Road adhesion coefficient hybrid estimation method for vehicle ESP system

Abstract

The invention discloses a road adhesion coefficient hybrid estimation method for a vehicle ESP system. The method comprises the following steps: firstly, constructing a mapping relation between a road adhesion coefficient and a vehicle motion state parameter based on a seven-degree-of-freedom vehicle model and a Dugoff tire model; secondly, two-channel unscented Kalman filtering is designed to carry out real-time estimation on vehicle motion state parameters and road adhesion coefficients, and a graph optimization method is introduced to carry out global correction on estimation results; aiming at an unscented Kalman filter scaling parameter optimization problem, an ETARBMO algorithm is provided to realize parameter adaptive setting; and finally, combining the dual-channel adaptive unscented Kalman filtering with a graph optimization method to form a hybrid estimation method, and alternately improving the estimation precision. According to the method, the estimation precision and robustness of the road adhesion coefficient are effectively improved, and a more accurate and reliable road state sensing basis is provided for an ESP system.

Description

Technical Field

[0001] This invention relates to the field of vehicle electronic stability program (ESP) systems, and more specifically to a hybrid estimation method for road adhesion coefficient for vehicle ESP systems. Background Technology

[0002] Accurate estimation of the road adhesion coefficient is crucial for ESP systems, helping to improve control system response efficiency, reduce the risk of vehicle loss of control, and enhance safety performance. Existing methods for estimating the road adhesion coefficient mainly fall into the following categories: In filter-based estimation methods, particle filters are applied to straight-line and turning driving conditions, and different estimators have been developed to improve estimation accuracy and robustness; road adhesion coefficients are estimated by combining tire dynamics characteristics with parameters such as fuzzy rules and wheel slip ratio; a fusion estimation framework based on multi-source information quality assessment has been designed, achieving higher estimation accuracy; and an adaptive singular value decomposition unscented Kalman estimator with a noise estimation algorithm has been proposed to suppress the negative definiteness of the state covariance matrix in unscented Kalman filtering. Regarding neural network applications, gated recurrent units and deep neural networks have been used for road adhesion coefficient estimation, but these methods fail to fully consider time-series characteristics. To address this issue, research has developed road adhesion coefficient estimation methods based on long short-term memory networks, and other studies have introduced fuzzy forgetting factors into unscented Kalman filtering to improve estimation accuracy. In recent years, methods have been developed that fuse the fruit fly optimization algorithm with neural networks to estimate the road adhesion coefficient. Other studies have proposed an improved extreme learning machine based on the sparrow search algorithm, significantly improving the estimation accuracy of the road adhesion coefficient. However, most existing estimation methods rely on fixed parameters or single sensor data, limiting the system's global adaptability and making it difficult to achieve real-time, accurate estimation, thus requiring further improvement. Summary of the Invention

[0003] The purpose of this invention is to provide a hybrid estimation method for road surface adhesion coefficient for vehicle ESP systems, in order to overcome the shortcomings of existing estimation methods in terms of noise interference, parameter sensitivity, and global error accumulation.

[0004] To achieve the above objectives, the present invention adopts the following technical solution: A hybrid estimation method for road adhesion coefficient for vehicle ESP systems includes the following steps: Step 1: Based on the seven-DOF vehicle model and the Dugoff tire model, construct the mapping relationship between vehicle motion state parameters and road adhesion coefficient; Step 2: Design a dual-channel unscented Kalman filter algorithm to estimate the vehicle's motion state parameters and road adhesion coefficient from Step 1, respectively; Step 3: Combine the graph optimization estimation method to correct the state values ​​in the dual-channel unscented Kalman filter algorithm in Step 2, and obtain the graph optimization method for estimating the road surface adhesion coefficient. Step 4: Design the Red-billed Blue Magpie algorithm that integrates the decreasing exploration leap evolution strategy and the addiction escape mechanism. Tune the scaling parameters in the dual-channel unscented Kalman filter algorithm in Step 2 to realize dual-channel adaptive unscented Kalman filtering. Step 5: Combine the dual-channel adaptive unscented Kalman filter algorithm in Step 4 with the graph optimization estimation method in Step 3 to form a hybrid dual-channel adaptive unscented Kalman filter-graph optimization road adhesion coefficient estimation method. The road adhesion coefficient is estimated by alternating between the dual-channel adaptive unscented Kalman filter and the graph optimization road adhesion coefficient estimation method.

[0005] Furthermore, the seven-degree-of-freedom vehicle model includes longitudinal degrees of freedom, lateral degrees of freedom, yaw degrees of freedom, and rotational degrees of freedom for the four wheels, as detailed below: (1) (2) (3) (4) (5) Wherein, formula (1) is the differential equation of the longitudinal kinematics of the vehicle, and in the formula, For the overall vehicle quality, For longitudinal vehicle speed, The lateral speed is the speed of the vehicle. The yaw rate is angular velocity. For the turning angle of the wheel, , The longitudinal forces are respectively the left and right front wheels. , These are the lateral forces on the left and right front wheels, respectively. , These are the longitudinal forces of the left and right rear wheels, respectively; Formula (2) is the differential equation of the vehicle's lateral kinematics, where, , These are the lateral forces of the left and right rear wheels, respectively; Formula (3) is the differential equation of the vehicle's yaw kinematics, where, For the whole vehicle about z Moment of inertia of the shaft The front axle track. The rear axle track. This is the distance from the center of gravity to the front axle. Let be the distance from the center of mass to the rear axis; Equation (4) is the dynamic equation for the four-degree-of-freedom rotational equilibrium, where Let be the moment of inertia of the wheel about its center. , , , These are the angular velocities of the left front wheel, right front wheel, left rear wheel, and right rear wheel, respectively. , , , These are the driving torques for the left front wheel, right front wheel, left rear wheel, and right rear wheel, respectively. , , , These are the vertical loads on the four wheels. The rolling resistance coefficient of the wheel. Let be the wheel radius; formula (5) is for calculating the vertical load on the four wheels, where The height of the vehicle's center of gravity. For the car along x Axial acceleration, For the car along y Axial acceleration.

[0006] Furthermore, the explicit mathematical relationship between tire force and road adhesion coefficient constructed by the Dugoff tire model is as follows: (6) In the formula, This is the longitudinal force acting on the tire. The road surface adhesion coefficient, For tire slip ratio, For correction functions, L These are the nonlinear characteristic parameters of the slip process. For longitudinal stiffness, The lateral force acting on the tire. For lateral stiffness, a This refers to the tire slip angle. This represents the effect of slip speed on tire force.

[0007] Furthermore, the dual-channel unscented Kalman filter design includes the following steps: Step 2.1: Design Channel 1: Estimate vehicle motion state parameters based on UKF1 filtering algorithm; Step 2.2: Design Channel Two: Estimate the road surface adhesion coefficient based on the UKF2 filtering algorithm.

[0008] Furthermore, in step 2.1, the nonlinear system for estimating the vehicle's motion state parameters is established as follows: (7) (8) (9) (10) (11) Wherein, formula (7) is the state equation of the nonlinear system, and in the formula, To estimate the state vector of a nonlinear system for the vehicle's motion state parameters, To estimate the control input vector of a nonlinear system for estimating the vehicle's motion state parameters, The process noise of state estimation for nonlinear systems that estimate vehicle motion state parameters. For longitudinal vehicle speed, The lateral speed is the speed of the vehicle. The yaw rate is angular velocity. , , , These are the angular velocities of the left front wheel, right front wheel, left rear wheel, and right rear wheel, respectively. For longitudinal acceleration, For lateral acceleration, This is the yaw acceleration. , , , Let be the angular accelerations of the left front wheel, right front wheel, left rear wheel, and right rear wheel, respectively; Equation (8) is the measurement equation for this nonlinear system, where To estimate the observation vector of a nonlinear system for estimating the vehicle's motion state parameters, Observation noise for state estimation of a nonlinear system for estimating vehicle motion state parameters. For the car along x Axial acceleration, For the car along y Axial acceleration, , For the left and right wheel turning angles, , , , These are the driving torques for the left front wheel, right front wheel, left rear wheel, and right rear wheel, respectively. Total braking torque; The expressions for each term in the formula are shown in formula (12): (12).

[0009] Furthermore, in step 2.2, the nonlinear system for estimating the road surface adhesion coefficient is established as follows: (13) (14) (15) (16) (17) Wherein, formula (13) is the state equation of the nonlinear system, and in the formula, To estimate the state vector of a nonlinear system for road surface adhesion coefficient, To estimate the control input vector of the nonlinear system for the road surface adhesion coefficient, The process noise for state estimation of the nonlinear system for estimating road surface adhesion coefficient. , , , Let be the road adhesion coefficients of the left front wheel, right front wheel, left rear wheel, and right rear wheel, respectively; formula (14) is the measurement equation of this nonlinear system, where To estimate the observation vector of the nonlinear system for the road surface adhesion coefficient, Observation noise for state estimation of nonlinear systems for estimating road adhesion coefficient. m For the overall vehicle quality, , The longitudinal forces are respectively the left and right front wheels. , The lateral forces are for the left and right front wheels, respectively. , The longitudinal forces are respectively the left and right rear wheels. , These are the lateral forces on the left and right rear wheels, respectively. The expressions for each term in the formula are shown in formula (18): (18).

[0010] In the formula, Let Z be the moment of inertia of the entire vehicle about the z-axis. The front axle track. The rear axle track. 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.

[0011] Furthermore, the graph-optimized road surface adhesion coefficient estimation method corrects the state values ​​in the dual-channel unscented Kalman filter algorithm by constructing a graph optimization objective function, as follows: The objective function for graph optimization is as follows: (19) In the formula, e k For systematic error, h ( x k ) is the observation equation, It is the inverse of the covariance matrix; A first-order expansion of the systematic error yields: (20) In the formula, yes e k about x k The derivative of is expressed as the Jacobian matrix. The expression for the change in road surface adhesion coefficient, with a linear assumption at the estimation point, is as follows: (twenty one) In the formula, The value before the change; The changes in the objective function are represented as follows: (twenty two) By using the command (22) The derivative is 0, which gives equation (23). Solving the linear equation system of equation (23) yields the state correction. This refers to the change in the road surface adhesion coefficient, specifically as follows: (twenty three) In the formula, , Due to systematic error e k and Jacobian matrix Calculations show that e k and The expression is as follows: (twenty four) (25) In equation (24), and They represent and Measured value express The estimated value of the yaw rate after passing through the UKF1 filtering algorithm.

[0012] Furthermore, the Red-billed Blue Magpie algorithm, which integrates a decreasing exploratory leap evolution strategy with an addiction escape mechanism, includes the following steps: Step 4.1: Randomly initialize the population positions and establish the initial search space; Step 4.2: Based on the initial population, update the individual positions using a decreasing exploration transition strategy; Step 4.3: During the individual location update process, monitor the changes in individual fitness. When it is detected that the fitness value of an individual is continuously lower than the average level of the population within the preset iteration threshold, the addiction escape mechanism is triggered to relocate the individual to another region of the solution space to avoid premature convergence of the population. Step 4.4: After completing the position update and escape operations, recalculate the fitness values ​​of all individuals based on the updated population positions, select the current optimal solution, and provide a benchmark for the next iteration; Step 4.5: Determine if the maximum number of iterations has been reached. If it has, output the optimal scaling parameter value for the dual-channel unscented Kalman filter algorithm; if it has not been reached, return to step 4.2.

[0013] Furthermore, in step 4.2, the formula for calculating the number of individual exploration jumps in the decreasing exploration jump evolution strategy is as follows: (26) In the formula, Ev t The number of individuals that underwent the exploratory jump in this iteration. Ev t-1 This represents the number of individuals that successfully evolved in the previous iteration. As an evolutionary factor, s 1 and s 3 represents the deceleration rate control parameter. s 2. To control the degree of nonlinearity of the decrease, s 4. Control the magnitude of the exponential decay term; Exploratory migration of individuals within a population occurs in three ways, determined by random numbers, as shown in the following formulas: (27) In the formula, The golden ratio, This represents the optimal individual in the population. This represents an individual after an exploratory leap. For a random individual in the population, l A random number between [-1, 1] b Let be a constant, and define the shape of the logarithmic spiral. r is a random number uniformly distributed between [0,1], used to select the exploratory jump mode of the population. Rand 4 and Rand 5 represents a random number between 0 and 1, used to introduce disturbances; The formula for population evolution is as follows: (28) In the formula, This indicates the fitness of an individual after the exploration jump. This indicates the fitness of an individual before the leap.

[0014] Furthermore, in step 4.3, the position update formula for the addicted individual in the addiction escape mechanism is as follows: (29) In the formula, For the sake of individual addiction, , , The new positions obtained through the three exploratory jump methods are selected, and the individual with the best fitness is chosen to replace the original addicted individual.

[0015] Compared with the prior art, the present invention has the following beneficial technical effects: This invention estimates the vehicle motion state and road adhesion coefficient separately using dual-channel unscented Kalman filtering, and performs global correction using graph optimization methods, which effectively suppresses error accumulation and improves estimation accuracy and robustness.

[0016] This invention addresses the problem of optimizing scaling parameters in unscented Kalman filters by proposing the ETARBMO (Red-billed Blue Magpie Optimization Algorithm), which integrates a decreasing exploratory transition strategy with an addictive escape mechanism. This enhances parameter adaptability and avoids premature convergence.

[0017] This invention organically combines filtering and graph optimization through a hybrid estimation framework, which significantly reduces mean square error and mean absolute error while ensuring real-time performance, providing the ESP system with a more accurate and reliable basis for road condition perception. Attached Figure Description

[0018] The accompanying drawings are provided to further understand the invention and constitute a part of this invention. The illustrative embodiments of the invention and their descriptions are used to explain the invention and do not constitute an improper limitation of the invention.

[0019] Figure 1 This is a flowchart illustrating the present invention; Figure 2 This is a flowchart of the dual-channel unscented Kalman filter estimation method involved in this invention; Figure 3 This is a flowchart of the hybrid estimation method involved in this invention; Figure 4 This is a comparison diagram of the estimation results of the present invention under different road surfaces, where (a) is a high road surface, (b) is a low road surface, and (c) is a connecting road surface. Detailed Implementation

[0020] The present invention will now be described in further detail with reference to the accompanying drawings: See Figure 1The flowchart of this invention is shown below. This invention proposes a road surface hybrid estimation method for vehicle ESP systems. First, a seven-DOF vehicle model and a Dugoff tire model are established to construct the mapping relationship between vehicle motion state parameters and road surface adhesion coefficient. Then, as shown in the flowchart... Figure 2 The design presents a dual-channel unscented Kalman filter to estimate the vehicle's motion state parameters and road adhesion coefficient in step 1. Then, a graph optimization estimation method is introduced to globally correct the filtering results. Furthermore, to address the scaling parameter optimization problem of the unscented Kalman filter, the ETARBMO algorithm, which integrates a decreasing exploratory transition evolution strategy and an addiction escape mechanism, is proposed to achieve a dual-channel adaptive unscented Kalman filter. Finally, as shown... Figure 3 The dual-channel adaptive unscented Kalman filter algorithm in step 4 is combined with the graph optimization estimation method in step 3 to form a hybrid dual-channel adaptive unscented Kalman filter-graph optimization road adhesion coefficient estimation method. The estimation accuracy of the road adhesion coefficient is gradually improved by alternating between the two methods.

[0021] The specific steps are as follows: Step 1: Based on the seven-DOF vehicle model and the Dugoff tire model, construct the mapping relationship between vehicle motion state parameters and road adhesion coefficient: The seven-DOF vehicle model includes longitudinal, lateral, and yaw degrees of freedom, as well as rotational degrees of freedom for the four wheels, as detailed below: (1) (2) (3) (4) (5) Wherein, formula (1) is the differential equation of the longitudinal kinematics of the vehicle, and in the formula, For the overall vehicle quality, For longitudinal vehicle speed, The lateral speed is the speed of the vehicle. The yaw rate is angular velocity. For the turning angle of the wheel, , The longitudinal forces are respectively the left and right front wheels. , These are the lateral forces on the left and right front wheels, respectively. , These are the longitudinal forces of the left and right rear wheels, respectively; Formula (2) is the differential equation of the vehicle's lateral kinematics, where, , These are the lateral forces of the left and right rear wheels, respectively; Formula (3) is the differential equation of the vehicle's yaw kinematics, where, For the whole vehicle aboutz Moment of inertia of the shaft The front axle track. The rear axle track. This is the distance from the center of gravity to the front axle. Let be the distance from the center of mass to the rear axis; Equation (4) is the dynamic equation for the four-degree-of-freedom rotational equilibrium, where Let be the moment of inertia of the wheel about its center. , , , These are the angular velocities of the left front wheel, right front wheel, left rear wheel, and right rear wheel, respectively. , , , These are the driving torques for the left front wheel, right front wheel, left rear wheel, and right rear wheel, respectively. , , , These are the vertical loads on the four wheels. The rolling resistance coefficient of the wheel. Let be the wheel radius; formula (5) is for calculating the vertical load on the four wheels, where The height of the vehicle's center of gravity. For the car along Axial acceleration, For the car along Axial acceleration.

[0022] The Dugoff tire model establishes the explicit mathematical relationship between tire force and road adhesion coefficient as follows: (6) In the formula, This is the longitudinal force acting on the tire. The road surface adhesion coefficient, For tire slip ratio, For correction functions, L These are the nonlinear characteristic parameters of the slip process. For longitudinal stiffness, The lateral force acting on the tire. For lateral stiffness, a This refers to the tire slip angle. This represents the effect of slip speed on tire force.

[0023] Step 2: Design a dual-channel unscented Kalman filter algorithm to estimate the vehicle's motion state parameters and road adhesion coefficient from Step 1: Step 2.1: Design Channel 1: Estimate vehicle motion state parameters based on UKF1 filtering algorithm; (7) (8) (9) (10) (11) Wherein, formula (7) is the state equation of the nonlinear system, and in the formula, To estimate the state vector of a nonlinear system for the vehicle's motion state parameters, To estimate the control input vector of a nonlinear system for estimating the vehicle's motion state parameters, The process noise of state estimation for nonlinear systems that estimate vehicle motion state parameters. For longitudinal vehicle speed, The lateral speed is the speed of the vehicle. The yaw rate is angular velocity. , , , These are the angular velocities of the left front wheel, right front wheel, left rear wheel, and right rear wheel, respectively. For longitudinal acceleration, For lateral acceleration, This is the yaw acceleration. , , , Let be the angular accelerations of the left front wheel, right front wheel, left rear wheel, and right rear wheel, respectively; Equation (8) is the measurement equation for this nonlinear system, where To estimate the observation vector of a nonlinear system for estimating the vehicle's motion state parameters, Observation noise for state estimation of a nonlinear system for estimating vehicle motion state parameters. For the car along x Axial acceleration, For the car along y Axial acceleration, , For the left and right wheel turning angles, , , , These are the driving torques for the left front wheel, right front wheel, left rear wheel, and right rear wheel, respectively. Total braking torque; The expressions for each term in the formula are shown in formula (12): (12) Step 2.2: Design Channel Two: Estimate the road surface adhesion coefficient based on the UKF2 filtering algorithm; (13) (14) (15) (16) (17) Wherein, formula (13) is the state equation of the nonlinear system, and in the formula, To estimate the state vector of a nonlinear system for road surface adhesion coefficient, To estimate the control input vector of the nonlinear system for the road surface adhesion coefficient, The process noise for state estimation of the nonlinear system for estimating road surface adhesion coefficient. , , , Let be the road adhesion coefficients of the left front wheel, right front wheel, left rear wheel, and right rear wheel, respectively; formula (14) is the measurement equation of this nonlinear system, where To estimate the observation vector of the nonlinear system for the road surface adhesion coefficient, Observation noise for state estimation of nonlinear systems for estimating road adhesion coefficient. m For the overall vehicle quality, , The longitudinal forces are respectively the left and right front wheels. , The lateral forces are for the left and right front wheels, respectively. , The longitudinal forces are respectively the left and right rear wheels. , These are the lateral forces of the left and right rear wheels, respectively. The expressions for each term in the formula are shown in formula (18): (18) In the formula, Let Z be the moment of inertia of the entire vehicle about the z-axis. The front axle track. The rear axle track. 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.

[0024] Step 3: Combining the graph optimization estimation method, the state values ​​in the dual-channel unscented Kalman filter in Step 2 are corrected to obtain the graph optimization method for estimating the road surface adhesion coefficient. The objective function for graph optimization is as follows: (19) In the formula, e k For systematic error, h ( xk ) is the observation equation, This is the inverse of the covariance matrix. A first-order expansion of the systematic error yields: (20) In the formula, yes e k about x k The derivative of is expressed as the Jacobian matrix. The expression for the change in road surface adhesion coefficient, with a linear assumption at the estimation point, is as follows: (twenty one) In the formula, The values ​​before the change are shown. The changes in the objective function are represented as follows: (twenty two) By using the command (22) The derivative is 0, which gives equation (23). Solving the linear equation system of equation (23) yields the state correction. This refers to the change in the road surface adhesion coefficient, specifically as follows: (twenty three) In the formula, , Can be caused by systematic error e k and Jacobian matrix Calculations show that e k and The expression is as follows: (twenty four) (25) In equation (24), and They represent and Measured value express The estimated value of the yaw rate after passing through the UKF1 filtering algorithm.

[0025] Step 4: Design the ETARBMO algorithm, which integrates a decreasing exploration leap evolution strategy with an addiction escape mechanism: Step 4.1: Randomly initialize the population positions and establish the initial search space; Step 4.2: Based on the initial population, update the individual positions using a decreasing exploration transition strategy; The formula for calculating the number of individual exploration jumps is as follows: (26) In the formula, Ev t The number of individuals that underwent the exploratory jump in this iteration. Ev t-1 This represents the number of individuals that successfully evolved in the previous iteration. As an evolutionary factor, s 1 and s 3 represents the deceleration rate control parameter. s 2. To control the degree of nonlinearity of the decrease, s 4. Control the magnitude of the exponential decay term.

[0026] Exploratory migration of individuals within a population occurs in three ways, determined by random numbers, as shown in the following formulas: (27) In the formula, The golden ratio, This represents the optimal individual in the population. This represents an individual after an exploratory leap. For a random individual in the population, l A random number between [-1, 1] b Let be a constant, and define the shape of the logarithmic spiral. r is a random number uniformly distributed between [0,1], used to select the exploratory jump mode of the population. Rand 4 and Rand 5 represents a random number between 0 and 1, used to introduce disturbances.

[0027] The formula for population evolution is as follows: (28) In the formula, This indicates the fitness of an individual after the exploration jump. This indicates the fitness of an individual before the leap.

[0028] Step 4.3: During the individual location update process, monitor the changes in individual fitness. When it is detected that the fitness value of an individual is continuously lower than the average level of the population within the preset iteration threshold, the addiction escape mechanism is triggered to relocate the individual to another region of the solution space to avoid premature convergence of the population. The position update formula for addicted individuals in the addiction escape mechanism is as follows: (29) In the formula, For the sake of individual addiction, , , The new positions obtained through the three exploratory jump methods are selected, and the individual with the best fitness is chosen to replace the original addicted individual.

[0029] Step 4.4: After completing the position update and escape operations, recalculate the fitness values ​​of all individuals based on the updated population positions, select the current optimal solution, and provide a benchmark for the next iteration; Step 4.5: Determine if the maximum number of iterations has been reached. If it has, output the optimal scaling parameter value for the dual-channel unscented Kalman filter algorithm; if it has not been reached, return to step 2.2.

[0030] Step 5: Combine the dual-channel adaptive unscented Kalman filter algorithm in Step 4 with the graph optimization estimation method in Step 3 to form a hybrid dual-channel adaptive unscented Kalman filter-graph optimization road adhesion coefficient estimation method. By alternating between the two methods, the estimation accuracy of the road adhesion coefficient is gradually improved.

[0031] Figure 4 To compare the estimation results under different road surfaces, where (a) represents a high road surface, (b) represents a low road surface, and (c) represents a connecting road surface, from... Figure 4 It can be seen that the hybrid dual-channel adaptive filtering-graph optimization road adhesion coefficient estimation method (G-ETARBMO-DUKF) has the fastest convergence speed and higher estimation accuracy compared with the dual-channel unscented Kalman filtering method (DUKF) and the dual-channel adaptive unscented Kalman filtering method (ETARBMO-DUKF), and its estimation results are closer to the actual adhesion coefficient value.

[0032] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit its scope of protection. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that after reading the present invention, they can still make various changes, modifications or equivalent substitutions to the specific implementation of the invention, but these changes, modifications or equivalent substitutions are all within the scope of protection of the pending claims of the invention.

Claims

1. A hybrid estimation method for road adhesion coefficient of vehicle ESP system, characterized in that, Includes the following steps: Step 1: Based on the seven-DOF vehicle model and the Dugoff tire model, construct the mapping relationship between vehicle motion state parameters and road adhesion coefficient; Step 2: Design a dual-channel unscented Kalman filter algorithm to estimate the vehicle's motion state parameters and road adhesion coefficient from Step 1, respectively; Step 3: Combine the graph optimization estimation method to correct the state values ​​in the dual-channel unscented Kalman filter algorithm in Step 2, and obtain the graph optimization method for estimating the road surface adhesion coefficient. Step 4: Design the Red-billed Blue Magpie algorithm that integrates the decreasing exploration leap evolution strategy and the addiction escape mechanism. Tune the scaling parameters in the dual-channel unscented Kalman filter algorithm in Step 2 to realize dual-channel adaptive unscented Kalman filtering. Step 5: Combine the dual-channel adaptive unscented Kalman filter algorithm in Step 4 with the graph optimization estimation method in Step 3 to form a hybrid dual-channel adaptive unscented Kalman filter-graph optimization road adhesion coefficient estimation method. The road adhesion coefficient is estimated by alternating between the dual-channel adaptive unscented Kalman filter and the graph optimization road adhesion coefficient estimation method.

2. The hybrid estimation method for road adhesion coefficient of vehicle ESP system according to claim 1, characterized in that, The seven-degree-of-freedom vehicle model includes longitudinal, lateral, and yaw degrees of freedom, as well as rotational degrees of freedom for the four wheels, as detailed below: （1） （2） （3） （4） （5） Wherein, formula (1) is the differential equation of the longitudinal kinematics of the vehicle, and in the formula, For the overall vehicle quality, For longitudinal vehicle speed, The lateral speed is the speed of the vehicle. The yaw rate is angular velocity. For the turning angle of the wheel, , The longitudinal forces are respectively the left and right front wheels. , These are the lateral forces on the left and right front wheels, respectively. , These are the longitudinal forces of the left and right rear wheels, respectively; Formula (2) is the differential equation of the vehicle's lateral kinematics, where, , These are the lateral forces of the left and right rear wheels, respectively; Formula (3) is the differential equation of the vehicle's yaw kinematics, where, For the whole vehicle about z Moment of inertia of the shaft The front axle track. The rear axle track. This is the distance from the center of gravity to the front axle. Let be the distance from the center of mass to the rear axis; Equation (4) is the dynamic equation for the four-degree-of-freedom rotational equilibrium, where Let be the moment of inertia of the wheel about its center. , , , These are the angular velocities of the left front wheel, right front wheel, left rear wheel, and right rear wheel, respectively. , , , These are the driving torques for the left front wheel, right front wheel, left rear wheel, and right rear wheel, respectively. , , , These are the vertical loads on the four wheels. The rolling resistance coefficient of the wheel. Let be the wheel radius; formula (5) is for calculating the vertical load on the four wheels, where The height of the vehicle's center of gravity. For the car along x Axial acceleration, For the car along y Axial acceleration.

3. The hybrid estimation method for road adhesion coefficient of vehicle ESP system according to claim 1, characterized in that, The explicit mathematical relationship between tire force and road adhesion coefficient constructed by the Dugoff tire model is as follows: （6） In the formula, This is the longitudinal force acting on the tire. The road surface adhesion coefficient, For tire slip ratio, For correction functions, L These are the nonlinear characteristic parameters of the slip process. For longitudinal stiffness, The lateral force acting on the tire. For lateral stiffness, a This refers to the tire slip angle. This represents the effect of slip speed on tire force.

4. The hybrid estimation method for road adhesion coefficient of vehicle ESP system according to claim 1, characterized in that, The dual-channel unscented Kalman filter design includes the following steps: Step 2.1: Design Channel 1: Estimate vehicle motion state parameters based on UKF1 filtering algorithm; Step 2.2: Design Channel Two: Estimate the road surface adhesion coefficient based on the UKF2 filtering algorithm.

5. The hybrid estimation method for road adhesion coefficient of vehicle ESP system according to claim 4, characterized in that, In step 2.1, the nonlinear system for estimating the vehicle's motion state parameters is established as follows: （7） （8） （9） （10） （11） Wherein, formula (7) is the state equation of the nonlinear system, and in the formula, To estimate the state vector of a nonlinear system for the vehicle's motion state parameters, To estimate the control input vector of a nonlinear system for estimating the vehicle's motion state parameters, The process noise of state estimation for nonlinear systems that estimate vehicle motion state parameters. For longitudinal vehicle speed, The lateral speed is the speed of the vehicle. The yaw rate is angular velocity. , , , These are the angular velocities of the left front wheel, right front wheel, left rear wheel, and right rear wheel, respectively. For longitudinal acceleration, For lateral acceleration, This is the yaw acceleration. , , , Let be the angular accelerations of the left front wheel, right front wheel, left rear wheel, and right rear wheel, respectively; Equation (8) is the measurement equation for this nonlinear system, where To estimate the observation vector of a nonlinear system for estimating the vehicle's motion state parameters, Observation noise for state estimation of a nonlinear system for estimating vehicle motion state parameters. For the car along x Axial acceleration, For the car along y Axial acceleration, , For the left and right wheel turning angles, , , , These are the driving torques for the left front wheel, right front wheel, left rear wheel, and right rear wheel, respectively. Total braking torque; The expressions for each term in the formula are shown in formula (12): （12）。 6. The hybrid estimation method for road adhesion coefficient of vehicle ESP system according to claim 4, characterized in that, In step 2.2, the nonlinear system for estimating the road surface adhesion coefficient is established as follows: （13） （14） （15） （16） （17） Wherein, formula (13) is the state equation of the nonlinear system, and in the formula, To estimate the state vector of a nonlinear system for road surface adhesion coefficient, To estimate the control input vector of the nonlinear system for the road surface adhesion coefficient, The process noise for state estimation of the nonlinear system for estimating road surface adhesion coefficient. , , , Let be the road adhesion coefficients of the left front wheel, right front wheel, left rear wheel, and right rear wheel, respectively; formula (14) is the measurement equation of this nonlinear system, where To estimate the observation vector of the nonlinear system for the road surface adhesion coefficient, Observation noise for state estimation of nonlinear systems for estimating road adhesion coefficient. m For the overall vehicle quality, , The longitudinal forces are respectively the left and right front wheels. , The lateral forces are for the left and right front wheels, respectively. , The longitudinal forces are respectively the left and right rear wheels. , These are the lateral forces on the left and right rear wheels, respectively. The expressions for each term in the formula are shown in formula (18): （18） In the formula, Let Z be the moment of inertia of the entire vehicle about the z-axis. The front axle track. The rear axle track. 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.

7. The hybrid estimation method for road adhesion coefficient of vehicle ESP system according to claim 4, characterized in that, The graph-optimized road surface adhesion coefficient estimation method corrects the state values ​​in the dual-channel unscented Kalman filter algorithm by constructing a graph-optimized objective function, as follows: The objective function for graph optimization is as follows: （19） In the formula, e k For systematic error, h ( x k ) is the observation equation, It is the inverse of the covariance matrix; A first-order expansion of the systematic error yields: （20） In the formula, yes e k about x k The derivative of is expressed as the Jacobian matrix. The expression for the change in road surface adhesion coefficient, with a linear assumption at the estimation point, is as follows: （21） In the formula, The value before the change; The changes in the objective function are represented as follows: （22） By using the command (22) The derivative is 0, which gives equation (23). Solving the linear equation system of equation (23) yields the state correction. This refers to the change in the road surface adhesion coefficient, specifically as follows: （23） In the formula, , Due to systematic error e k and Jacobian matrix Calculations show that e k and The expression is as follows: （24） （25） In equation (24), and They represent and Measured value express The estimated value of the yaw rate after passing through the UKF1 filtering algorithm.

8. The hybrid estimation method for road adhesion coefficient of vehicle ESP system according to claim 1, characterized in that, The Red-billed Blue Magpie algorithm, which integrates a decreasing exploration leap evolution strategy with an addiction escape mechanism, includes the following steps: Step 4.1: Randomly initialize the population positions and establish the initial search space; Step 4.2: Based on the initial population, update the individual positions using a decreasing exploration transition strategy; Step 4.3: During the individual location update process, monitor the changes in individual fitness. When it is detected that the fitness value of an individual is continuously lower than the average level of the population within the preset iteration threshold, the addiction escape mechanism is triggered to relocate the individual to another region of the solution space to avoid premature convergence of the population. Step 4.4: After completing the position update and escape operations, recalculate the fitness values ​​of all individuals based on the updated population positions, select the current optimal solution, and provide a benchmark for the next iteration; Step 4.5: Determine if the maximum number of iterations has been reached. If it has, output the optimal scaling parameter value for the dual-channel unscented Kalman filter algorithm; if it has not been reached, return to step 4.

2.

9. The hybrid estimation method for road adhesion coefficient of vehicle ESP system according to claim 8, characterized in that, In step 4.2, the formula for calculating the number of individual exploration jumps in the decreasing exploration jump evolution strategy is as follows: （26） In the formula, Ev t The number of individuals that underwent the exploratory jump in this iteration. Ev t-1 This represents the number of individuals that successfully evolved in the previous iteration. As an evolutionary factor, s 1 and s 3 represents the deceleration rate control parameter. s 2. To control the degree of nonlinearity of the decrease, s 4. Control the magnitude of the exponential decay term; Exploratory migration of individuals within a population occurs in three ways, determined by random numbers, as shown in the following formulas: （27） In the formula, The golden ratio, This represents the optimal individual in the population. This represents an individual after an exploratory leap. For a random individual in the population, l A random number between [-1, 1] b Let be a constant, and define the shape of the logarithmic spiral. r is a random number uniformly distributed between [0,1], used to select the exploratory jump mode of the population. Rand 4 and Rand 5 represents a random number between 0 and 1, used to introduce disturbances; The formula for population evolution is as follows: （28） In the formula, This indicates the fitness of an individual after the exploration jump. This indicates the fitness of an individual before the leap.

10. The hybrid estimation method for road adhesion coefficient of vehicle ESP system according to claim 9, characterized in that, In step 4.3, the position update formula for the addicted individual in the addiction escape mechanism is as follows: （29） In the formula, For the sake of individual addiction, , , The new positions obtained through the three exploratory jump methods are selected, and the individual with the best fitness is chosen to replace the original addicted individual.

Images