A road adhesion coefficient estimation method based on a dynamics model and support vector regression

Abstract

This invention discloses a method for estimating the road adhesion coefficient based on a dynamic model and support vector regression, relating to the field of vehicle parameter estimation and control. The method includes the following steps: S1, establishing a training data acquisition module to collect data on the vehicle's four-wheel wheel speed, vehicle speed, longitudinal acceleration, lateral velocity, lateral acceleration, yaw rate, steering wheel angle, tire lateral force, tire longitudinal force, and road adhesion coefficient; S2, using the features in the training dataset formed in S1 as input to the support vector regression (SVR) to obtain a pseudo-measured value of the road adhesion coefficient; S3, establishing a three-degree-of-freedom vehicle dynamics model and a Dugoff tire model, and designing an AEKF algorithm model; S4, using the pseudo-measured value of the road adhesion coefficient obtained in step S2 as an extension of the measurement vector in step S3 of the AEKF algorithm to achieve optimal road adhesion coefficient estimation. Compared with existing technologies, this invention comprehensively considers the model accuracy problem and the nonlinear relationship capture capability of support vector regression, and the estimation method that integrates the two can simultaneously possess robustness and accuracy.

Description

Technical Field

[0001] This invention relates to a method for estimating the road surface adhesion coefficient based on a dynamic model and support vector regression, belonging to the field of vehicle parameter estimation and control, and particularly to a fusion method for estimating the vehicle road surface adhesion coefficient. Background Technology

[0002] Intelligent vehicles, as a core component of intelligent transportation systems, are intelligent machines capable of autonomous driving, intelligent navigation, environmental perception, and decision-making and planning by utilizing computer science, artificial intelligence, and sensor technology. [1] The movement of intelligent vehicles is a comprehensive manifestation of the interaction between the vehicle and the road surface. [2] The coefficient of friction (COP), as one of the indicators measuring the maximum force between the tire and the road surface, plays an indispensable role in the vehicle control system due to its accurate estimation, which affects the stable and safe driving of the vehicle. The essence of COP estimation is to obtain the COP directly or indirectly by observing the changes in the response of parameters closely related to the COP, such as tire force, tire slip ratio, and tire side slip angle, to different road surfaces.

[0003] Currently, in the research field of vehicle road adhesion coefficient estimation, estimation methods are mainly divided into three categories: the first is experiment-based methods, the second is model-based methods, and the third is machine learning-based methods. [3] The experimental approach obtains sensor data through various onboard sensors. While this method can provide relatively accurate data, it is costly, and sensors are prone to hysteresis and failure, making it difficult to guarantee the real-time requirements of intelligent vehicles. [4] Model-based methods estimate the road adhesion coefficient by combining a simplified vehicle dynamics model with an estimation algorithm. This method relies on accurate vehicle dynamics and tire models. However, vehicles are highly nonlinear systems, and current estimation algorithms have significant limitations in handling nonlinear relationships, resulting in inaccuracies in estimation accuracy. [5] Machine learning-based methods utilize data-driven models to directly learn the nonlinear mapping relationship between sensor signal input and pavement adhesion coefficient output. Purely data-driven methods rely on the establishment of high-quality, full-condition datasets; the lack of sufficient training data will prevent these methods from effectively obtaining the pavement adhesion coefficient. [6] .

[0004] In existing technologies, the road surface adhesion coefficient is difficult to measure directly or requires expensive sensors. [7] Modern vehicles' CAN bus can provide parameters such as steering wheel angle, wheel speed, vehicle speed, acceleration, and yaw rate.

[0005] Therefore, how to accurately estimate the road surface adhesion coefficient using existing sensors on intelligent vehicles to improve vehicle driving safety is a key issue that urgently needs to be addressed in the upgrade of intelligent driving.

[0006] References for this application: [1]Kumar N, Rahman SS, Dhakad N. Fuzzy Inference Enabled DeepReinforcement Learning-Based Traffic Light Control for IntelligentTransportation System[J]. IEEE Transactions on Intelligent TransportationSystems, 2020, 22(8): 4919-4928. [2]Guo J, He C, Li J, et al. Slope Estimation Method of ElectricVehicles Based on Improved Sage-Husa Adaptive Kalman Filter[J]. Energies, 2022, 15(11): 4126. [3]Khaleghian S, Emami A, Taheri S. A Technical Survey on Tire-RoadFriction Estimation[J]. Friction, 2017, 5: 123-146. [4]Du Y, Liu C, Song Y, et al. Rapid Estimation of Road Friction for Anti-Skid Autonomous Driving[J]. IEEE Transactions on IntelligentTransportation Systems, 2019, 21(6): 2461-2470. [5] Wu, Yufan. Estimation of vehicle state and road adhesion coefficient of four-wheel independent drive electric vehicle [D]. Jilin University, 2020. [6] Wang Juncheng, Wang Fahui. Estimation of road surface adhesion coefficient based on variable weight PSO-Elman neural network [J]. Journal of Zhejiang University (Engineering Science), 2024, 58(03): 622-634. [7] Wang Lianbing. A method for estimating the state of intelligent vehicles and road adhesion coefficient by multi-information fusion [D]. Jilin: Jilin University, 2024. Summary of the Invention

[0007] The purpose of this invention is to overcome or at least partially solve the above problems by providing a method for estimating the road surface adhesion coefficient of intelligent vehicles based on a physical model and data-driven approach, which addresses the issues of high cost, inaccurate accuracy, and insufficient generalization ability of existing road surface adhesion coefficient estimation methods.

[0008] Terminology Explanation: 1. SVR: Support Vector Regression (SVR) 2. AEKF: Adaptive Extended Kalman Filter 3. RBF: Radial Basic Function 4. GWO: Grey Wolf Optimizer 5. MSE: Mean Squared Error 6. EKF: Extended Kalman Filter Step S1: Establish a training data acquisition module to collect the vehicle's four-wheel wheel speed, vehicle speed, longitudinal acceleration, lateral velocity, lateral acceleration, yaw rate, steering wheel angle, tire lateral force, tire longitudinal force, and road adhesion coefficient. Step S2: Use the features in the training dataset formed in S1 as input to SVR to obtain the pseudo-measured value of the road adhesion coefficient. Step S3: Establish a three-degree-of-freedom vehicle dynamics model and a Dugoff tire model, and design the AEKF algorithm model; The pseudo-measured values ​​of the road adhesion coefficient obtained in steps S4 and S2 are used as an extension of the measurement vector of the AEKF algorithm in step S3 to achieve the optimal road adhesion coefficient estimation.

[0009] Furthermore, step S1 specifically includes: Step S11: Based on the vehicle's longitudinal driving conditions, set a sinusoidal sweep frequency signal and design three target vehicle speed control input curves for acceleration, deceleration, and constant speed. Step S12: Based on the vehicle's lateral driving conditions, set different frequencies and amplitudes of steering wheel angle control input curves for dual lane change input, sinusoidal steering input, and steady-state rotation input.

[0010] Step S13: Select three different road surface types, and execute S11 and S12 under each road surface condition.

[0011] Step S14: Arrange the above target vehicle speed, steering wheel angle, and target adhesion coefficient to make the training dataset have high coverage of working conditions.

[0012] Furthermore, step S2 specifically includes: Step S21: The training dataset is determined by the four-wheel wheel speed, vehicle speed, longitudinal acceleration, lateral velocity, lateral acceleration, yaw rate, steering wheel angle, tire lateral force, tire longitudinal force, and road adhesion coefficient; the SVR input vector expression is as follows:

[0013] In the formula, For four-wheel wheel speed; For vehicle speed; It is longitudinal acceleration; Lateral velocity; It is lateral acceleration; This refers to the yaw rate; Steering wheel angle; This refers to the lateral force of the tire; This refers to the longitudinal force of the tire; where, , respectively representing left front, right front, left back, and right back.

[0014] Step S22: The SVR model is a multiple input multiple output model, and the RBF kernel function is selected. Its hyperparameters are autonomously optimized through GWO.

[0015] Furthermore, step S22 specifically includes the following steps: (1) The basic SVR model is based on a given dataset. , , .in, For the first Input features of each sample express It is 3D vector Indicates the feature dimension; For the first The target value for each sample It is the space of real numbers; This represents the total number of samples in the dataset.

[0016] The standard SVR model expression is:

[0017] The objective function is:

[0018] In the formula, The output predicted by the model; The input feature vector; This is the weight vector; For bias terms; For regularization terms, minimize Norms are used to control model complexity; For error terms, For regularization parameters; for Insensitive loss function Tolerance for error.

[0019]

[0020] (2) Introducing slack variables and The objective function for optimizing the SVR model is:

[0021] The error limit is:

[0022] (3) Introducing Lagrange multipliers and The objective function for optimizing the SVR model is:

[0023] (4) Introduce the RBF kernel function:

[0024] In the formula, Center of the kernel function; The width of the kernel function; for arrive The Euclidean distance between them.

[0025] (5) The SVR expression combining the RBF kernel function is:

[0026] Step S23: Further, the SVR model uses the GWO algorithm to adjust the regularization parameters. Error tolerance Performing a minimum hyperparameter search avoids getting trapped in local optima, improving the generalization ability and prediction performance of SVR; using MSE as the fitness function of the GWO algorithm; minimizing the MSE corresponding to the fitness function improves the SVR model. and The optimization process for SVR is then carried out.

[0027] Furthermore, the optimization process of the GWO algorithm includes: setting the initial parameters in the algorithm, iterating sequentially, and reaching the maximum number of iterations or the algorithm stabilizing indicates that the optimal hyperparameters have been found.

[0028] Furthermore, the MSE expression as the fitness function is:

[0029] In the formula, This is the actual value. These are predicted values.

[0030] Furthermore, step S3 specifically includes the following steps: Step S31, the detailed expression of the three-degree-of-freedom vehicle dynamics model is as follows: ; Step S32, the relevant expressions for the Dugoff tire model are as follows:

[0031]

[0032]

[0033]

[0034]

[0035]

[0036]

[0037]

[0038] In the formula, For the overall vehicle weight; To bypass Moment of inertia of the shaft; This is the distance from the center of gravity to the front axle; This is the distance from the center of mass to the rear axle; , These are the half track widths of the front and rear axles, respectively. This refers to the driving torque of the wheels. The moment of inertia of the wheel; The angular velocity of the wheel. The effective rolling radius of the wheel; The coefficient of friction for each wheel on the road surface; The sideslip angle of each wheel; For longitudinal stiffness; Lateral stiffness; The vertical force on each tire; The slip ratio of each wheel; Vertical forces on the left and right front tires; Vertical forces on the left and right rear tires; Let be the acceleration due to gravity, taken as 9.8; The height of the vehicle's center of gravity above the ground; It is a normalization function used to ensure that the tire force increases linearly under low slip and reaches the friction limit under high slip, which helps to simulate the dynamic behavior of the tire under different road conditions. It is a dimensionless parameter; The speed-affecting factor was determined through testing. The effective rolling radius of the tire; The center speed of each wheel.

[0039] Step S33: The AEKF algorithm model is based on the EKF algorithm, dynamically adjusting the process noise covariance matrix and the measurement noise covariance matrix. The road adhesion coefficient estimation model used in this invention is a nonlinear model. Further, step S33 specifically includes the following steps: (1) Consider the following nonlinear system model:

[0040] In the formula, for The state vector of the system at any given time; for The control input vector at each time step; for Measurement vectors of the time-series system; and These are the state transition function and the measurement function, respectively. and For independent Gaussian white noise, where , , The process noise covariance matrix is... This is the measurement noise covariance matrix.

[0041] (2) Specific implementation process of the AEKF algorithm: 1) In At that time, set the initial state estimate. Initial state estimation covariance matrix Process noise covariance matrix Measurement noise covariance matrix ; 2) Change the state transition function exist Perform a first-order Taylor expansion at this point:

[0042] Let the state transition Jacobian matrix be... State function exist The first-order Taylor expansion at point is expressed as:

[0043] Therefore, the state vector equation in the nonlinear system model can be expressed as:

[0044] The expressions for the state prediction value and the state error covariance are as follows:

[0045]

[0046] 3) Measurement function exist Perform a first-order Taylor expansion at this point:

[0047] Let the measurement Jacobian matrix Measurement function exist The first-order Taylor expansion at point is expressed as:

[0048] Therefore, the measurement equations in the nonlinear system model can be expressed as:

[0049] The expressions for the measurement prediction value and the measurement prediction error covariance are as follows:

[0050]

[0051] 4) Covariance:

[0052] 5) Update state estimation and state estimation error covariance :

[0053]

[0054] 6) Calculate the Kalman gain:

[0055] 7) Adaptive process noise covariance matrix Measurement noise covariance matrix Adjustment: Innovation (the difference between the actual measured state and the predicted state) sequence:

[0056] Variance of the innovation sequence:

[0057] Dynamic adjustment process noise covariance matrix :

[0058] Dynamically adjust the measurement noise covariance matrix :

[0059] In the formula, , This is the adaptive gain factor.

[0060] Furthermore, step S4 specifically includes: using the road surface adhesion coefficient obtained by training the model through SVR as a pseudo-measurement value, which is used as part of the measurement vector in AEKF to participate in the filtering process; the adaptive mechanism of the AEKF algorithm dynamically adjusts the noise covariance matrix to optimize the estimation of the road surface adhesion coefficient.

[0061] Furthermore, step S4 specifically includes the following steps: Step S41: Obtain the road surface adhesion coefficient as a pseudo-measurement value by training the model using SVR; Step S42: The pseudo-measured value is used as part of the AEKF input to participate in the filtering process; the adaptive mechanism of the AEKF algorithm dynamically adjusts the noise covariance matrix to estimate the road surface adhesion coefficient.

[0062] The beneficial effects of this invention are: 1. This invention combines the efficient nonlinear modeling of SVR with the noise adaptive mechanism of AEKF to improve the accuracy of road adhesion coefficient estimation; 2. This invention utilizes GWO to optimize SVR hyperparameters, which can improve the robustness and generalization of the SVR model; 3. The method proposed in this invention can be applied to intelligent vehicle control modules through model and algorithm programming, providing a new solution for intelligent vehicle control algorithms. Attached Figure Description

[0063] Figure 1 This is a flowchart illustrating a method for estimating road surface adhesion coefficient based on a dynamic model and support vector regression in an embodiment of the present invention. Figure 2This is a schematic diagram of the vehicle dynamics model provided in an embodiment of the present invention; Figure 3 This is a schematic diagram of the structure of GWO-SVR and AEKF working together in an embodiment of the present invention; Figure 4 This is a schematic diagram of the hardware structure of an electronic device for estimating the road surface adhesion coefficient based on a dynamic model and support vector regression, as described in an embodiment of the present invention. Detailed Implementation

[0064] To facilitate a better understanding of the present invention by those skilled in the art, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments. The following are merely exemplary and do not limit the scope of protection of the present invention.

[0065] This invention discloses a method for estimating the road surface adhesion coefficient of intelligent vehicles based on a physical model and data-driven approach, comprising the following steps: Step S1: Establish a training data acquisition module to collect the vehicle's four-wheel wheel speed, vehicle speed, longitudinal acceleration, lateral velocity, lateral acceleration, yaw rate, steering wheel angle, tire lateral force, tire longitudinal force, and road adhesion coefficient. Step S2: Use the features in the training dataset formed in step S1 as input to SVR to obtain a pseudo-measured value of the road surface adhesion coefficient. Step S3: Construct a three-degree-of-freedom vehicle dynamics model and a Dugoff tire model, and design the AEKF algorithm model; The pseudo-measured values ​​of the road adhesion coefficient obtained in steps S4 and S2 are used as an extension of the measurement vector of the AEKF algorithm in S3 to achieve the optimal road adhesion coefficient estimation.

[0066] To better understand the above technical solutions, exemplary embodiments of the present invention will be described in detail below with reference to the accompanying drawings. These are merely exemplary embodiments of the present invention; however, it should be understood that the present invention can be implemented in various forms and is not limited to the embodiments described herein. These embodiments are provided to enable those skilled in the art to understand the present invention more clearly and thoroughly.

[0067] Example 1 Please see Figures 1 to 3 This invention proposes an embodiment of a method for estimating the road surface adhesion coefficient of intelligent vehicles based on a dynamic model and support vector regression. The specific steps include: Step S1: Establish a training data acquisition module to collect the vehicle's four-wheel wheel speed, vehicle speed, longitudinal acceleration, lateral velocity, lateral acceleration, yaw rate, steering wheel angle, tire lateral force, tire longitudinal force, and road adhesion coefficient. Furthermore, step S1 specifically includes the following steps: Step S11: Using CarSim and Simulink, based on the vehicle's longitudinal driving conditions, set a sinusoidal sweep frequency signal and design three target vehicle speed control input curves for acceleration, deceleration, and constant speed. Step S12: Based on the vehicle's lateral driving conditions, set different frequencies and amplitudes of steering wheel angle control input curves for dual lane change input, sinusoidal steering input, and steady-state rotation input.

[0068] Step S13: Select three different typical road surface types. Please refer to Table 1. Perform S11 and S12 for each road surface condition.

[0069] Table 1 Three typical pavement types

[0070] Step S14: Arrange the above target vehicle speed, steering wheel angle, and target adhesion coefficient to make the training dataset have high coverage of working conditions.

[0071] Furthermore, step S14 specifically includes the following steps: 1) Data preprocessing: Check the first five rows of the training dataset, check for missing values ​​in the data table, and check the mean, standard deviation, minimum value, quantiles, and maximum value of the data in the dataset; 2) Feature engineering: Extracting feature variables and label variables; 3) Data set splitting: The data table is split into 70% training set and 30% test set.

[0072] Step S2: Use the features in the training dataset formed in step S1 as input to SVR to obtain the pseudo-measured value of the road surface adhesion coefficient.

[0073] Furthermore, step S2 specifically includes the following steps: Step S21: The training dataset is determined by the four-wheel wheel speed, vehicle speed, longitudinal acceleration, lateral velocity, lateral acceleration, yaw rate, steering wheel angle, tire lateral force, tire longitudinal force, and road adhesion coefficient; the SVR input vector expression is as follows:

[0074] In the formula, For four-wheel wheel speed; For vehicle speed; It is longitudinal acceleration; Lateral velocity; It is lateral acceleration; This refers to the yaw rate; Steering wheel angle; This refers to the lateral force of the tire; This refers to the longitudinal force of the tire; where, , respectively representing left front, right front, left back, and right back.

[0075] Step S22: The SVR model is a multiple input multiple output model, and the RBF kernel function is selected. Its hyperparameters are autonomously optimized through GWO.

[0076] 1) The basic SVR model uses a given dataset , , .in, For the first Input features of each sample express It is 3D vector Indicates the feature dimension; For the first The target value for each sample It is the space of real numbers; This represents the total number of samples in the dataset.

[0077] The standard SVR model expression is:

[0078] The objective function is:

[0079] In the formula, The output predicted by the model; The input feature vector; This is the weight vector; For bias terms; For regularization terms, minimize Norms are used to control model complexity; For error terms, For regularization parameters; for Insensitive loss function This represents the error tolerance.

[0080]

[0081] 2) Introduce slack variables and The objective function for optimizing the SVR model is:

[0082] The error limit is:

[0083] 3) Introduce Lagrange multipliers and The objective function for optimizing the SVR model is:

[0084] Introducing the RBF kernel function:

[0085] In the formula, Center of the kernel function; The width of the kernel function; for arrive The Euclidean distance between them.

[0086] 4) The SVR expression combining the RBF kernel function is:

[0087] Step S23: Further, the SVR model uses the GWO algorithm to adjust the regularization parameters. Error tolerance To avoid getting trapped in local optima, a hyperparameter search is performed to improve the generalization ability and prediction performance of SVR. The MSE (Mean Sequence Equation) is used as the fitness function of the GWO (Geometric Optimization) algorithm. The regularization parameter in the SVR model is adjusted by minimizing the MSE corresponding to the fitness function. and error tolerance The optimization process for SVR is then carried out.

[0088] Furthermore, the MSE expression as the fitness function is:

[0089] In the formula, This is the actual value. These are predicted values.

[0090] Furthermore, in step S23, the GWO algorithm optimizes the regularization parameters of the SVR model. Error tolerance The specific steps include: 1) Define the SVR hyperparameter search space and initialize the GWO population; set 15 hyperparameter combinations, each containing For specific values, please refer to Table 2 for GWO parameter settings; Table 2 GWO Parameter Settings

[0091] Again, use , The current value is used to train the SVR model, and its MSE is calculated as the fitness function. Then, the optimal hyperparameter combination is obtained through GWO algorithm optimization. The optimal hyperparameter combination is identified based on the MSE value, where the combination that produces the lowest MSE is the optimal combination; the combination with the second lowest MSE is the suboptimal combination; and the combination with the third lowest MSE is the third optimal combination.

[0092] Furthermore, the GWO optimization process is as follows: 1) Update the positions of other hyperparameter combinations based on the optimal, second-best, and third-best combinations. The distances are calculated as follows:

[0093] In the formula, , , These represent the distances between the current combination and the optimal, second-best, and third-best combinations, respectively. The position of the optimal combination; The position of the second-best combination; The position of the third best combination; The position of the current combination; The components are randomly combined by the computer during the calculation process.

[0094] 2) Combination position update:

[0095] In the formula, The components are randomly combined by the computer during the calculation process; 4) The final new position of the combination:

[0096] In the formula, This is the current iteration step; This is the next iteration step.

[0097] 5) Update fitness: Retrain the SVR model with the updated hyperparameter combination and calculate the new MSE; 6) Iteration: After completing the combination position update, calculate the new MSE. If the new MSE is better than the historical MSE, then update. , , This continues until the maximum number of iterations is reached. This represents the optimal combination of hyperparameters for the SVR model.

[0098] Finally, the regularization hyperparameters found using GWO were used. Error tolerance The optimal combination of hyperparameters is used to retrain the SVR model, and the resulting optimized model is used to predict the pseudo-measured values ​​of the road adhesion coefficient.

[0099] Step S3: Establish a three-degree-of-freedom vehicle dynamics model and a Dugoff tire model, and design the AEKF algorithm. Please refer to Table 3 for some vehicle parameters. Table 3 Vehicle Information

[0100] Furthermore, step S3 specifically includes the following steps: Step S31, the expression for the three-degree-of-freedom vehicle dynamics model is as follows: ; In the formula, It is longitudinal acceleration; It is lateral acceleration; Let be the rotational angular acceleration about the z-axis; For the overall vehicle weight; To bypass Moment of inertia of the shaft; This is the distance from the center of gravity to the front axle; This is the distance from the center of mass to the rear axle; , These are the half track widths of the front and rear axles, respectively. Step S32, the expression for the Dugoff tire model is as follows:

[0101]

[0102]

[0103]

[0104]

[0105]

[0106]

[0107]

[0108] In the formula, Let be the acceleration due to gravity, taken as 9.8; The height of the vehicle's center of gravity above the ground; For longitudinal stiffness; Lateral stiffness; The sideslip angle of each wheel; The slip ratio of each wheel; It is a dimensionless parameter; As a speed-affecting factor; It is a normalization function used to ensure that the tire force increases linearly under low slip and reaches the friction limit under high slip, which helps to simulate the dynamic behavior of the tire under different road conditions. The center speed of each wheel; The effective rolling radius of the wheel; The vertical force on each tire; This refers to the driving torque of the wheels. The moment of inertia of the wheel; The angular acceleration of the wheel; This represents the road surface adhesion coefficient for each wheel.

[0109] Step S33: Based on steps S31 and S32, design the basic AEKF algorithm to estimate the road surface adhesion coefficient; 1) The nonlinear system model is:

[0110] In the formula, for The state vector of the system at any given time; for The control input vector at each time step; for Measurement vectors of the time-series system; and These are the state transition function and the measurement function, respectively. and For independent Gaussian white noise, where , , The process noise covariance matrix is... This is the measurement noise covariance matrix.

[0111] 2) Define state variables: Control input vector: Measurement vector: ; 3) Initialize the state vector Covariance matrix Initial value of noise covariance and Sliding window parameters: ① Initial vehicle speed 40km / h km / h , ; ② Initial state vector ; ② Initial covariance matrix ; ③ Initial process noise covariance matrix Measurement noise covariance :

[0112]

[0113] ④ Set the sliding window size to 15 and set the initial window to an empty set.

[0114] 4) Predict the state and error covariance based on the nonlinear system model: ① Using a three-degree-of-freedom vehicle dynamics model, in Estimating state at any time With control input ,predict Moment State ,Right now

[0115] in, This is the state transition function. This represents process noise. Its covariance is...

[0116] ② Predicting state estimation covariance The uncertainties of the three-degree-of-freedom vehicle dynamics model need to be considered for prediction. State error covariance at time 1 ,Right now

[0117] in, The state transition Jacobian matrix is ​​the linearized state transition function. Partial derivative with respect to the state vector, ; 5) Measurement residual calculation:

[0118] 6) Noise covariance in the adaptive update process Measurement noise covariance Dynamically update process noise covariance using a sliding window. Measurement noise covariance

[0119] 7) Utilizing measurement With control input vector Correct the predicted state: ① Through measurement function Utilizing the predicted state and control input vector ,predict Time measurement value ,Right now

[0120] ③ Calculate the Kalman gain:

[0121] in, To measure the Jacobian matrix, i.e., to linearize the measurement function. Partial derivative with respect to the state vector, ; 8) Update state estimation State estimation covariance : ;

[0122] 9) Repeat steps 4) to 7) until the loop ends, and output the road surface adhesion coefficient estimate.

[0123] The pseudo-measured values ​​of the road adhesion coefficient obtained in steps S4 and S2 are used as an extension of the measurement vector of the AEKF algorithm in step S3 to output the optimal road adhesion coefficient estimate.

[0124] Furthermore, step S4 specifically includes the following steps: ① Enhance the observation equation by using the pseudo-measured value of the road adhesion coefficient estimated by SVR as an extension of the measurement vector. The measurement vector is then redefined as...

[0125] ② Extend the measurement vector to measure the Jacobian matrix. Add calculation, denoted as ; ③ Use the extended version ,Right now Perform Kalman gain calculation:

[0126] ④ Use Update state estimation State estimation covariance :

[0127]

[0128] ⑤ Use Updated state estimation State estimation covariance Repeat steps 7 through 9.

[0129] For further details, please refer to Figure 4The present invention also provides an electronic device and a computer-readable storage medium, wherein the storage medium stores a computer program for a road surface adhesion coefficient estimation method based on a physical model and data, and the computer program, when executed by a processor, is used to implement the method provided in the above embodiments.

[0130] The readable storage medium can be a computer storage medium or a communication medium. A communication medium includes any medium that facilitates the transfer of computer programs from one location to another. A computer storage medium can be any available medium accessible to a general-purpose or special-purpose computer. For example, a readable storage medium is coupled to a processor, enabling the processor to read information from and write information to the readable storage medium. Of course, the readable storage medium can also be a component of the processor. The processor and the readable storage medium can reside in an Application-Specific Integrated Circuit (ASIC). Alternatively, the ASIC can be located in a user equipment. Of course, the processor and the readable storage medium can also exist as discrete components in a communication device. The readable storage medium can be a read-only memory (ROM), random access memory (RAM), CD-ROM, magnetic tape, floppy disk, and optical data storage device, etc.

[0131] In the embodiments of the above-described device, the processor used in this embodiment is based on an STM32F767 development board. The steps of the method disclosed in this invention can be directly manifested as being executed by a hardware processor, or being executed by a combination of hardware and software modules within the processor.

[0132] The above are merely embodiments of this application and are not intended to limit this application. Commonly known structures and characteristics in the solution are not described in detail here. Those skilled in the art are aware of all common technical knowledge in the field prior to the application date or priority date, are aware of all prior art in that field, and have the ability to apply conventional experimental methods prior to that date. Those skilled in the art can, under the guidance of this application, improve and implement this solution in combination with their own capabilities. Some typical well-known structures or methods should not be obstacles for those skilled in the art to implement this application. It should be noted that those skilled in the art can make several modifications and improvements without departing from the structure of this invention. These should also be considered within the scope of protection of this invention, and will not affect the effectiveness of the invention or the practicality of the patent.

[0133] It should be noted that the division of the various modules in the above system is merely a logical functional division. In actual implementation, they can be all or partly concentrated on a single physical entity, or they can be physically separated. These modules can be implemented entirely in software through processing elements, entirely in hardware, or partially in software through processing elements and partially in hardware. For example, the receiving module can be a separate processing element, or it can be integrated into a chip in the aforementioned device. Alternatively, it can be stored as program code in the memory of the aforementioned device and implemented by a chip in the device. Another option is to store program code in the memory of the aforementioned device, which can then be called and executed by a processing element of the device. The implementation of other modules is similar. Furthermore, these modules can be all or partly integrated together, or implemented independently. The processing elements mentioned here can be integrated circuits with signal processing capabilities. During implementation, the steps of the above method or the various modules can be completed through the integrated logic circuits in the hardware of the processor element or through human instructions in software form.

[0134] For example, these modules can be one or more integrated circuits configured to implement the above methods, such as one or more specific integrated circuits, one or more microprocessors, or one or more field-configurable gate arrays, etc. As another example, when a module is implemented through processing element scheduler code, the processing element can be a general-purpose processor, such as a central processing unit or other processor capable of calling program code. Furthermore, these modules can be integrated together to implement a system-on-a-chip.

[0135] The present invention provides a terminal device, including a memory, a processor, and a computer program stored in the memory and capable of running on the processor. The memory stores the computer program capable of running on the processor. When the processor loads and executes the computer program, it employs the steps of the above-mentioned intelligent vehicle road surface adhesion coefficient estimation method based on physical model and data drive.

[0136] It should be noted that the terminal device can be a computer device such as a desktop computer, a laptop computer, or a cloud server, and the terminal device includes, but is not limited to, a processor and a memory. For example, the terminal device may also include input / output devices, network access devices, and buses.

[0137] Furthermore, the processor can be a central processing unit. Of course, depending on the actual use, it can also be other general-purpose processors, digital signal processors, application-specific integrated circuits, thread-programmable gate arrays, or other programmable logic devices, discrete hardware components, etc. The general-purpose processor can be a microprocessor or any conventional processor, etc., and this application does not limit it in this regard.

[0138] In addition, the present invention provides a storage medium containing computer-executable instructions, characterized in that the computer-executable instructions, when executed by a computer processor, are used to execute the above-mentioned physical model and data-driven fusion estimation algorithm.

[0139] The computer program can be stored in a computer-readable medium. The computer program includes computer program code, which can be in the form of source code, object code, executable file, or certain middleware. The computer-readable medium includes any entity or device capable of carrying computer program code, recording media, USB flash drive, portable hard drive, magnetic disk, optical disk, computer memory, read-only memory, random access memory, electrical carrier signal, telecommunication signal, and software distribution medium, etc. It should be noted that the computer-readable medium includes, but is not limited to, the above-mentioned components.

[0140] It should be noted that in this article, such as First and second Relational terms such as "comprising," "including," or other variations are used merely to distinguish one entity or operation from another, without necessarily requiring or implying any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or other variations are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

Claims

1. A method for estimating road surface adhesion coefficient based on a dynamic model and support vector regression, characterized in that, Includes the following steps: S1. Establish a training data acquisition module to collect data on the vehicle's four-wheel wheel speed, vehicle speed, longitudinal acceleration, lateral velocity, lateral acceleration, yaw rate, steering wheel angle, tire lateral force, tire longitudinal force, and road adhesion coefficient. S2. Use the features in the training dataset formed in S1 as input to SVR to obtain the pseudo-measured value of the road adhesion coefficient. S3. Establish a three-degree-of-freedom vehicle dynamics model and a Dugoff tire model, and design the AEKF algorithm model; S4. The pseudo-measured values ​​of the road surface adhesion coefficient obtained in step S2 are used as an extension of the measurement vector of the AEKF algorithm in step S3 to achieve the optimal estimation of the road surface adhesion coefficient.

2. The method for estimating road surface adhesion coefficient based on a dynamic model and support vector regression according to claim 1, characterized in that, Step S1 specifically includes: S11. Based on the longitudinal driving conditions of the vehicle, a sinusoidal sweep frequency signal is set, and three target vehicle speed control input curves for acceleration, deceleration, and constant speed are designed. S12. Based on the vehicle's lateral driving conditions, set different frequencies and amplitudes of steering wheel angle control input curves for dual lane change input, sinusoidal steering input, and steady-state rotation input. S13. Select three different road surface types, and execute S11 and S12 under each road surface condition. S14. Arrange the above target vehicle speed, steering wheel angle, and target adhesion coefficient to make the training dataset have high coverage of working conditions.

3. The method for estimating road surface adhesion coefficient based on a dynamic model and support vector regression according to claim 1, characterized in that, The training dataset is determined by four-wheel wheel speed, vehicle speed, longitudinal acceleration, lateral velocity, lateral acceleration, yaw rate, steering wheel angle, tire lateral force, tire longitudinal force, and road adhesion coefficient; the SVR input vector expression is as follows: In the formula, For four-wheel wheel speed; For vehicle speed; It is longitudinal acceleration; Lateral velocity; It is lateral acceleration; This refers to the yaw rate; Steering wheel angle; This refers to the lateral force of the tire; This refers to the longitudinal force of the tire; where, , respectively representing left front, right front, left back, and right back.

4. The method for estimating road surface adhesion coefficient based on a dynamic model and support vector regression according to claim 1, characterized in that, The regularization parameters and error tolerance of the SVR model are autonomously optimized by GWO, with the parameters set to optimize the regularization parameter range as []. The optimized error tolerance range is []. The number of races is 10; the maximum number of iterations is 4. The SVR model is trained using the optimal combination of parameters to obtain pseudo-measured values ​​of the road adhesion coefficient, which serve as an extension of the measurement vector in the AEKF algorithm.

5. The method for estimating road surface adhesion coefficient based on a dynamic model and support vector regression according to claim 4, characterized in that, The measurement vector in the AEKF algorithm needs to be redefined In the formula, This represents the estimated road adhesion coefficient for each wheel output by the SVR model.

Images