Road surface unevenness estimation method and device, terminal and medium

Abstract

The invention provides a road surface unevenness estimation method and device, a terminal and a medium, and the method comprises the steps: inputting an obtained random filtering white noise road surface into a vehicle quarter nonlinear suspension model to generate a dynamic response amount, and obtaining a sprung acceleration and an unsprung acceleration; inputting the sprung acceleration and the unsprung acceleration into an NARX neural network to predict the road surface unevenness to obtain a first estimated road surface unevenness value; obtaining a second estimated road surface unevenness value according to the sprung acceleration, the unsprung acceleration, the first estimated road surface unevenness value and an international road surface unevenness index; and fusing the first estimated road surface unevenness value and the second estimated road surface unevenness value by adopting a fusion algorithm to obtain a road surface unevenness estimated value. Through the fusion algorithm, the problem that the pavement unevenness estimation precision is reduced due to nonlinearity and uncertainty of the model can be effectively relieved, and the pavement unevenness estimation efficiency is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of vehicle engineering, and particularly to a method, device, terminal and medium for estimating road surface unevenness. Background Art

[0002] Road surface unevenness is the main excitation during vehicle driving, which has a great impact on vehicle ride comfort and passenger comfort. At the same time, road surface unevenness information is of great significance for vehicle dynamics control, especially suspension control. In suspension control, it is necessary to meet the dual requirements of comfort and safety. However, for various types of uneven roads, a single control parameter is difficult to achieve the optimal control effect. Therefore, accurate road surface unevenness information can provide a basis for parameter adjustment of the adaptive control algorithm.

[0003] Currently, road surface unevenness recognition technologies can be divided into three categories: direct measurement method, indirect measurement method, and dynamic response-based method. The direct measurement method refers to using professional road profile meters and other equipment for measurement. Generally, it requires special fixed costs and is relatively high, making it difficult to be commercialized on a large scale. Most of the indirect measurement methods are based on technologies such as lidar and millimeter-wave radar. Their estimation accuracy is high, but they are greatly affected by weather conditions. The estimation methods based on dynamic response can be divided into model-based methods and data-driven methods. In the model-based methods, the linear Kalman filter method is widely used. It requires a simple model and small computational complexity. However, for non-linear models, the non-linearity of its parameters and the mismatch between the model parameters and the actual system will be the key problems in practical applications. For data-based methods, they utilize the potential features in the learning response signal without specifically analyzing the physical process. However, when facing unknown road surfaces, it is often difficult to maintain a high estimation accuracy. In addition, considering that most current road surface recognition algorithms are designed based on linear suspension parameters, and in semi-active or active suspensions, the spring stiffness or damping coefficient often has non-linear situations, which will lead to a decrease in estimation accuracy. Summary of the Invention

[0004] Aiming at the deficiencies of the existing technology, the present invention proposes a method for estimating road surface unevenness, aiming to improve the estimation accuracy of road surface unevenness on unknown roads and be able to effectively estimate road surface unevenness under the condition that the selected parameters are non-linear.

[0005] In a first aspect, an embodiment of the present application provides a method for estimating road surface unevenness, including:

[0006] Inputting the obtained random filtered white noise road surface into a quarter-vehicle non-linear suspension model to generate dynamic response quantities, obtaining the acceleration above the spring and the acceleration below the spring;

[0007] Input the above-spring acceleration and below-spring acceleration into a NARX neural network for predicting road surface unevenness to obtain a first estimated road surface unevenness value;

[0008] Obtain a second estimated road surface unevenness value based on the above-spring acceleration, below-spring acceleration, first estimated road surface unevenness value, and international road surface unevenness index;

[0009] Use a fusion algorithm to fuse the first estimated road surface unevenness value and the second estimated road surface unevenness value to obtain an estimated road surface unevenness value.

[0010] Optionally, the step of obtaining a second estimated road surface unevenness value based on the above-spring acceleration, below-spring acceleration, first estimated road surface unevenness value, and international road surface unevenness index includes:

[0011] Obtain an adaptive covariance matrix based on the first estimated road surface unevenness value and the international road surface unevenness index;

[0012] Input the above-spring acceleration, below-spring acceleration, and adaptive covariance matrix into a Kalman filter to obtain a second estimated road surface unevenness value.

[0013] Optionally, the step of inputting the above-spring acceleration, below-spring acceleration, and adaptive covariance matrix into a Kalman filter to obtain a second estimated road surface unevenness value includes:

[0014] Input the above-spring acceleration and below-spring acceleration into the state space equation of a nonlinear system, and combine with the adaptive covariance matrix to update the state of the Kalman filter, obtaining a state space predicted value and a square root factor of the error covariance matrix;

[0015] Update the state variables of the Kalman filter based on the above-spring acceleration, below-spring acceleration, state space predicted value, and square root factor of the error covariance matrix to obtain a second estimated road surface unevenness value.

[0016] Optionally, the step of obtaining an adaptive covariance matrix based on the first estimated road surface unevenness value and the international road surface unevenness index includes:

[0017] Obtain a road surface grade based on the first estimated road surface unevenness value and the international road surface unevenness index;

[0018] Convert the road surface grade to obtain an adaptive covariance matrix.

[0019] Optionally, the step of using a fusion algorithm to fuse the first estimated road surface unevenness value and the second estimated road surface unevenness value to obtain an estimated road surface unevenness value includes:

[0020] Take the first estimated road surface unevenness value within the first preset time range as the road surface unevenness estimated value;

[0021] If the current time is not within the first preset time range and the sprung acceleration estimated by the Kalman filter and the sprung acceleration meet the preset conditions, then take the second road surface unevenness value as the road surface unevenness estimated value;

[0022] If the current time is not within the first preset time range and the sprung acceleration estimated by the Kalman filter and the sprung acceleration do not meet the preset conditions, then take the first road surface unevenness value as the road surface unevenness estimated value;

[0023] If the Kalman filter does not converge within the second preset time range, then take the first estimated road surface unevenness value as the road surface unevenness estimated value;

[0024] Or, the preset condition is that when the variable value η is less than the preset variable threshold η k it meets the preset conditions;

[0025] Among them, The sprung acceleration value estimated by the Kalman filter, is the sprung acceleration.

[0026] Optionally, the obtaining of the sprung acceleration and the unsprung acceleration according to the random filtered white noise road surface and the quarter-vehicle nonlinear suspension model includes:

[0027] The quarter-vehicle nonlinear suspension model is:

[0028] Among them, z r is the random filtered white noise road surface, is the sprung acceleration, is the unsprung acceleration; m s is the mass of the sprung object, m u is the mass of the unsprung suspension, F k is the nonlinear force of the spring, F c is the nonlinear force of the suspension damper, k t is the elastic coefficient of the lower spring, z u is the unsprung displacement;

[0029] The nonlinear formula of the vehicle suspension spring is:

[0030]

[0031] Among them, is the linear damping coefficient, is the nonlinear damping coefficient, is the damper asymmetry correlation coefficient; is the speed of the sprung mass movement, is the speed of the unsprung suspension movement;

[0032] The non - linear formula for the vehicle suspension damping is:

[0033] wherein, is the linear spring stiffness coefficient, is the non - linear spring stiffness coefficient, z s is the sprung displacement.

[0034] Optionally, the inputting the sprung acceleration and the unsprung acceleration into the NARX neural network for predicting the road surface unevenness to obtain the first estimated road surface unevenness value includes:

[0035] According to the NARX neural network for the sprung acceleration, the unsprung acceleration and the historical data of the road surface unevenness, obtaining the first estimated road surface unevenness value, wherein the historical data of the road surface unevenness is obtained by inputting the historical data of the sprung acceleration and the unsprung acceleration into the NARX neural network for predicting the road surface unevenness.

[0036] In a second aspect, an embodiment of the present application provides a road surface unevenness estimation device, including:

[0037] An acceleration determination module, configured to input the obtained random filtered white noise road surface into a quarter - vehicle non - linear suspension model to generate dynamic response quantities, and obtain the sprung acceleration and the unsprung acceleration;

[0038] A first determination module, configured to input the sprung acceleration and the unsprung acceleration into the NARX neural network for predicting the road surface unevenness, and obtain the first estimated road surface unevenness value;

[0039] A second value determination module, configured to obtain a second estimated road surface unevenness value according to the sprung acceleration, the unsprung acceleration, the first estimated road surface unevenness value and the international road surface unevenness index;

[0040] A fusion module, configured to fuse the first estimated road surface unevenness value and the second estimated road surface unevenness value by using a fusion algorithm to obtain a road surface unevenness estimated value.

[0041] In a third aspect, an embodiment of the present application provides a terminal device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the road surface unevenness estimation method described in any item of the first aspect above is implemented.

[0042] Fourthly, an embodiment of the present application provides a computer-readable storage medium storing a computer program, and when the computer program is executed by a processor, the road surface unevenness estimation method described in any one of the above first aspects is implemented.

[0043] Fifthly, an embodiment of the present application provides a computer program product, and when the computer program product runs on a terminal device, the terminal device is enabled to execute the road surface unevenness estimation method described in any one of the above first aspects.

[0044] The beneficial effects of the embodiments of the present application compared with the prior art are as follows:

[0045] In the embodiments of the present application, the obtained random filtered white noise road surface is input into a quarter-vehicle nonlinear suspension model to generate dynamic response quantities, and the acceleration above the spring and the acceleration below the spring are obtained; the acceleration above the spring and the acceleration below the spring are input into a NARX neural network for predicting the road surface unevenness, and a first estimated road surface unevenness value is obtained; according to the acceleration above the spring, the acceleration below the spring, the first estimated road surface unevenness value, and the international road surface unevenness index, a second estimated road surface unevenness value is obtained; a fusion algorithm is used to fuse the first estimated road surface unevenness value and the second estimated road surface unevenness value to obtain a road surface unevenness estimation value. The fusion algorithm can effectively reduce the problem of reduced road surface unevenness estimation accuracy caused by the nonlinearity and uncertainty of the model and improve the efficiency of road surface unevenness estimation. Description of the Drawings

[0046] In order to more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the following will briefly introduce the drawings required for use in the description of the specific embodiments or the prior art. In all the drawings, similar elements or parts are generally denoted by similar reference numerals. In the drawings, the elements or parts are not necessarily drawn to scale.

[0047] Figure 1 is a schematic flowchart of the road surface unevenness estimation method provided by an embodiment of the present application;

[0048] Figure 2 is a schematic diagram of a quarter-vehicle suspension model in the road surface unevenness estimation method of the present invention;

[0049] Figure 3 is a schematic diagram of a NARX prediction network in the road surface unevenness estimation method of the present invention;

[0050] Figure 4 is a schematic flowchart of the second embodiment of the road surface unevenness estimation method of the present invention;

[0051] Figure 5 is a road surface grade classification diagram based on the international unevenness index of the present invention;

[0052] Figure 6 This is the contrast diagram of road surface grades and adaptive covariance matrices in the road surface unevenness estimation method of the present invention;

[0053] Figure 7 This is the comparative verification of the digital-analog fusion model in the road surface unevenness estimation method of the present invention Figure 1 ;

[0054] Figure 8 This is the comparative verification of the digital-analog fusion model in the road surface unevenness estimation method of the present invention Figure 2 ;

[0055] Figure 9 This is a schematic structural diagram of the road surface unevenness estimation device provided by an embodiment of the present application;

[0056] Figure 10 This is a schematic structural diagram of the terminal device provided by an embodiment of the present application. Detailed implementation manners

[0057] In the following description, specific details such as specific system architectures and technologies are presented for the purpose of illustration rather than limitation, so as to thoroughly understand the embodiments of the present application. However, those skilled in the art should clearly understand that the present application can also be implemented in other embodiments without these specific details. In other cases, detailed descriptions of well-known systems, devices, circuits, and methods are omitted to avoid unnecessary details from interfering with the description of the present application.

[0058] It should be understood that when used in the specification and appended claims of the present application, the term "comprising" indicates the presence of the described features, wholes, steps, operations, elements, and / or components, but does not exclude the presence or addition of one or more other features, wholes, steps, operations, elements, components, and / or their combinations.

[0059] It should also be understood that the term " / and / " used in the specification and appended claims of the present application refers to any combination and all possible combinations of one or more of the associated listed items, and includes these combinations.

[0060] The execution subject of the road surface unevenness estimation method provided by the embodiment of the present application can be a road surface unevenness estimation device. The obtained random filtered white noise road surface is input into a quarter-vehicle nonlinear suspension model to generate dynamic response quantities, obtaining the acceleration of the sprung mass and the acceleration of the unsprung mass. The acceleration of the sprung mass and the acceleration of the unsprung mass are input into a NARX neural network for predicting the road surface unevenness, obtaining a first estimated road surface unevenness value. According to the acceleration of the sprung mass, the acceleration of the unsprung mass, the first estimated road surface unevenness value, and the international road surface unevenness index, a second estimated road surface unevenness value is obtained. A fusion algorithm is used to fuse the first estimated road surface unevenness value and the second estimated road surface unevenness value to obtain a road surface unevenness estimation value.

[0061] Figure 1 FIG. shows a schematic flowchart of road surface unevenness estimation provided by an embodiment of the present application. By way of example and not limitation, this method can be applied to the above-mentioned road surface unevenness estimation device, or it can be a method for a user or operator to perform operations and judgments on the road surface unevenness estimation device. As Figure 1 shown, this method may include:

[0062] S10, input the obtained random filtered white noise road surface into a quarter-vehicle nonlinear suspension model to generate dynamic response quantities, obtaining the acceleration of the sprung mass and the acceleration of the unsprung mass;

[0063] The road surface unevenness estimation device obtains a random filtered white noise road surface and a quarter-vehicle nonlinear suspension model, and inputs the random filtered white noise road surface into the quarter-vehicle nonlinear suspension model to generate dynamic response quantities, obtaining the acceleration of the sprung mass and the acceleration of the unsprung mass.

[0064] Among them, as Figure 2 shown, specifically consider the nonlinearity of the suspension spring force and the nonlinearity of the damping force, and input the random filtered white noise road surface into the established quarter-vehicle nonlinear suspension model to generate dynamic response quantities: the acceleration of the sprung mass and the acceleration of the unsprung mass.

[0065] The dynamic differential equation of the quarter-suspension system is shown as follows:

[0066]

[0067] where, z r is the random filtered white noise road surface, is the acceleration of the sprung mass, is the acceleration of the unsprung mass; m s is the mass of the sprung object, m u is the mass of the unsprung suspension, F k is the nonlinear force of the spring, F c is the nonlinear force of the suspension damping, k t is the equivalent stiffness of the tire, z uis the displacement below the spring;

[0068] Considering the non - linear force of the vehicle suspension spring, the specific formula is:

[0069]

[0070] Wherein, is the linear damping coefficient, is the non - linear damping coefficient, is the damper asymmetry correlation coefficient; is the speed of the object above the spring, is the speed of the suspension movement below the spring;

[0071] Considering the non - linear force of the vehicle suspension damper, the specific formula is:

[0072]

[0073] Wherein, is the linear spring stiffness coefficient, is the non - linear spring stiffness coefficient, z s is the displacement above the spring.

[0074] Substituting the non - linear formula of the vehicle suspension spring and the non - linear formula of the vehicle suspension damper into the quarter - vehicle non - linear suspension model, the complete dynamic formula can be obtained:

[0075]

[0076] Furthermore, it can be calculated to obtain the acceleration above the spring and the acceleration below the spring.

[0077] In a possible implementation, before obtaining the dynamic response quantities such as the acceleration above the spring and the acceleration below the spring by inputting the obtained random filtered white noise road surface into the quarter - vehicle non - linear suspension model, it may include: establishing a random filtered white noise road surface model. Wherein, the specific formula of the random filtered white noise road surface model is:

[0078]

[0079] Wherein, z r (t) represents the road surface roughness signal in the time domain, v represents the vehicle speed, n c is the lower cut - off frequency of the road spatial frequency, w(t) is the unit white noise, G q (n0) is the power spectral density of the road surface unevenness at the reference spatial frequency, and n0 is the reference spatial frequency.

[0080] S20, input the sprung acceleration and unsprung acceleration into the NARX neural network for predicting road surface unevenness, and obtain the first estimated road surface unevenness value;

[0081] After determining the sprung acceleration and unsprung acceleration, the road surface unevenness estimation device inputs the sprung acceleration and unsprung acceleration into the NARX neural network for predicting road surface unevenness, and obtains the first estimated road surface unevenness value.

[0082] As an implementation manner, as Figure 3 shown, inputting the sprung acceleration and unsprung acceleration into the NARX neural network for predicting road surface unevenness and obtaining the first estimated road surface unevenness value may include: obtaining the first estimated road surface unevenness value according to the sprung acceleration, unsprung acceleration and historical road surface unevenness data of the suspension by the NARX neural network, wherein the historical road surface unevenness data is obtained by inputting the historical data of the sprung acceleration and unsprung acceleration into the NARX neural network for predicting road surface unevenness.

[0083] Specifically, after determining the sprung acceleration and unsprung acceleration, the road surface unevenness estimation device obtains the first estimated road surface unevenness value according to the sprung acceleration, unsprung acceleration and historical road surface unevenness data of the suspension by the NARX neural network, wherein the historical road surface unevenness data is obtained by inputting the historical data of the sprung acceleration and unsprung acceleration into the NARX neural network for predicting road surface unevenness.

[0084] When predicting through the NARX neural network, refer to Figure 3 shown, specifically as follows:

[0085] Set the number of hidden layers:

[0086] where m is the number of neurons in the output layer, n is the number of input neurons, and a is a constant from 0 to 10. /

[0087] The hidden layer function formula is: f(x) = 1 / (1 + e -x );

[0088] The output layer function formula is: g(x) = x;

[0089] The weights and thresholds of the neural network are adjusted by the training algorithm, and here the trainlm algorithm is taken as the training algorithm.

[0090] Using the sprung acceleration and unsprung acceleration to predict the road surface unevenness, its core is to predict the current output through the past output and past external input, and the basic formula is

[0091] y(t) = f(y(t - 1), y(t - 2),..., y(t - n y ), u(t - 1), u(t - 2), …, y(t - n u ))

[0092] where y(t) is the output, i.e., the road surface unevenness, and u(t) is the input, i.e., the acceleration above the spring and the acceleration below the spring. n y and n u are the delay orders respectively.

[0093] The NARX neural network usually includes an input layer, a hidden layer and an output layer. The specific structure is as shown in the original text Figure 3 as follows: Input layer: Receives the delayed road surface unevenness and acceleration data. Hidden layer: Processes the input using the non-linear activation function Sigmoid. Its activation function formula is Finally, the output layer: Outputs the predicted value of the road surface unevenness.

[0094] Assume that the hidden layer has m neurons and the activation function is f(x). Then the output result of the hidden layer is expressed as:

[0095]

[0096] where is the first weight coefficient, is the second weight coefficient, and b h is the bias.

[0097] The output layer calculates the output road surface unevenness y(t):

[0098]

[0099] where is the weight from the hidden layer to the output layer, and b r is the bias.

[0100] Through the NARX neural network, the acceleration data above and below the spring of the suspension can be used as inputs to predict the road surface unevenness data. The model realizes the prediction through the non-linear mapping of historical data and current inputs.

[0101] S30. Obtain a second estimated road surface unevenness value according to the acceleration above the spring, the acceleration below the spring, the first estimated road surface unevenness value and the international road surface unevenness index;

[0102] After determining the acceleration above the spring, the acceleration below the spring, the first estimated road surface unevenness value and the international road surface unevenness index, the road surface unevenness estimation device obtains a second estimated road surface unevenness value according to the acceleration above the spring, the acceleration below the spring, the first estimated road surface unevenness value and the international road surface unevenness index.

[0103] Further, referring to Figure 4 , Figure 4 , which is a schematic flowchart of the second embodiment of the road surface unevenness estimation method of the present invention. Based on the above Figure 4 illustrated embodiment, according to the sprung acceleration, unsprung acceleration, first estimated road surface unevenness value, and international road surface unevenness index, a second estimated road surface unevenness value is obtained, specifically including:

[0104] S31 Obtain an adaptive covariance matrix according to the first estimated road surface unevenness value and the international road surface unevenness index;

[0105] After the road surface unevenness estimation device determines the first estimated road surface unevenness value, an adaptive covariance matrix is obtained according to the first estimated road surface unevenness value and the international road surface unevenness index;

[0106] As an implementation manner, obtaining an adaptive covariance matrix according to the first estimated road surface unevenness value and the international road surface unevenness index may include: obtaining a road surface grade according to the first estimated road surface unevenness value and the international road surface unevenness index; converting the road surface grade to obtain an adaptive covariance matrix.

[0107] In this embodiment, after the road surface unevenness estimation device determines the first estimated road surface unevenness value, the road surface unevenness grade is calculated through the international road surface unevenness index.

[0108] Where can be approximated as the road surface unevenness value, and the IRI international road surface unevenness index can be calculated by the following formula: Where L is the distance traveled by the vehicle, and t is the time period during which the vehicle travels.

[0109] After obtaining the IRI international road surface unevenness index, the road surface unevenness grade can be determined by querying the Figure 5 table. The G q (n0) value is used for road modeling, and the calculated IRI value needs to obtain the road surface grade according to the look-up table.

[0110] In this embodiment, referring to Figure 6 shown, after the road surface unevenness estimation device determines the road surface grade, the table shown in Figure 6 can be queried to convert the road surface grade to obtain an adaptive covariance matrix. The adaptive covariance matrix is a matrix composed of z s , zu, and .

[0111] For the update of the adaptive noise covariance matrix related to the self-adaptive square root cubature Kalman, the following formula is used for the update: Qk = γQ k-1 ;

[0112] where γ is the adaptive noise covariance factor, and Q k is the system noise covariance matrix at the current time k, and Q k-1 is the system noise covariance matrix at the previous time k−1. The value of γ is an update parameter, which is expressed as Q k / Q k-1 , that is, the ratio of the Q value at this moment to the Q value at the previous moment. The Q value at each moment can be obtained by looking up the table according to the road surface grade obtained in the previous step.

[0113] S32 inputs the sprung mass acceleration, unsprung mass acceleration, and adaptive covariance matrix into a Kalman filter to obtain a second estimated road surface roughness value.

[0114] After obtaining the sprung mass acceleration, unsprung mass acceleration, and adaptive covariance matrix, the road surface roughness estimation device inputs the sprung mass acceleration, unsprung mass acceleration, and adaptive covariance matrix into a Kalman filter to obtain a second estimated road surface roughness value.

[0115] As an implementation, inputting the sprung mass acceleration, unsprung mass acceleration, and adaptive covariance matrix into a Kalman filter to obtain a second estimated road surface roughness value may include: inputting the sprung mass acceleration and unsprung mass acceleration into the state space equation of a nonlinear system, and combining with the adaptive covariance matrix to perform state update of the Kalman filter to obtain a state space predicted value and a square root factor of the error covariance matrix; according to the sprung mass acceleration, unsprung mass acceleration, state space predicted value, and square root factor of the error covariance matrix, perform state variable update of the Kalman filter to obtain a second estimated road surface roughness value.

[0116] Based on the model, use square root cubature Kalman filter to estimate road surface roughness and update it using the latest process noise covariance matrix. The specific steps are as follows:

[0117] Establish the state space equation of the nonlinear system:

[0118]

[0119] where x(t) is the state variable, represents the process noise, θ(t) is zero-mean Gaussian noise, v represents the vehicle speed, and n c is the lower cut-off frequency of the road spatial frequency, w(t) is unit white noise, G q (n0) is the power spectral density of road surface unevenness at the reference spatial frequency, z(t) is the observed variable, and V(t) = [v1(t), v2(t)] TFor measuring noise, v i (t) is zero-mean Gaussian noise, and the state variables are selected as which is the derivative of the state variable, and the observation variable is selected as

[0120] Discretize the continuous state-space equation to obtain the following formula:

[0121]

[0122]

[0123] where x k is the state variable at time k, z k is the observation variable at time k, W k-1 is the process noise at time k - 1, V k is the measurement noise at time k, F k,k-1 is the spring force at the previous time, F c,k-1 is the damping force at the previous time, which can be expressed as:

[0124]

[0125] where, is the linear coefficient of spring stiffness, is the non-linear coefficient of spring stiffness, is the linear damping coefficient, is the non-linear damping coefficient, is the coefficient related to the asymmetry of the damper.

[0126] In the discrete non-linear equation, the process noise W k-1 and the measurement noise V k are zero-mean Gaussian white noise, and their covariance matrices can be expressed as:

[0127]

[0128] where, R k is the positive definite measurement covariance matrix, and δ kj is the Kronecker delta function. Q k-1 is the adaptive covariance matrix at the previous time, is the transpose of the process noise matrix at time j - 1, is the transpose of the observation noise matrix at time j.

[0129] Thus, the square root cubature Kalman equations can be summarized as

[0130] (1) Initialization:

[0131] Assume the initial state and the relevant square root factor of the corresponding error covariance matrix is S 0|0 , S 0|0 has the following expression:

[0132] S 0|0 = [Chol(P 0|0 )] T ;

[0133] where Chol is the Cholesky decomposition and P 0|0 is the error covariance matrix.

[0134] (2) State update

[0135] Calculate the cubature points according to the state transition equation and propagate them. The specific formula is as follows:

[0136]

[0137] where, the value of the i-th state cubature point, is the state estimate value at the previous moment, is the state of the i-th cubature point at the previous moment after passing through the state transition matrix , S k-1|k-1 is the square root factor of the error covariance at the previous moment, ξ i is the i-th column of the weight matrix, and the weight matrix is where I n is an n×n matrix.

[0138] where, according to the state transition equation formula, 2n can be calculated. Bring these 2n points into whose expression is Thus, 2n

[0139] Calculate the predicted value of the state space and the square root factor of its error covariance matrix. The specific formula is as follows:

[0140]

[0141] is the predicted value of the state at time k, S kk-1 is the square root factor of the error covariance matrix, Q k-1 is the system error covariance matrix at the previous moment, weighted central moment, and can be defined as:

[0142]

[0143] (3) Measurement update

[0144] Update the volume points according to the measurement equation and propagate them. The specific formula is as follows:

[0145]

[0146] where, is the predicted value of the i-th state volume point at time k, and ξ i is the i-th column of the weight matrix, and the weight matrix is where I n is an n×n matrix. According to the measurement equation formula, 2n predicted observation variables can be calculated Bring these 2n points into respectively. Its expression is Thus, 2n

[0147] Calculate the predicted value of the measurement and the square root factor of its measurement error covariance matrix. The specific formula is as follows:

[0148]

[0149] where, is the predicted value of the observation variable, is the square root factor of the measurement error covariance matrix. Similarly, R k is the measurement noise covariance matrix at this time, and the weighted central moment Z kk-1 can be defined as:

[0150]

[0151] Calculate the measurement covariance matrix and the cross-covariance matrix

[0152]

[0153] The weighted central moment X kk-1 is defined as:

[0154]

[0155] Calculate the Kalman gain matrix K k :

[0156]

[0157] where, is the measurement covariance matrix, is the cross-covariance matrix.

[0158] Finally, update the state variable and the square root factor S of the error covariance matrix kk :

[0159]

[0160] where is the predicted value of the state at time k given the state at time k-1, K k is the Kalman gain matrix at time k, z k is the true observed value at time k, is the predicted value of the observed variable at time k given the state at time k-1.

[0161] Thus, through the adaptive covariance matrix at the previous time, the second estimated road surface unevenness value and S k|k process noise covariance matrix can be obtained.

[0162] When performing the square root cubature Kalman equation at the next time, k = k + 1, using the current time's Q k adaptive covariance matrix, calculate the square root factor S of the error covariance matrix at the next time k|k-1 .

[0163] S40. Use a fusion algorithm to fuse the first estimated road surface unevenness value and the second estimated road surface unevenness value to obtain an estimated road surface unevenness value.

[0164] After determining the first estimated road surface unevenness value and the second estimated road surface unevenness value, the road surface unevenness estimation device uses a fusion algorithm to fuse the first estimated road surface unevenness value and the second estimated road surface unevenness value to obtain an estimated road surface unevenness value.

[0165] As an implementation, using a fusion algorithm to fuse the first estimated road surface unevenness value and the second estimated road surface unevenness value to obtain an estimated road surface unevenness value may include: taking the first estimated road surface unevenness value within the first preset time range as the estimated road surface unevenness value; if the current time is not within the first preset time range and the sprung acceleration estimated by the Kalman filter satisfies a preset condition with the sprung acceleration, then taking the second road surface unevenness value as the estimated road surface unevenness value; if the current time is not within the first preset time range and the sprung acceleration estimated by the Kalman filter does not satisfy the preset condition with the sprung acceleration, then taking the first road surface unevenness value as the estimated road surface unevenness value; if the Kalman filter does not converge within the second preset time range, then taking the first estimated road surface unevenness value as the estimated road surface unevenness value;

[0166] Among them, the preset condition is that when the variable value η is less than the preset variable threshold η k the preset condition is satisfied;

[0167] Among them, the value of the sprung mass acceleration estimated by the Kalman filter, is obtained by differentiation, and is the sprung mass acceleration. η k can be 0.1 or 0.2.

[0168] Specifically, to solve the problem that the adaptive cubature Kalman filter (PASCKF) has a slow convergence speed at the initial moment, within the first 5 seconds, the NARX neural network is used for estimation. At the same time, it also outputs the process noise correction for the next time step to the PASCKF estimator, thereby improving the estimation accuracy and convergence speed of the PASCKF estimator.

[0169] (2) To solve the accuracy problem of the PASCKF estimator, a variable η is introduced, and its definition is:

[0170]

[0171] Among them, is the value of the sprung mass acceleration estimated by PASCKF, represents the sprung mass acceleration obtained from the sensor. Set the threshold η k = 0.1. If η < η k , it indicates that the value calculated by PASCKF is accurate. In this case, the value from PASCKF will be directly output. If η ≥ η k , it indicates that the value calculated by PASCKF is inaccurate. In this case, the value from NARX will be directly output.

[0172] (3) Considering the problem that PASCKF cannot converge and correctly identify the road surface, if PASCKF cannot converge within 10 seconds and NARX can stably identify the road surface, then the estimator will directly output the value of NARX.

[0173] As Figure 8 and Figure 9 shown, it is a comparison and verification diagram of the road surface unevenness estimation value and the random filtered white noise road surface. It can be seen from the figure that the algorithm proposed in this paper can effectively estimate the road surface unevenness and has a significant improvement compared with the original SCKF algorithm.

[0174] In summary, the obtained random filtered white noise road surface is input into the quarter-vehicle nonlinear suspension model to generate dynamic response quantities, obtaining the acceleration above the spring and the acceleration below the spring; the acceleration above the spring and the acceleration below the spring are input into the NARX neural network for predicting the road surface unevenness, obtaining the first estimated road surface unevenness value; according to the acceleration above the spring, the acceleration below the spring, the first estimated road surface unevenness value and the international road surface unevenness index, a second estimated road surface unevenness value is obtained; a fusion algorithm is used to fuse the first estimated road surface unevenness value and the second estimated road surface unevenness value to obtain the road surface unevenness estimated value. It can effectively reduce the problem of reduced accuracy of road surface unevenness estimation caused by the nonlinearity and uncertainty of the model, and improve the efficiency of road surface unevenness estimation. Applying the square root cubature Kalman technique to the problem of road surface unevenness estimation of nonlinear suspensions can effectively estimate the road surface unevenness under the condition that the suspension parameters are nonlinear. Incorporating the road surface unevenness predicted by the NARX neural network into the square root cubature Kalman filter estimation framework improves the estimation accuracy of the road surface unevenness and prevents the problem of non-convergence of the adaptive cubature Kalman during the road change process.

[0175] Consistent with the above, please refer to Figure 9 , Figure 9 which is a schematic structural diagram of a road surface unevenness estimation device provided by an embodiment of the present application. As Figure 9 shown, the device includes:

[0176] An acceleration determination module 901, configured to input the obtained random filtered white noise road surface into the quarter-vehicle nonlinear suspension model to generate dynamic response quantities, obtaining the acceleration above the spring and the acceleration below the spring;

[0177] A first determination module 902, configured to input the acceleration above the spring and the acceleration below the spring into the NARX neural network for predicting the road surface unevenness, obtaining the first estimated road surface unevenness value;

[0178] A second determination module 903, configured to obtain a second estimated road surface unevenness value according to the acceleration above the spring, the acceleration below the spring, the first estimated road surface unevenness value and the international road surface unevenness index;

[0179] A fusion module 904, configured to use a fusion algorithm to fuse the first estimated road surface unevenness value and the second estimated road surface unevenness value to obtain the road surface unevenness estimated value.

[0180] An embodiment of the present application further provides a terminal device, which includes: at least one processor, a memory, and a computer program stored in the memory and executable on the at least one processor, and when the processor executes the computer program, the steps in the embodiment of the road surface unevenness estimation method are implemented.

[0181] An embodiment of the present application also provides a computer storage medium. The computer storage medium stores a computer program for electronic data exchange, and the computer program enables a computer to execute some or all of the steps of any one of the road surface unevenness estimation methods described in the above method embodiments.

[0182] An embodiment of the present application also provides a computer program product. The computer program product includes a non-transitory computer-readable storage medium storing a computer program, and the computer program enables a computer to execute some or all of the steps of any one of the road surface unevenness estimation methods described in the above method embodiments.

[0183] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, to implement all or part of the processes in the above method embodiments of the present application, a computer program can be used to instruct relevant hardware to complete. The computer program can be stored in a computer-readable storage medium. When the computer program is executed by a processor, the steps of the above method embodiments can be implemented. Among them, the computer program includes computer program code, and the computer program code can be in the form of source code, object code, executable file, or some intermediate form, etc. The computer-readable storage medium can at least include: any entity or device that can carry the computer program code to the device / terminal device, recording medium, computer memory, read-only memory (ROM), random access memory (RAM), electrical carrier signal, telecommunication signal, and software distribution medium. For example, a USB flash drive, a mobile hard disk, a magnetic disk, or an optical disc, etc. In some jurisdictions, according to legislation and patent practice, the computer-readable storage medium may not be an electrical carrier signal and a telecommunication signal.

[0184] In the above embodiments, the descriptions of the various embodiments have their own emphases. For the parts not detailed or recorded in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.

[0185] Those of ordinary skill in the art can realize that the units and algorithm steps of the examples described in combination with the embodiments disclosed herein can be implemented by electronic hardware, or by a combination of computer software and electronic hardware. Whether these functions are executed in a hardware or software manner depends on the specific application and design constraints of the technical solution. Professional technicians can use different methods to implement the described functions for each specific application, but such implementation should not be considered to exceed the scope of the present application.

[0186] In the embodiments provided in the present application, it should be understood that the disclosed device / terminal device and method can be implemented in other ways. For example, the device / terminal device embodiments described above are merely illustrative. For example, the division of the modules or units is only a logical function division. In actual implementation, there may be other division methods. For example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the displayed or discussed coupling or direct coupling or communication connection between each other can be through some interfaces. The indirect coupling or communication connection of the device or unit can be in electrical, mechanical or other forms.

[0187] The units described as separate components may or may not be physically separated. The components displayed as units may or may not be physical units, that is, they may be located in one place or distributed to multiple network units. Some or all of the units can be selected according to actual needs to achieve the purpose of the solution of this embodiment.

Claims

1. A method for estimating road surface unevenness, characterized in that Including: The obtained random filtered white noise road surface is input into a quarter-vehicle nonlinear suspension model to generate dynamic response quantities, obtaining the acceleration above the spring and the acceleration below the spring; The acceleration above the spring and the acceleration below the spring are input into a NARX neural network for predicting road surface unevenness, obtaining a first estimated road surface unevenness value; According to the acceleration above the spring, the acceleration below the spring, the first estimated road surface unevenness value, and the international road surface unevenness index, a second estimated road surface unevenness value is obtained; A fusion algorithm is used to fuse the first estimated road surface unevenness value and the second estimated road surface unevenness value to obtain an estimated road surface unevenness value.

2. The road surface unevenness estimation method according to claim 1, wherein The step of obtaining the second estimated road surface unevenness value according to the acceleration above the spring, the acceleration below the spring, the first estimated road surface unevenness value, and the international road surface unevenness index includes: According to the first estimated road surface unevenness value and the international road surface unevenness index, an adaptive covariance matrix is obtained; The acceleration above the spring, the acceleration below the spring, and the adaptive covariance matrix are input into a Kalman filter to obtain the second estimated road surface unevenness value.

3. The road surface unevenness estimation method according to claim 2, characterized in that, The step of inputting the acceleration above the spring, the acceleration below the spring, and the adaptive covariance matrix into the Kalman filter to obtain the second estimated road surface unevenness value includes: The acceleration above the spring and the acceleration below the spring are input into the state space equation of a nonlinear system, and combined with the adaptive covariance matrix, the state of the Kalman filter is updated to obtain a state space predicted value and a square root factor of the error covariance matrix; According to the acceleration above the spring, the acceleration below the spring, the state space predicted value, and the square root factor of the error covariance matrix, the state variables of the Kalman filter are updated to obtain the second estimated road surface unevenness value.

4. The road surface unevenness estimation method according to claim 2, wherein The step of obtaining the adaptive covariance matrix according to the first estimated road surface unevenness value and the international road surface unevenness index includes: According to the first estimated road surface unevenness value and the international road surface unevenness index, a road surface grade is obtained; The road surface grade is transformed to obtain the adaptive covariance matrix.

5. The road surface unevenness estimation method according to any one of claims 1 to 4, characterized in that The step of using a fusion algorithm to fuse the first estimated road surface unevenness value and the second estimated road surface unevenness value to obtain an estimated road surface unevenness value includes: Taking the first estimated road surface unevenness value within a first preset time range as the estimated road surface unevenness value; If the current time is not within the first preset time range and the acceleration above the spring estimated by the Kalman filter satisfies a preset condition with the acceleration above the spring, then taking the second road surface unevenness value as the estimated road surface unevenness value; If the current time is not within the first preset time range and the acceleration above the spring estimated by the Kalman filter does not satisfy the preset condition with the acceleration above the spring, then taking the first road surface unevenness value as the estimated road surface unevenness value; If the Kalman filter does not converge within a second preset time range, then taking the first estimated road surface unevenness value as the estimated road surface unevenness value; Or, the preset condition is that when the variable value η is less than the preset variable threshold η k it meets the preset condition; Among them, is the value of the acceleration above the spring estimated by the Kalman filter, is the acceleration above the spring.

6. The road surface unevenness estimation method according to any one of claims 1 to 4, characterized in that The step of obtaining the acceleration above the spring and the acceleration below the spring according to the random filtered white noise road surface and the quarter-vehicle nonlinear suspension model includes: The quarter-vehicle nonlinear suspension model is as follows: where z r is the random filtered white noise road surface, is the acceleration above the spring, is the acceleration below the spring; m s is the mass of the object above the spring, m u is the mass of the suspension below the spring, F k is the non-linear force of the spring, F c is the non-linear force of the suspension damper, k t is the elastic coefficient of the lower spring, and zu is the displacement below the spring; The nonlinear formula of the vehicle suspension spring is: Among them, is the linear damping coefficient, is the non-linear damping coefficient, is the damper asymmetry correlation coefficient; is the speed of the sprung mass movement, is the speed of the unsprung suspension movement; The non-linear formula for vehicle suspension damping is as follows: Among them, is the linear spring stiffness coefficient, is the non-linear spring stiffness coefficient, and zs is the displacement above the spring.

7. The method for estimating road roughness according to any one of claims 1 to 4, characterized in that: The step of inputting the acceleration above the spring and the acceleration below the spring into a NARX neural network for predicting road surface unevenness, obtaining a first estimated road surface unevenness value, includes: According to the NARX neural network, the first estimated road surface unevenness value is obtained from the historical data of the sprung acceleration, unsprung acceleration, and road surface unevenness. The historical data of the road surface unevenness is obtained by inputting the historical data of the sprung acceleration and unsprung acceleration into the NARX neural network for road surface unevenness prediction.

8. A road surface unevenness estimation device, characterized in that, It includes: An acceleration determination module, configured to input the obtained random filtered white noise road surface into a quarter-vehicle nonlinear suspension model to generate dynamic response quantities, and obtain the sprung acceleration and unsprung acceleration; A first determination module, configured to input the sprung acceleration and unsprung acceleration into the NARX neural network for road surface unevenness prediction, and obtain a first estimated road surface unevenness value; A second determination module, configured to obtain a second estimated road surface unevenness value according to the sprung acceleration, unsprung acceleration, first estimated road surface unevenness value, and international road surface unevenness index; A fusion module, configured to fuse the first estimated road surface unevenness value and the second estimated road surface unevenness value by using a fusion algorithm to obtain an estimated road surface unevenness value.

9. A terminal device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the road surface unevenness estimation method according to any one of claims 1 to 7.

10. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by the processor, it implements the road surface unevenness estimation method according to any one of claims 1 to 7.

Images