardized scores. The formula for standardizing fractions is expressed as follows: in: Indicates standardized score, This indicates the initial overall performance score. This represents the mean of the initial set of historical comprehensive performance scores. This represents the standard deviation of the historical initial overall performance score set. Substituting the initial overall performance score of 86.4 into the formula, the standardized score is calculated. It is 0.28.

[0040] In practice, the standardized scores are linearly mapped to a preset final output range for the comprehensive performance score to obtain the comprehensive performance score. The final output range for the comprehensive performance score is set to 0 to 100. The linear mapping function is as follows: ,in: This represents the final overall performance score. The standardized score will be used to determine the final overall performance score. Substituting into the mapping function, the final overall performance score is calculated to be 52.8. Data comparison shows that the initial overall performance score without calibration is 86.4, while the overall performance score after distribution calibration is 52.8. The score result has been adjusted to a scale based on the historical distribution.

[0041] See Figure 4 This is a radar chart of the performance of a car's rear-wheel active steering system. Its core purpose is to display the system's following accuracy, stability, and overall score under different operating conditions from multiple dimensions. In automotive chassis system performance evaluation, this type of chart is used for multi-dimensional performance visualization and comparison, supporting the analytical process of "sub-item evaluation → comprehensive assessment → optimization direction." This chart is a comprehensive display tool for evaluating the performance of a car's rear-wheel active steering system, intuitively presenting the system's performance shortcomings under different operating conditions; comparing the correlation between sub-item performance and the overall score; and providing direction for system optimization.

[0042] In one embodiment of the present invention, a standardized test condition protocol covering low-speed steering, medium-speed lane changing, and high-speed steady-state turning is constructed. According to the standardized test condition protocol, a data acquisition system is deployed in a real vehicle test track or a high-fidelity dynamics simulation platform. The data acquisition system synchronously acquires the control command timing of the rear wheel steering actuator, the feedback timing of the rear wheel steering angle sensor, the timing of the vehicle yaw rate, and the timing of the vehicle body lateral acceleration. The acquired raw timing data is preprocessed by timestamp synchronization correction, outlier removal, and missing value interpolation to form a set of operating data.

[0043] In practical implementation, before obtaining the operational data set of the rear-wheel active steering system of the vehicle under test under preset test conditions, it is necessary to complete the construction of a standardized test environment and the collection and preprocessing of raw data. Constructing a standardized test condition protocol covering low-speed steering, medium-speed lane changing, and high-speed steady-state turning is the first step. The standardized test condition protocol clearly defines the parameters of each test scenario in document form. For example, the low-speed steering condition specifies that the vehicle performs sinusoidal steering at a speed of 20 km / h, with a steering wheel angle amplitude of 90 degrees and a frequency of 0.2 Hz; the medium-speed lane changing condition specifies that the vehicle performs double lane change operations at a speed of 80 km / h; and the high-speed steady-state turning condition specifies that the vehicle performs circular driving with a fixed radius at a speed of 120 km / h. Data comparison shows that the low-speed steering condition mainly stimulates the low-speed agility of the rear-wheel steering system, while the high-speed steady-state turning condition mainly assesses the high-speed stability of the system.

[0044] According to standardized test condition protocols, a data acquisition system is deployed in a real vehicle test track or a high-fidelity dynamics simulation platform. The data acquisition system deployed in the real vehicle test track includes a vehicle bus data logger, a high-precision inertial measurement unit, and a steering angle sensor acquisition module. The data acquisition system deployed in the high-fidelity dynamics simulation platform directly listens to and records all signals output by the simulation model through a software interface. The data acquisition system ensures that it can synchronously record the control commands of the rear wheel steering actuator, the feedback from the rear wheel steering angle sensor, the vehicle yaw rate, and the vehicle's lateral acceleration signals at a frequency of no less than 100 Hz. The data acquisition system synchronously acquires the timing of the control commands of the rear wheel steering actuator, the feedback timing of the rear wheel steering angle sensor, the timing of the vehicle yaw rate, and the timing of the vehicle's lateral acceleration. The synchronization mechanism relies on a unified global clock source, such as a GPS clock or a high-precision network time protocol, ensuring that the timestamp deviation of all acquisition channels is less than 1 millisecond.

[0045] In some embodiments, the acquired raw time-series data undergoes preprocessing operations including timestamp synchronization correction, outlier removal, and missing value imputation to form a running dataset. Timestamp synchronization correction aligns the sampling points of all data channels based on a unified global clock source; the correction formula is expressed as: in: Indicates the first The corrected timestamps for each sampling point Indicates the first The timestamp of the original record of each sampling point This represents the time offset compensation calculated through clock synchronization analysis. Outlier removal uses the Laida criterion; for any signal sequence, its mean is calculated. and standard deviation Values ​​not in the range The sampling points within the range were identified as outliers and removed.

[0046] Optionally, missing value interpolation addresses continuous data loss caused by outlier removal or transient signal loss, employing a linear interpolation method. For example, if two consecutive sampling points in the vehicle yaw rate timing sequence are found to have been removed due to outliers, linear interpolation is performed using the values ​​of adjacent valid sampling points to fill the gaps. After preprocessing, the four timing data streams are strictly aligned on the time axis, contain no invalid values, and have consistent data lengths, collectively forming the operational data set for subsequent performance evaluation. Data comparison shows that the original rear wheel steering angle sensor feedback timing sequence contained three obvious pulse-like outliers. After outlier removal and interpolation, the signal curve became smooth and continuous.

[0047] See Figure 5 This is a multi-condition performance contribution analysis chart for a car's rear-wheel active steering system. It primarily displays the overall performance score and the individual contributions of following accuracy and stability under different test conditions. This chart is a tool for analyzing the component contributions of a car's rear-wheel active steering system performance evaluation, intuitively identifying the dominant factors in performance under different conditions; pinpointing weaknesses in stability contribution; and providing direction for system optimization. In automotive chassis system performance evaluation, this type of chart is used to analyze the contribution ratio of individual performance components under multiple conditions, supporting the process of "performance breakdown → weakness identification → optimization direction."

[0048] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0049] 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 performance evaluation method for a vehicle's rear-wheel active steering system, characterized in that, The method includes: Obtain the set of operating data of the rear-wheel active steering system of the vehicle to be evaluated under preset test conditions. The set of operating data includes the control command timing of the rear-wheel steering actuator, the feedback timing of the rear-wheel steering angle sensor, the timing of the vehicle yaw rate, and the timing of the vehicle body lateral acceleration. The timing sequence of the control command of the rear wheel steering actuator and the timing sequence of the feedback of the rear wheel angle sensor are aligned and the difference is calculated to generate an angle command following error sequence. Based on the steering command following error sequence and the preset dynamic following accuracy threshold, the error distribution statistical processing is performed to generate the dynamic following accuracy evaluation result of the rear wheel active steering system; The vehicle yaw rate time series and the vehicle body lateral acceleration time series are jointly analyzed and processed to generate a vehicle lateral dynamic response feature vector. The vehicle's lateral dynamic response feature vector is matched with the desired response by calling a preset stability benchmark model to generate yaw and lateral dynamic response error vectors. Based on the yaw and lateral dynamic response error vectors and the preset stability error tolerance range, the stability deviation is calculated to generate the dynamic stability evaluation result of the rear wheel active steering system. By combining the dynamic following accuracy evaluation results and the dynamic stability evaluation results, a comprehensive performance score for the vehicle's rear-wheel active steering system is generated.

2. The performance evaluation method for a vehicle rear-wheel active steering system according to claim 1, characterized in that, The timing alignment and difference calculation of the control command timing for the rear wheel steering actuator and the feedback timing of the rear wheel steering angle sensor are performed to generate a steering angle command following error sequence, including: The timing of the control commands for the rear wheel steering actuator is remapped at sampling points to make its sampling frequency consistent with the sampling frequency of the feedback timing of the rear wheel steering angle sensor. The dynamic time warping algorithm is invoked to perform non-linear time axis alignment processing on the remapped control command timing and the feedback timing of the rear wheel steering angle sensor to eliminate the phase delay between the two. Calculate the absolute difference between the command angle value of each sampling point in the aligned control command timing and the actual angle value of the corresponding sampling point in the feedback timing of the rear wheel angle sensor; The absolute difference sequence is subjected to low-pass filtering to remove high-frequency noise interference, generating a smoothed corner command following error sequence.

3. The performance evaluation method for a vehicle rear-wheel active steering system according to claim 1, characterized in that, The step of performing error distribution statistical processing based on the steering command following error sequence and a preset dynamic following accuracy threshold to generate the dynamic following accuracy evaluation result of the rear wheel active steering system includes: The cornering command following error sequence is divided into multiple consecutive error analysis time windows; Within each error analysis time window, the root mean square error value, the maximum peak error value, and the proportion of duration during which the error exceeds the preset dynamic tracking accuracy threshold are calculated for the corner command following error sequence. Based on the root mean square error value, the maximum peak error value, and the duration ratio, combined with the preset graded scoring rules, the local following accuracy score for each error analysis time window is calculated. The local following accuracy scores for all error analysis time windows are processed by time-weighted averaging to generate the dynamic following accuracy evaluation results. The weight of the time-weighted averaging is positively correlated with the severity of the test conditions corresponding to the error analysis time window.

4. The performance evaluation method for a vehicle rear-wheel active steering system according to claim 1, characterized in that, The step of jointly analyzing the vehicle yaw rate time series and the vehicle body lateral acceleration time series to generate a vehicle lateral dynamic response feature vector includes: The yaw rate time series and the lateral acceleration time series of the vehicle body are subjected to frequency domain transformation processing to obtain the yaw rate spectrum characteristics and the lateral acceleration spectrum characteristics, respectively. Energy distribution features within a preset frequency band are extracted from the yaw rate spectral features, and energy distribution features within the same preset frequency band are extracted from the lateral acceleration spectral features. Calculate the coherence function values ​​of the yaw rate spectral characteristics and the lateral acceleration spectral characteristics within the preset frequency band to obtain the frequency domain coherence characteristics; The energy distribution features of the yaw rate spectrum, the energy distribution features of the lateral acceleration spectrum, and the frequency domain coherence features are vectorized and concatenated to generate the vehicle's lateral dynamic response feature vector.

5. The performance evaluation method for a vehicle rear-wheel active steering system according to claim 1, characterized in that, The process of calling a preset stability benchmark model to perform expected response matching processing on the vehicle's lateral dynamic response feature vector to generate yaw and lateral dynamic response error vectors includes: The vehicle's lateral dynamic response feature vector is input into the feature parsing layer of the stability benchmark model, which is constructed based on an ideal linear two-degree-of-freedom vehicle model. The feature parsing layer maps the vehicle's lateral dynamic response feature vector into an actual yaw rate gain vector and an actual lateral acceleration gain vector. The desired response calculation layer of the stability benchmark model is invoked to calculate the desired yaw rate gain vector and the desired lateral acceleration gain vector based on the steering wheel angle input and vehicle speed under the current test conditions. Calculate the difference between the actual yaw rate gain vector and the desired yaw rate gain vector, and the difference between the actual lateral acceleration gain vector and the desired lateral acceleration gain vector, respectively. The two difference sequences are combined to form the yaw and lateral dynamic response error vector.

6. The performance evaluation method for a vehicle rear-wheel active steering system according to claim 1, characterized in that, The process of calculating the stability deviation based on the yaw and lateral dynamic response error vectors and a preset stability error tolerance range generates the dynamic stability evaluation results of the rear-wheel active steering system, including: Principal component analysis was performed on the yaw and lateral dynamic response error vectors to extract the principal component eigenvectors characterizing the main dynamic deviation modes and their corresponding variance contribution rates. The Mahalanobis distance between the yaw and lateral dynamic response error vectors in the high-dimensional space formed by the preset stability error tolerance interval boundary is calculated based on the principal component eigenvectors. The Mahalanobis distance is normalized to obtain the original stability deviation score; The original stability deviation score is weighted and corrected based on the variance contribution rate to generate a corrected stability deviation score. The corrected stability deviation score is mapped to a preset stability level range to generate the dynamic stability evaluation result.

7. The performance evaluation method for a vehicle rear-wheel active steering system according to claim 1, characterized in that, The process of integrating the dynamic following accuracy evaluation results and the dynamic stability evaluation results to generate a comprehensive performance score for the vehicle's rear-wheel active steering system includes: The dynamic tracking accuracy evaluation results and the dynamic stability evaluation results are quantified into tracking accuracy values ​​and stability values, respectively. Based on the type of the current test condition, the corresponding following accuracy weight coefficient and stability weight coefficient are queried from the preset weight configuration database. Under low-speed steering conditions, the following accuracy weight coefficient is higher than the stability weight coefficient, and under high-speed steering conditions, the stability weight coefficient is higher than the following accuracy weight coefficient. Based on the retrieved following accuracy weight coefficient and stability weight coefficient, the following accuracy value and the stability value are linearly weighted and summed to generate an initial comprehensive performance score; A preset performance scoring calibration model is invoked, and the initial comprehensive performance score is calibrated based on historical evaluation data to ensure that the comprehensive performance score conforms to the preset historical score distribution pattern.

8. The performance evaluation method for a vehicle rear-wheel active steering system according to claim 7, characterized in that, The step of calling a preset performance scoring calibration model and performing distribution calibration processing on the initial comprehensive performance score based on historical evaluation data includes: Retrieve from the historical evaluation database the initial comprehensive performance score set of historical systems that are the same as or similar to the current vehicle's rear-wheel active steering system model; Calculate the mean and standard deviation of the initial comprehensive performance score set; Based on the mean and standard deviation, the initial comprehensive performance score is standardized using Z-scores to obtain a standardized score. The standardized score is linearly mapped to the preset final output range of the comprehensive performance score to obtain the comprehensive performance score.

9. The performance evaluation method for a vehicle rear-wheel active steering system according to claim 1, characterized in that, Before acquiring the set of operational data of the rear-wheel active steering system of the vehicle under test under preset test conditions, the following steps are included: Establish a standardized test condition protocol covering low-speed steering, medium-speed lane changing, and high-speed steady-state turning; According to the standardized test condition protocol, deploy a data acquisition system in a real vehicle test track or a high-fidelity dynamics simulation platform; The data acquisition system synchronously acquires the control command timing of the rear wheel steering actuator, the feedback timing of the rear wheel steering angle sensor, the vehicle yaw rate timing, and the vehicle body lateral acceleration timing. The collected raw time-series data undergoes preprocessing operations such as timestamp synchronization correction, outlier removal, and missing value imputation to form the running data set.

10. A performance evaluation system for a vehicle's rear-wheel active steering system, comprising a memory, a processor, and a computer program stored in the memory and running on the processor, characterized in that, When the processor executes the computer program, it implements the steps of the performance evaluation method for a vehicle rear-wheel active steering system as described in any one of claims 1 to 9.

Images