A road unevenness recognition method, a multi-axle vehicle, a medium, and a product

Abstract

This application discloses a road surface roughness identification method, a multi-axle vehicle, a medium, and a product, relating to the field of vehicle driving technology. The method includes determining the road surface grade pre-classification result for the current period based on the acquired vertical acceleration time-series data set for the current period; combining the road surface grade pre-classification result of the previous period to determine the forgetting factor and process noise covariance matrix of the unknown input Kalman filter algorithm for the current period, thereby determining the Kalman gain for the current period; combining the system's discrete state-space equation and the unknown input Kalman filter algorithm to determine the road surface excitation signal for the current period; when the current period is greater than a preset value, calculating the road surface roughness power spectral density for the current period based on the road surface excitation signals of each period within a preset time window, and determining the road surface grade for the current period based on the road surface roughness classification criteria. This application can improve the response sensitivity and accuracy of multi-axle vehicles in identifying road surface grades.

Description

Technical Field

[0001] This application relates to the field of vehicle driving technology, and in particular to a method for identifying road surface unevenness, multi-axle vehicles, media, and products. Background Technology

[0002] Road surface roughness is a key external excitation source in vehicle driving systems, significantly affecting the vehicle's vertical vibration characteristics and dynamic loads on the vehicle body structure. Obtaining real-time, accurate road surface roughness information is crucial for active suspension adjustment, optimizing chassis control strategies, and improving driving performance.

[0003] Traditional methods for measuring road surface unevenness primarily rely on specialized measuring equipment such as laser profilers. However, these precision instruments are mainly used in road maintenance and inspection, and pavement quality acceptance technologies. Their identification process typically requires the test vehicle to maintain a low-speed and stable driving state to ensure the continuity and effectiveness of data acquisition. Offline measurement methods are not only expensive in terms of equipment cost and maintenance, but also cannot meet the real-time environmental information perception needs of heavy vehicles under complex speed change conditions, greatly limiting their feasibility for large-scale online applications in engineering vehicles.

[0004] In contrast, contact identification algorithms can be based on onboard sensors, offering lower costs and reflecting the true coupling state between the vehicle and the road surface. However, existing contact identification algorithms typically rely on the assumption of a stationary stochastic process, using a fixed, pre-defined covariance matrix of system noise and measurement process noise for state estimation. For conditions with varying road surface roughness levels, state estimation exhibits a lag, affecting the algorithm's convergence stability. Summary of the Invention

[0005] The purpose of this application is to provide a method for identifying road surface roughness, a multi-axle vehicle, a medium, and a product, which can improve the response sensitivity and accuracy of multi-axle vehicles in identifying road surface grades.

[0006] To achieve the above objectives, this application provides the following solution: Firstly, this application provides a method for identifying road surface roughness, including: Obtain the vertical acceleration time series data set for the current period; each vertical acceleration time series data in the vertical acceleration time series data set corresponds to a sampling point, and the sampling point is the center of the bottom of the multi-axle vehicle or the center of any wheel in the multi-axle vehicle.

[0007] Based on the current period's vertical acceleration time series data set, the road surface grade is pre-classified to determine the pre-classification result of the road surface grade for the current period.

[0008] Based on the road surface grade pre-classification results of the current period and the road surface grade pre-classification results of the previous period, the forgetting factor and process noise covariance matrix of the unknown input Kalman filter algorithm for the current period are determined, and the Kalman gain for the current period is determined based on the forgetting factor and process noise covariance matrix of the unknown input Kalman filter algorithm for the current period.

[0009] Based on the system's discrete state-space equation, the unknown input Kalman filter algorithm, the current period's vertical acceleration time-series data set, and the Kalman gain, the road excitation signal for the current period is determined. The system's discrete state-space equation is obtained by discretizing the state-space equation of the multi-axle vehicle, and the state-space equation is established based on the vehicle dynamics model of the multi-axle vehicle.

[0010] When the current period is greater than a preset value, the road surface roughness power spectral density of the current period is calculated based on the road surface excitation signal of each period within the preset time window, and the road surface roughness classification standard is combined to determine the road surface grade of the current period; wherein, the preset time window is a time window of preset length with the current period as the last time period.

[0011] Secondly, this application provides a multi-axle vehicle, including: a body, a suspension system, a running gear system, a memory, a processor, and a computer program stored in the memory and capable of running on the processor, wherein the processor executes the computer program to implement the steps of the road surface roughness identification method described above.

[0012] Thirdly, this application provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the road surface roughness identification method described above.

[0013] Fourthly, this application provides a computer program product, including a computer program that, when executed by a processor, implements the steps of the road surface roughness identification method described above.

[0014] According to the specific embodiments provided in this application, this application has the following technical effects: This application provides a road surface roughness identification method, a multi-axle vehicle, a medium, and a product. First, the road surface grade is pre-classified based on the current period's vertical acceleration time-series data. Then, combined with the road surface grade pre-classification results from the previous period, the forgetting factor and process noise covariance matrix of the unknown input Kalman filter algorithm for the current period are determined (equivalent to prior knowledge for the unknown input Kalman filter algorithm). Since the road surface grade pre-classification results of the current and previous periods can quickly and roughly determine the changes in road surface roughness, they are used as prior information to determine the prior knowledge for the unknown input Kalman filter algorithm in the current period. Based on this, the Kalman gain of the unknown input Kalman filter algorithm is dynamically adjusted to adapt to varying road surface roughness conditions and improve the identification response sensitivity. Furthermore, when determining the road surface excitation signal for the current period, a state-space equation is first established for the vehicle dynamics model of the multi-axle vehicle and discretized. Then, the road surface excitation signal for the current period is determined in combination with the unknown input Kalman filter algorithm for the current period. This avoids introducing model bias due to simplified models when applied to multi-axle vehicles, improving the accuracy of road surface grade identification for multi-axle vehicles. Attached Figure Description

[0015] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0016] Figure 1 This is an application environment diagram of a road surface roughness identification method according to an embodiment of this application; Figure 2 A schematic flowchart illustrating a road surface roughness identification method according to an embodiment of this application; Figure 3 A simplified structural diagram of a multi-axle vehicle provided in an embodiment of this application; Figure 4 A flowchart illustrating a method for dynamically adjusting an unknown input Kalman filter algorithm and calculating a road excitation signal based on a road surface grade pre-classification result, provided in an embodiment of this application; Figure 5 This is a structural schematic diagram of a multi-axle vehicle provided in an embodiment of this application.

[0017] Attached image labels: 102 terminal, 104 server. Detailed Implementation

[0018] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0019] To make the above-mentioned objectives, features and advantages of this application more apparent and understandable, the application will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0020] The road surface unevenness identification method provided in this application embodiment can be applied to, for example, Figure 1 In the application environment shown, terminal 102 communicates with server 104 via a network. A data storage system can store the data that server 104 needs to process. The data storage system can be set up independently, integrated into server 104, or placed in the cloud or on another server. Terminal 102 can send the current period's vertical acceleration time-series data set to server 104. After receiving the current period's vertical acceleration time-series data set, server 104 performs pre-classification of road surface grade based on the current period's vertical acceleration time-series data set to determine the current period's road surface grade pre-classification result. Based on the current period's road surface grade pre-classification result and the previous period's road surface grade pre-classification result, server 104 determines the forgetting factor and process noise covariance matrix of the unknown input Kalman filter algorithm for the current period, and determines the Kalman gain for the current period based on the forgetting factor and process noise covariance matrix of the unknown input Kalman filter algorithm for the current period. Based on the system's discrete state-space equation, the unknown input Kalman filter algorithm, the current period's vertical acceleration time-series data set, and the Kalman gain, server 104 determines the road surface excitation signal for the current period. When the current period is greater than a preset value, server 104 calculates the road surface roughness power spectral density for the current period based on the road surface excitation signals of each period within a preset time window, and determines the road surface grade for the current period in conjunction with the road surface roughness classification standard. Server 104 can feed back the obtained road surface grade for the current period to terminal 102. Furthermore, in some embodiments, the road surface roughness identification method can also be implemented by either server 104 or terminal 102. For example, terminal 102 can directly perform road surface roughness identification processing on the current period's vertical acceleration time-series data set, or server 104 can obtain the current period's vertical acceleration time-series data set from the data storage system and perform road surface roughness identification processing on the current period's vertical acceleration time-series data set.

[0021] The terminal 102 can be, but is not limited to, various desktop computers, laptops, smartphones, tablets, IoT devices, multi-axle vehicles, and portable wearable devices. IoT devices can include smart speakers, smart TVs, smart air conditioners, and smart in-vehicle devices. Portable wearable devices can include smartwatches, smart bracelets, and head-mounted devices. The server 104 can be implemented using a standalone server or a server cluster composed of multiple servers, or it can be a cloud server.

[0022] In one exemplary embodiment, such as Figure 2 As shown, a method for identifying road surface unevenness is provided. This method is executed by a computer device, specifically by a terminal or server alone, or by both a terminal and a server. In this embodiment, the method is applied to... Figure 1 Taking server 104 as an example, the explanation includes the following steps 201 to 205. Wherein: Step 201: Obtain the vertical acceleration time series data set for the current period; each vertical acceleration time series data in the vertical acceleration time series data set corresponds to a sampling point, and the sampling point is the center of the bottom of the multi-axle vehicle body or the center of any wheel.

[0023] Step 202: Based on the current period's vertical acceleration time series data set, pre-classify the road surface grade to determine the pre-classification result of the road surface grade for the current period.

[0024] Step 203: Based on the road surface grade pre-classification results of the current period and the road surface grade pre-classification results of the previous period, determine the forgetting factor and process noise covariance matrix of the unknown input Kalman filter algorithm for the current period, and determine the Kalman gain for the current period based on the forgetting factor and process noise covariance matrix of the unknown input Kalman filter algorithm for the current period.

[0025] Step 204: Based on the system discrete state-space equation, the unknown input Kalman filter algorithm, the current period's vertical acceleration time-series data set, and the Kalman gain, determine the road excitation signal for the current period; the system discrete state-space equation is the equation obtained by discretizing the state-space equation of the multi-axle vehicle, and the state-space equation is established based on the vehicle dynamics model of the multi-axle vehicle.

[0026] Step 205: When the current period is greater than a preset value, calculate the road surface roughness power spectral density of the current period based on the road surface excitation signal of each period within the preset time window, and determine the road surface grade of the current period in combination with the road surface roughness classification standard; wherein, the preset time window is a time window with the current period as the last time period.

[0027] Steps 201 to 205 are implemented as follows: First, the road surface grade is pre-classified based on the current period's vertical acceleration time series data set. Then, combined with the road surface grade pre-classification results of the previous period, the forgetting factor and process noise covariance matrix of the unknown input Kalman filter algorithm for the current period are determined (equivalent to the prior knowledge of the unknown input Kalman filter algorithm). Since the road surface grade pre-classification results of the current period and the previous period can quickly and roughly determine the changes in road surface roughness, they are used as prior information to determine the prior knowledge of the unknown input Kalman filter algorithm for the current period. Based on this, the Kalman gain of the unknown input Kalman filter algorithm is dynamically adjusted to adapt to the changing road surface roughness conditions and improve the identification response sensitivity.

[0028] Furthermore, when determining the road excitation signal for the current period, the state-space equation of the vehicle dynamics model of the multi-axle vehicle is first established and discretized. Then, the road excitation signal is determined by combining the improved unknown input Kalman filter algorithm. This avoids the introduction of model bias due to the simplified model when applied to multi-axle vehicles and improves the identification accuracy of multi-axle vehicles under varying road surface roughness levels.

[0029] In another exemplary embodiment of this application, step 202 is replaced by the following steps 2021-2022: Step 2021: Based on the vertical acceleration time series data set of the current period, determine the classification features of each sampling point in the current period; the classification features include kurtosis factor, skewness factor, margin factor, peak factor and waveform factor.

[0030] As an example, based on the time series data set of vertical acceleration for the current period collected in step 201, dimensionless parameters are selected as classification features by analyzing the time domain characteristics of the data set.

[0031] To eliminate the differences in dynamic response signals under different road surface conditions, dimensionless parameters such as kurtosis factor, skewness factor, margin factor, peak factor, and waveform factor were selected as key characteristic indicators.

[0032] The formula for calculating the kurtosis factor is as follows: (1); In the formula, The kurtosis factor is a classification feature used to characterize the flatness of the data distribution. This is the vertical acceleration time series data corresponding to the c-th sampling point. for The first in n Vertical acceleration data at time t, N for Number of data points for The average value of the data. for The root mean square value of the data.

[0033] The formula for calculating the skewness factor is: (2); In the formula, The skewness factor is a classification feature used to reflect the directional characteristics of the road surface excitation signal.

[0034] The formula for calculating the margin factor is as follows: (3); In the formula, As a margin factor, this classification feature is used to reflect the high-frequency characteristics of road surface excitation. for The peak value of the data.

[0035] The formula for calculating the peak factor is: (4); In the formula, As the peak factor, this classification feature is used to detect extreme outliers in vertical acceleration time series data.

[0036] The formula for calculating the waveform factor is: (5); In the formula, The waveform factor is used to detect the distribution of vertical acceleration time series data.

[0037] Step 2022: Based on the classification features of all sampling points in the current period and the trained road surface grade preclassifier, determine the road surface grade preclassification result for the current period; the trained road surface grade preclassifier is a classifier obtained by training a preclassification model built based on the AdaBoot algorithm.

[0038] As an example, AdaBoost (Adaptive Boosting) is an ensemble learning algorithm whose core idea is to train multiple weak classifiers sequentially and adaptively adjust sample weights based on their performance, ultimately combining them into a strong classifier.

[0039] Before training, a sample set and a pre-classification model are constructed. A sample in the sample set is the classification feature corresponding to all sampling points (also called sampling channels or channels) within a period. Different samples in the sample set correspond to different periods. The pre-classification model includes multiple weak classifiers, and each subsequent weak classifier is used to focus on the previously misclassified sample.

[0040] The specific training process is as follows: Step 2.2.1 Initialize the weight distribution.

[0041] Based on the input sample set, the weight distribution of the weak classifiers in the pre-classification model is determined. Initialization is performed, represented as: (6); in, The weight distribution of the pre-classification model in the first iteration (initialization). m The sample number. Nz The total number of samples, This refers to the road surface grade, corresponding to the eight grades A through H as defined by ISO 8608.

[0042] Step 2.2.2 Weak classifier output construction.

[0043] With the first t Taking the nth iteration as an example, in the nth iteration t In the next iteration, a weak classifier is constructed. , indicating the first m Input Samples Road surface grade The confidence probability is calculated using the following formula: (7).

[0044] Step 2.2.3 Calculate the pseudo loss.

[0045] The pseudo-loss of the weak classifier in the AdaBoost algorithm This is used to quantify the model's ability to distinguish between true and false labels at the t-th iteration. Based on the classification features in the samples, the t-th iteration is obtained. t The pseudo-loss of the next iteration The calculation formula is: (8); in, , indicating the first m The classification feature vector is the concatenation of the classification features of all channels (corresponding to sampling points) of a sample. Indicates the first m The kurtosis factor of the first channel of each sample. Indicates the first m The skewness factor of the first channel of each sample. Indicates the first m The margin factor of the first channel of each sample. Indicates the first m The peak factor of the first channel of each sample, Indicates the first mThe waveform factor of the first channel of each sample. Indicates the first m The waveform factor of the 9th channel of a sample. For the first m The true value of the road surface grade for each sample. It is a set of error road surface grades.

[0046] Step 2.2.4 Calculate the normalization factor.

[0047] For the t The weights of the next iteration are normalized to obtain the normalization factor. The calculation formula is: (9); in, Update the parameters for the weights.

[0048] Step 2.2.5 Update the weight distribution.

[0049] Based on the pseudo-loss calculated in step 2.2.3 and the normalization factor calculated in step 2.2.4, we can obtain... t Weight distribution in +1 iteration The calculation formula is: (10).

[0050] Step 2.2.6 Calculate the weighted cumulative confidence level.

[0051] Based on steps 2.2.1 to 2.2.5, after iteration, the weighted cumulative confidence scores of each weak classifier can be calculated to obtain the weighted cumulative confidence scores of the classifiers. Represented as: (11); Where M is the total number of iterations for training the pre-classifier.

[0052] Thus, according to steps 2.2.1 to 2.2.6, the training process of the pre-classification model is completed (and a trained road surface grade pre-classifier is obtained).

[0053] Compare confidence scores for all road surface grades The maximum confidence level corresponds to the pavement grade pre-classification result, that is, the pavement grade pre-classification result for the current period k is determined. The expression is: (12); Indicates the current period k The classification feature vector corresponding to the vertical acceleration time series data set, current period k Road surface grade pre-classification results The numerical value represents the road surface roughness level. These correspond to road surfaces of grades A through H, respectively.

[0054] In another exemplary embodiment of this application, the simplified structure of the multi-axle vehicle includes a body, a suspension system, and a running gear consisting of multiple wheels.

[0055] The vehicle body has three degrees of freedom of motion relative to the ground, including vertical vertical jump, lateral rotation about the vehicle's longitudinal axis, and pitch rotation about the vehicle's lateral axis; each wheel has one degree of freedom of motion relative to the ground, including vertical vertical jump.

[0056] The suspension system includes multiple axles arranged longitudinally along the vehicle, each axle having a suspension node at each end, and each suspension node being connected to a wheel.

[0057] The process of determining the state-space equations is as follows: The vertical displacement of each suspension node is determined based on the simplified structure.

[0058] Based on the vertical displacement and all the degrees of freedom of motion, the differential equations of dynamics for multi-axle vehicles are determined.

[0059] Using the vertical acceleration response of the vehicle body and the vertical acceleration response of each wheel as observations, the dynamic differential equation of the multi-axle vehicle is transformed to obtain the state-space equation of the multi-axle vehicle.

[0060] As an example, a whole vehicle dynamics model is established, and based on the dynamic equations of this model, a standard state-space expression for state observation is obtained. The specific steps are as follows: Step 4.1 Simplify the vehicle system structure.

[0061] like Figure 3 As shown, taking an 8-wheel (4-axle) vehicle as an example (the number of axles can be added or removed to expand the model for other vehicle structures), the main structure of a multi-axle vehicle is simplified to a body, a suspension system, and a running gear consisting of multiple wheels. The body has three degrees of freedom relative to the ground: vertical sway, roll rotation about the vehicle's longitudinal axis, and pitch rotation about the vehicle's lateral axis. Each wheel has one degree of freedom relative to the ground, including vertical sway. The suspension system includes a first axle, a second axle, a third axle, and a fourth axle arranged along the vehicle's longitudinal direction (from front to rear). Each axle has a suspension node at both ends, and each suspension node is connected to one of the wheels. In road surface unevenness identification, the multi-axle vehicle's heading remains basically stable, and yaw motion is ignored.

[0062] Figure 3 middle, This indicates the vertical displacement of the vehicle body. Indicates the pitch angle of the vehicle body. Indicates the body roll angle. , , , and The vertical movement corresponds to the left wheel of the first axle, the right wheel of the first axle, the left wheel of the second axle, the left wheel of the third axle, and the left wheel of the fourth axle, respectively. Indicates the first i The vertical jump corresponding to each wheel, the subscript i =1,2,…,8 correspond to the left wheel of the first axle, the right wheel of the first axle, the left wheel of the second axle, the right wheel of the second axle, the left wheel of the third axle, the right wheel of the third axle, the left wheel of the fourth axle, and the right wheel of the fourth axle, respectively. and The road surface excitations are respectively for the left and right wheels of the first axle, with... Indicates the first i The road surface excitation corresponding to each wheel, the subscript i =1,2,…,8 correspond to the left wheel of the first axle, the right wheel of the first axle, the left wheel of the second axle, the right wheel of the second axle, the left wheel of the third axle, the right wheel of the third axle, the left wheel of the fourth axle, and the right wheel of the fourth axle, respectively.

[0063] Step 4.2 Calculate the vertical displacement of each suspension node.

[0064] Based on the simplified vehicle structure in step 4.1, the expressions for the vertical displacements of each suspension node of the multi-axle vehicle can be obtained as follows: (13); in, This is the distance from the vehicle's center of gravity to the first axle. This is the distance from the vehicle's center of gravity to the fourth axle. The wheelbase is the distance between the wheels. , , and The vertical displacements of the suspension nodes corresponding to the left and right wheels of the first axle, the left and right wheels of the fourth axle, respectively, are as follows: Indicates the first i Vertical displacement of the suspension node corresponding to each wheel, index i =1,2,…,8 correspond to the left wheel of the first axle, the right wheel of the first axle, the left wheel of the second axle, the right wheel of the second axle, the left wheel of the third axle, the right wheel of the third axle, the left wheel of the fourth axle, and the right wheel of the fourth axle, respectively.

[0065] Step 4.3 Construct the differential equations of vehicle dynamics.

[0066] Based on the vehicle's degrees of freedom and the vertical displacement of the suspension nodes in steps 4.1 and 4.2, the dynamic equations of the simplified vehicle structure are derived, resulting in the dynamic differential equations of a multi-axle vehicle as follows: (14); in, For the body mass of multi-axle vehicles, The moment of inertia of the vehicle body during roll. Let the vehicle body pitch moment of inertia be... It is the first i The mass of each wheel For the center of gravity of the car body to the first i The distance of each axis, It is the first i The stiffness coefficient of each suspension, It is the first i The damping coefficient of each suspension, It is the first i The stiffness coefficient of each wheel, where the subscript is... i =1,2,…,8 correspond to the left wheel of the first axle, the right wheel of the first axle, the left wheel of the second axle, the right wheel of the second axle, the left wheel of the third axle, the right wheel of the third axle, the left wheel of the fourth axle, and the right wheel of the fourth axle, respectively.

[0067] Step 4.4 Obtain the matrix form of the vehicle dynamics equations.

[0068] The vehicle dynamics differential equations are rearranged, and relevant parameter matrices are constructed to obtain the matrix form of the vehicle dynamics equations. The specific process is as follows: Step 4.4.1 Calculate the system mass matrix.

[0069] The mass matrix is ​​calculated using the following formula: (15); () represents the diagonal matrix construction operator, which places the elements given in parentheses sequentially on the main diagonal of the matrix, and sets all off-diagonal elements to 0.

[0070] Step 4.4.2 Calculate the suspension stiffness and damping matrix.

[0071] The dynamic differential equations from step 4.3 are rearranged into standard matrix form to obtain the stiffness matrix of the suspension system. and damping matrix The calculation formula is: (16); (17); in: ; .

[0072] Step 4.4.3 Construct the dynamic equation matrix of the multi-axle vehicle.

[0073] Based on the dynamic differential equation in step 4.3 and the parameter matrices in steps 4.4.1 to 4.4.2, formula (14) is rearranged into matrix form and expressed as: (18); in, For state variables, ; It is the system's road surface excitation input vector. ; It is a matrix of all zeros. This is the coefficient matrix of the road surface excitation. .

[0074] Step 4.5 Obtain the system state-space equations.

[0075] Based on step 4.4, the matrix equations are rearranged into the standard form of the state-space equations, using the vehicle body's vertical acceleration response and the vertical acceleration responses of each wheel as observables: (19); (20); in, A , B , H and D For intermediate parameters, the expression is: ; ; ; ; In the formula, X For multi-axle vehicles, these are state variables. , This refers to the state changes of multi-axle vehicles. Y For observation of multi-axle vehicles, , For inputs to multi-axle vehicles, Here is the stiffness matrix. Here is the damping matrix. For the quality matrix, It is a matrix of all zeros. This is the coefficient matrix of the road surface excitation. It is a matrix of all 1s, with subscripts... , , , and Indicates the number of dimensions of the matrix.

[0076] Step 4.6 Obtain the discrete state-space equations of the system.

[0077] Discretize the state-space equations obtained in step 4.5 to obtain the discrete-time state-space model as follows: (twenty one); (twenty two); in, , , For multi-axle vehicles in the next cycle k +1 state quantity, For multi-axle vehicles in the current cycle k The state variables, For multi-axle vehicles in the next cycle k +1 input, For multi-axle vehicles in the next cycle k +1 observation, For multi-axle vehicles in the current cycle k The state transition matrix, For multi-axle vehicles in the current cycle k The input gain matrix, The time difference between adjacent quantity periods. For time integral infinitesimal elements, It is a time variable.

[0078] In another exemplary embodiment of this application, step 203, which determines the Kalman gain for the current period based on the forgetting factor and process noise covariance matrix of the unknown input Kalman filter algorithm for the current period, specifically includes: Step 2031: Based on the discrete state-space equation of the system, the prior estimate of the error covariance matrix of the previous period, and the Kalman gain, determine the posterior estimate of the error covariance matrix of the previous period.

[0079] Step 2032: Based on the discrete state-space equation of the system, the posterior estimate of the error covariance matrix of the previous period, and the forgetting factor and process noise covariance matrix of the unknown input Kalman filter algorithm of the current period, determine the prior estimate of the error covariance matrix of the current period.

[0080] Step 2033: Determine the Kalman gain for the current period based on the discrete state-space equation of the system and the prior estimate of the error covariance matrix for the current period.

[0081] In another exemplary embodiment of this application, step 204 is replaced by the following steps: Step 2041: Based on the system's discrete state-space equations, the road excitation signal of the previous cycle, the prior state estimate, the vertical acceleration time series data set, and the Kalman gain, determine the posterior state estimate of the previous cycle.

[0082] Step 2042: Based on the system's discrete state-space equations, the posterior state estimate of the previous period, and the road excitation signal, determine the prior state estimate of the current period.

[0083] Step 2043: Based on the system's discrete state-space equations, the unknown input Kalman filter algorithm, the prior state estimate of the current period, the vertical acceleration time series data set, and the Kalman gain, determine the road excitation signal for the current period.

[0084] Through steps 203 and 204 above, a pre-classifier is used to perform prior classification and identification of road surface grades. Based on the pre-classification results, the forgetting factor and process noise covariance matrix of the unknown input Kalman filter algorithm are dynamically adjusted to adapt to varying road surface roughness levels. As an example, such as... Figure 4 As shown, a method is provided for dynamically adjusting the unknown input Kalman filter algorithm and calculating the road excitation signal based on the road surface grade pre-classification result, including: Step 5.1 State prediction based on pre-classifier.

[0085] Step 5.1.1 Update parameter configuration based on pavement grade pre-classification results.

[0086] The forgetting factor for the current period is determined based on a preset value and the change in road surface grade; the expression for the forgetting factor is: (twenty three); In the formula, For the current period k The forgetting factor, and This is a preset value. For the current period k Road surface grade, For the previous cycle k -1 is the road surface grade. The forgetting factor is used when the road surface grade remains unchanged. It is taken as a large value close to 1, such as the range of 0.985 to 0.995. The forgetting factor used when the road surface grade changes is taken as less than 1. The value of is 0.900 to 0.970.

[0087] The process noise covariance matrix for the current period is determined based on the road surface grade, the road surface grade change state, and a preset parameter table; the expression for the process noise covariance matrix is: (twenty four); In the formula, For the current period k The process noise covariance matrix, For the previous cycle k The process noise covariance matrix is ​​-1. This is obtained by looking up the preset parameter table. The corresponding process noise covariance matrix.

[0088] As an example, during the current cycle k and k The pre-classification results for road surface grades of -1 are consistent, that is, they satisfy the requirements. At that time, it was considered that the overall road surface conditions remained stable, and the currently used methods continued to be adopted. And set the forgetting factor to a value During the current cycle k Compared with the previous cycle k The pre-classification results for road surface grade -1 are inconsistent, i.e., they meet the requirements. If the road surface grade changes, the parameters from the previous cycle will no longer be used; instead, the parameters from the current cycle will be used according to the preset parameter table. k The corresponding value is obtained Take a value, and set the forgetting factor to a value. In other words, when the road surface condition is detected to be stable, The value approaches 1 to maintain steady-state accuracy; when the pre-classifier reports changes in road surface grade, the accuracy is reduced. The value is selected to increase the weight of the prediction error covariance and improve the sensitivity of the Kalman filter algorithm to the observations.

[0089] As an example, the preset parameter table is determined by pre-calibrating the sampled data of road surfaces of known grades. The process for determining the preset parameter table is as follows: For the current period k The road surface grade is determined by pre-setting a corresponding target process noise covariance matrix for different road surfaces, denoted as . .in It is the process noise covariance matrix of a Class C road surface. This is the ratio of the current road surface grade to the medium-grade road surface (Grade C road surface). .

[0090] A data segment of length can be collected under known Class C road surface conditions. The offline data, in which the sensor (measuring vertical acceleration) output signal sequence is denoted as... Several candidate baseline process noise covariance matrices are preset. Pre-calibration is performed sequentially, the identification process is carried out offline, and the predicted output value is compared with the actual response output. This process is repeated for the previous cycle. k -1 can be used to construct the residual as For the entire offline data, the residual mean square index can be obtained as follows: (25).

[0091] Among them, when constructing the residual, Indicates the previous period k The predicted output value is -1. .

[0092] Finally, in the candidate baseline process noise covariance matrix, the residual mean square index is selected. The smallest candidate matrix is ​​used as the baseline process noise covariance matrix. In one example, the baseline process noise covariance matrix for , It is a 22-order identity matrix.

[0093] The process noise covariance matrix reflects the equivalent process noise intensity of the vehicle state equation under road surface input. This equivalent intensity is affected not only by road surface roughness but also by the vehicle dynamics model, thus exhibiting a nonlinear relationship with the road surface roughness index. Therefore, in different application scenarios, a scaling factor can be established based on the geometric mean of the road surface roughness index corresponding to different road surface grades. (26); In the formula, for The geometric mean of the pavement roughness index for Class I pavement. The geometric mean of the road surface roughness index is for Class C road surfaces. This is the mapping coefficient, with a reference value range of 0.5 to 1.0. In one example, Set it to 0.5. .

[0094] In one example, the preset parameters for the process noise covariance matrix are shown in Table 1.

[0095] Table 1 Preset Parameters for Process Noise Covariance Matrix

[0096] Step 5.1.2 Calculate the prior state estimate.

[0097] Utilizing the previous cycle k The posterior state estimate of -1, combined with the discrete state-space equation of the system constructed in step 4.6, yields the current period. k The prior state estimate is expressed as: (27); Furthermore, the expression for the prior estimate of the error covariance matrix for the current period is: (28); in, For the current period k The prior estimate of the error covariance matrix, For the previous cycle k -1 is the posterior estimate of the error covariance matrix. The current period in the discrete state-space equations of the system k The state transition matrix, for transpose, Indicates the current period k The prior state estimate, Indicates the previous period k -1 is the posterior state estimate. Indicates the previous period k -1 represents the road surface excitation signal. (Subscript) k Represents the current cycle k subscript k -1 represents the previous cycle. k -1, subscript k / k -1 represents the previous cycle. k -1 for the current period k The prior estimate (i.e., the prior estimate for the current period), subscript k / k Represents the current cycle k The posterior estimate, subscript k -1 / k -1 represents the previous cycle. k The posterior estimate. Formula (27) is used to implement the process of step 2042. Formula (28) is used to implement the process of step 2032.

[0098] Step 5.2 Calculate the Kalman gain matrix.

[0099] Based on the discrete state-space equations of the system constructed in step 4.6 and the prior estimate of the error covariance matrix for the current period calculated in step 5.1, the Kalman gain for the current period can be calculated. Represented as: (29); in, To measure the noise covariance matrix, formula (29) is used to implement the process in step 2033.

[0100] Step 5.3 Construct the unknown input estimator (unknown input Kalman filter algorithm).

[0101] Based on the discrete state-space equations constructed in step 4.6 and the prior state estimates for the current period calculated in steps 5.1 and 5.2 respectively. Kalman gain for the current period It estimates the unknown road surface roughness input and calculates the road surface excitation signal for the current cycle. Represented as: (30); (31); in, It is the identity matrix. Let be the covariance matrix for the current period. For multi-axle vehicles in the current cycle k The observations (i.e., the set of vertical acceleration time-series data for the current period obtained in step 201, corresponding to...) Figure 4 (Sensor data in the data). Formulas (30) and (31) are used to implement the process of step 2043.

[0102] Step 5.4 Correct the state estimate.

[0103] Based on the calculation results of steps 4.6 and 5.1 to 5.3, the calculated posterior state estimate, Kalman gain, and unknown input estimate input state update equation are used to eliminate the bias of the unknown input on the state estimate, thus obtaining the posterior (optimal) state estimate for the current period. The data used for the next period's recursion is represented as: (32).

[0104] Formula (32) calculates the posterior state estimate of the current period. Similarly, the posterior state estimate of the previous period can be calculated to achieve the process of step 2041. This will not be elaborated here.

[0105] Step 5.5 Correct the error covariance matrix.

[0106] Considering the additional uncertainty introduced by the unknown input estimation, the posterior estimate of the error covariance matrix for the current period is... The correction and update are performed for the next cycle of recursion, and are represented as follows: (33).

[0107] Formula (33) calculates the posterior estimate of the error covariance matrix for the current period. Similarly, the posterior estimate of the error covariance matrix for the previous period can be calculated to achieve the process of step 2031. This will not be elaborated here.

[0108] Steps 5.1 to 5.5 completed the identification and recursive process of the unknown input Kalman filter algorithm, resulting in the unknown input estimate. The road surface excitation signal is identified.

[0109] In another exemplary embodiment of this application, the multi-axle vehicle includes multiple axles arranged longitudinally along the vehicle, with a wheel connected to each axle on its left and right sides respectively; step 205, which calculates the road surface roughness power spectral density for the current period based on the road surface excitation signal of each period within a preset time window, specifically includes: Step 2051: Determine the sampling delay of each axle based on the longitudinal distance of each axle relative to the vehicle's center of gravity and the vehicle speed in the current cycle.

[0110] Step 2052: Based on the road surface excitation signals of each cycle within a preset time window and all the sampling delays, determine the left-side fused road surface excitation signal and the right-side fused road surface excitation signal; the left-side fused road surface excitation signal is determined based on the road surface excitation signals of the left wheels of all the axles, and the right-side fused road surface excitation signal is determined based on the road surface excitation signals of the right wheels of all the axles.

[0111] Step 2053: Based on the left fused road surface excitation signal and the right fused road surface excitation signal, calculate the power spectral density of the left road surface roughness and the power spectral density of the right road surface roughness, respectively.

[0112] As an example, the ISO 8608 standard classifies road roughness levels based on pavement power spectral density (PSD), according to the pavement roughness power spectral density curve. The road surfaces are divided into eight grades, A to H. The standard values ​​of the power spectral density for grades A to H are shown in Table 2. This is the reference spatial frequency.

[0113] Step 6.1 Construct a single-sided fusion pavement excitation.

[0114] Based on the road excitation signals identified in steps 5.1 to 5.5, the road excitation estimation results corresponding to each wheel position can be obtained. Taking a 4-axle, 8-wheel vehicle as an example, we have: (34); in, The road surface excitation signals corresponding to the left wheels of the first, second, third, and fourth axles. The road surface excitation signals corresponding to the right wheels of the first, second, third, and fourth axles. The period corresponding to the road surface excitation signal. , a It is a positive integer. The sampling period.

[0115] Because there is an inter-axle time delay relationship when different wheels on the same side pass through the same road segment, in order to improve the consistency of road surface identification, this step performs time delay alignment on the road surface excitation estimation signals of the wheels on both sides, and constructs the left-side fused road surface excitation signal separately. and the right-side road surface excitation signal : (35); (36); in, This represents the sampling delay of the wheel position on the same side relative to the reference wheel position. In specific calculations, Take values ​​respectively The sampling delays of the second, third, and fourth axes relative to the first axis are calculated as follows: (37); In the formula, For vehicle speed, This refers to the longitudinal distance of each axis relative to the center of gravity of the vehicle body, as defined in step 4.2.

[0116] In applying discrete algorithms, a preset sliding time window (a time window with the current period as the last time period) is used. This preset time window includes road surface excitation signals from multiple periods. To achieve delay alignment, The value should be such that it can still be retrieved in the data segment after time delay alignment. Real-time valid data.

[0117] Step 6.2 Calculate the power spectral density of a single-sided road surface.

[0118] Based on the road excitation signal identified in step 6.1, the power spectral density of the road surface roughness on the left side can be obtained. Power spectral density of road surface roughness on the right side The calculation formula is: (38); (39); in, T The length of the preset time window.

[0119] Table 2 ISO 8608 Classification Standard for Road Surface Roughness

[0120] The road surface roughness index in Table 2 is the road surface power spectral density at the reference space frequency. The values ​​at the given location. Based on the pavement power spectral density calculated using formulas (38) and (39) and Table 2, the final identification result (pavement grade) can be obtained.

[0121] This application divides the multi-axle vehicle road roughness identification process based on dynamic response into a pre-classification process based on time-domain analysis and an identification process based on an unknown input Kalman filter algorithm. To address the identification lag problem of the unknown input Kalman filter algorithm under varying road conditions, combining road level-based pre-classification with the unknown input Kalman filter algorithm is a feasible solution to improve online identification efficiency. Based on the vehicle dynamic time-domain response, the road level is classified a priori, and parameters such as the forgetting factor and process noise covariance matrix of the unknown input Kalman filter algorithm are determined based on the prior classification results. This makes the current identification model more adaptable to transient changes in road conditions. The final road level is determined based on the vehicle dynamic frequency-domain response (calculating the road roughness power spectral density).

[0122] In one exemplary embodiment, a multi-axle vehicle is provided, the internal structure of which can be shown in the following diagram. Figure 5As shown, the multi-axle vehicle includes a body, suspension system, running gear, processor, memory, input / output (I / O) interface, and communication interface. The processor, memory, and I / O interface are connected via a system bus, and the communication interface is also connected to the system bus via the I / O interface. The processor provides computational and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system, computer programs, and a database. The internal memory provides the environment for the operating system and computer programs in the non-volatile storage media. The database stores road surface roughness identification data (e.g., time-series data sets of vertical acceleration for each cycle, road excitation signals, road surface roughness power spectral density, and road surface grade). The I / O interface allows the processor to exchange information with external devices. The communication interface allows communication with external terminals via a network connection. When the computer program is executed by the processor, it implements a road surface roughness identification method.

[0123] Those skilled in the art will understand that Figure 5 The structures shown are merely block diagrams of some structures related to the present application and do not constitute a limitation on the multi-axle vehicles to which the present application is applied. Specific multi-axle vehicles may include more or fewer components than shown in the figures, or combine certain components, or have different component arrangements. In an exemplary embodiment, a multi-axle vehicle is provided, including a body, a suspension system, a running gear system, a memory, and a processor. The memory stores a computer program, and the processor executes the computer program to implement the steps in the above-described method embodiments.

[0124] In one exemplary embodiment, a computer-readable storage medium is provided storing a computer program that, when executed by a processor, implements the steps in the above-described method embodiments.

[0125] In one exemplary embodiment, a computer program product is provided, including a computer program that, when executed by a processor, implements the steps in the above-described method embodiments.

[0126] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, data stored, data displayed, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties. Moreover, the collection, use and processing of the relevant data are carried out in compliance with the relevant data protection laws and policies of the country where the location is located, and with the authorization granted by the owner of the corresponding device.

[0127] Those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium. When executed, the computer program can include the processes of the embodiments described above. Any references to memory, databases, or other media used in the embodiments provided in this application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can take many forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM).

[0128] The databases involved in the embodiments provided in this application may include at least one type of relational database and non-relational database. Non-relational databases may include, but are not limited to, blockchain-based distributed databases. The processors involved in the embodiments provided in this application may be general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic devices, quantum computing-based data processing logic devices, etc., and are not limited to these.

[0129] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0130] This document uses specific examples to illustrate the principles and implementation methods of this application. The descriptions of the above embodiments are only for the purpose of helping to understand the methods and core ideas of this application. Furthermore, those skilled in the art will recognize that, based on the ideas of this application, there will be changes in the specific implementation methods and application scope. Therefore, the content of this specification should not be construed as a limitation of this application.

Claims

1. A method for identifying road surface roughness, characterized in that, The road surface unevenness identification method includes: Obtain the vertical acceleration time series data set for the current period; each vertical acceleration time series data in the vertical acceleration time series data set corresponds to a sampling point, and the sampling point is the center of the bottom of the multi-axle vehicle or the center of any wheel in the multi-axle vehicle; Based on the current period's vertical acceleration time series data set, the road surface grade is pre-classified to determine the pre-classification result of the road surface grade for the current period; Based on the road surface grade pre-classification results of the current period and the road surface grade pre-classification results of the previous period, the forgetting factor and process noise covariance matrix of the unknown input Kalman filter algorithm for the current period are determined, and the Kalman gain for the current period is determined based on the forgetting factor and process noise covariance matrix of the unknown input Kalman filter algorithm for the current period. Based on the system discrete state-space equation, the unknown input Kalman filter algorithm, the current period's vertical acceleration time-series data set, and the Kalman gain, the road excitation signal for the current period is determined; the system discrete state-space equation is the equation obtained by discretizing the state-space equation of the multi-axle vehicle, and the state-space equation is established based on the vehicle dynamics model of the multi-axle vehicle; When the current period is greater than a preset value, the road surface roughness power spectral density of the current period is calculated based on the road surface excitation signal of each period within the preset time window, and the road surface roughness classification standard is combined to determine the road surface grade of the current period; wherein, the preset time window is a time window with the current period as the last time period.

2. The road surface roughness identification method according to claim 1, characterized in that, Based on the current period's vertical acceleration time-series data set, the road surface grade is pre-classified to determine the pre-classification result for the current period, specifically including: Based on the current period's vertical acceleration time series data set, the classification features of each sampling point in the current period are determined; the classification features include kurtosis factor, skewness factor, margin factor, peak factor, and waveform factor. Based on the classification features of all sampling points in the current period and the trained road surface grade pre-classifier, the road surface grade pre-classification result for the current period is determined; the trained road surface grade pre-classifier is a classifier obtained by training a pre-classification model built on the AdaBoot algorithm.

3. The road surface roughness identification method according to claim 1, characterized in that, The simplified structure of the multi-axle vehicle includes a body, a suspension system, and a running gear consisting of multiple wheels; The vehicle body has three degrees of freedom of motion relative to the ground, including vertical vertical jump, lateral rotation about the vehicle's longitudinal axis, and pitch rotation about the vehicle's lateral axis; each wheel has one degree of freedom of motion relative to the ground, including vertical vertical jump. The suspension system includes multiple axles arranged longitudinally along the vehicle, each axle having a suspension node at each end, and each suspension node being connected to a wheel. The process of determining the state-space equations is as follows: The vertical displacement of each suspension node is determined based on the simplified structure described above; Based on the vertical displacement and all the degrees of freedom of motion, the differential equations of dynamics for the multi-axle vehicle are determined; Using the vertical acceleration response of the vehicle body and the vertical acceleration response of each wheel as observations, the dynamic differential equation of the multi-axle vehicle is transformed to obtain the state-space equation of the multi-axle vehicle.

4. The method for identifying road surface roughness according to claim 1, characterized in that, Based on the system's discrete state-space equations, the unknown input Kalman filter algorithm, the current period's vertical acceleration time-series data set, and the Kalman gain, the road surface excitation signal for the current period is determined, specifically including: Based on the system's discrete state-space equations, the road excitation signal of the previous cycle, the prior state estimate, the set of vertical acceleration time series data, and the Kalman gain, the posterior state estimate of the previous cycle is determined. Based on the system's discrete state-space equations, the posterior state estimate of the previous cycle, and the road excitation signal, the prior state estimate of the current cycle is determined. Based on the system's discrete state-space equations, the unknown input Kalman filter algorithm, the prior state estimate of the current cycle, the vertical acceleration time series data set, and the Kalman gain, the road excitation signal for the current cycle is determined.

5. The method for identifying road surface roughness according to claim 1, characterized in that, The Kalman gain for the current period is determined based on the forgetting factor and process noise covariance matrix of the unknown input Kalman filter algorithm for the current period, specifically including: Based on the discrete state-space equation of the system, the prior estimate of the error covariance matrix of the previous cycle, and the Kalman gain, the posterior estimate of the error covariance matrix of the previous cycle is determined. Based on the discrete state-space equation of the system, the posterior estimate of the error covariance matrix of the previous period, and the forgetting factor and process noise covariance matrix of the unknown input Kalman filter algorithm in the current period, the prior estimate of the error covariance matrix of the current period is determined. Based on the discrete state-space equation of the system and the prior estimate of the error covariance matrix for the current period, the Kalman gain for the current period is determined.

6. The method for identifying road surface roughness according to claim 5, characterized in that, The forgetting factor for the current cycle is determined based on a preset value and the change in road surface grade. The expression for the forgetting factor is: ； In the formula, For the current period k The forgetting factor, and This is a preset value. For the current period k Road surface grade, For the previous cycle k -1 is the road surface grade. The process noise covariance matrix for the current period is determined based on the road surface grade, the road surface grade change state, and a preset parameter table; the expression for the process noise covariance matrix is: ； In the formula, For the current period k The process noise covariance matrix, For the previous cycle k The process noise covariance matrix is ​​-1. This is obtained by looking up the preset parameter table. The corresponding process noise covariance matrix; The expression for the prior estimate of the error covariance matrix for the current period is: ； In the formula, For the current period k The prior estimate of the error covariance matrix, For the previous cycle k -1 is the posterior estimate of the error covariance matrix. The current period in the discrete state-space equations of the system k The state transition matrix, for The transpose of .

7. The method for identifying road surface roughness according to claim 1, characterized in that, The multi-axle vehicle includes multiple axles arranged longitudinally along the vehicle, with a wheel connected to each of the left and right sides of each axle. Based on the road excitation signals of each cycle within a preset time window, the road roughness power spectral density of the current cycle is calculated, specifically including: The sampling delay for each axle is determined based on the longitudinal distance of each axle relative to the vehicle's center of gravity and the vehicle speed in the current cycle. Based on the road surface excitation signals of each cycle within a preset time window and all the sampling delays, the left-side fused road surface excitation signal and the right-side fused road surface excitation signal are determined; the left-side fused road surface excitation signal is determined based on the road surface excitation signals of the left wheels of all the axles, and the right-side fused road surface excitation signal is determined based on the road surface excitation signals of the right wheels of all the axles. Based on the left-side fused road surface excitation signal and the right-side fused road surface excitation signal, the power spectral density of the left-side road surface roughness and the power spectral density of the right-side road surface roughness are calculated respectively.

8. A multi-axle vehicle, comprising: A vehicle body, a suspension system, a running gear system, a memory, a processor, and a computer program stored in the memory and capable of running on the processor, characterized in that the processor executes the computer program to implement the road surface roughness identification method according to any one of claims 1-7.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the road surface roughness identification method according to any one of claims 1-7.

10. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by the processor, it implements the road surface roughness identification method according to any one of claims 1-7.

Images