CN116588119A - A Vehicle State Estimation Method Based on Adaptive Tire Model Parameters - Google Patents

Abstract

The invention relates to a vehicle state estimation method based on tire model parameter self-adaption, which comprises the following steps: collecting real vehicle data; building a tire experience model based on the vehicle load transfer model and the wheel center speed calculation; establishing a vehicle state estimation scheme based on unscented Kalman filtering; screening and memorizing data fragments meeting the continuous excitation condition under the typical working condition; carrying out tire model parameter identification by adopting particle swarm optimization, and obtaining the relation of tire model parameters along with the change of the vehicle speed according to the fitting of optimal tire model parameters in different vehicle speed sections; substituting the relation of the tire model parameters along with the change of the vehicle speed into a vehicle state estimation scheme based on unscented Kalman filtering to obtain real-time vehicle state estimation. The tire model and parameter identification algorithm provided by the invention fully considers the nonlinear and transverse and longitudinal dynamics coupling relation of the vehicle, and can realize more reliable and accurate vehicle dynamics parameter identification by adaptively adjusting the tire model parameters through the optimization algorithm.

Description

A Vehicle State Estimation Method Based on Adaptive Tire Model Parameters

technical field

The invention relates to the technical field of vehicle state estimation, in particular to a vehicle state estimation method for time-varying tire model parameters.

Background technique

Accurate prediction and estimation of vehicle state is critical to the performance of vehicle control systems, such as autonomous driving, vehicle dynamics control, and chassis control. The vehicle's longitudinal velocity, lateral velocity and yaw rate directly determine the vehicle's motion state. The vehicle state estimation algorithm uses the existing vehicle model and sensor signal synthesis to estimate the above-mentioned vehicle motion state. Therefore, the accuracy of the vehicle model directly affects the accuracy of the vehicle state estimation. Due to the many disturbances in the real vehicle operation process, the accuracy of the vehicle model is affected by many aspects. For example, as the vehicle speed changes, the key parts of the vehicle model (tyre model parameters) also change, resulting in a decrease in the accuracy of the vehicle model with fixed parameters. Tire dynamics is an important part of vehicle dynamics. The lateral force and longitudinal force on the ground of the vehicle act on the vehicle system through the tires. However, tires are strongly nonlinear, and their performance is easily affected by driving conditions. Reliable parameter identification is currently a difficult problem that restricts the further improvement of vehicle control performance. At present, the main evaluation method for the identification of vehicle dynamics parameters is to obtain the true value of the parameter through experiments, and compare it with the estimated value of the parameter, so as to realize the optimization of the identification method. The main problem of this method lies in the two aspects of truth value acquisition and parameter identification algorithm:

1) Acquisition of the true value: the true value of the tire parameters is not suitable to be obtained through the test, coupled with the idealization of the test conditions, which deviates from the parameter identification requirements in the actual driving scene of the vehicle, so the acquisition of the true value has great uncertainty; When obtaining the true value of tire parameters, it is often necessary to use a large number of different types of sensors, which greatly increases the cost;

2) Parameter identification algorithm: Most parameter identification algorithms do not model the vehicle’s transverse and longitudinal dynamics, and often ignore the nonlinear characteristics of the vehicle system, and the accuracy of the model is not high; parameter optimization methods are prone to local optimal solutions , premature convergence and other issues.

Contents of the invention

The object of the present invention is to provide a vehicle state estimation method based on tire model parameter adaptation in order to overcome the above-mentioned defects in the prior art.

The purpose of the present invention can be achieved through the following technical solutions:

A vehicle state estimation method based on tire model parameter adaptation, the method comprising:

Collect real vehicle data, including measuring the inherent parameters of the vehicle and collecting vehicle motion parameters for typical working conditions of the vehicle;

Through the calculation of the vehicle load transfer model and the wheel center velocity, the tire empirical model is established; the vehicle state estimation scheme based on the unscented Kalman filter UKF is established;

Screen the data fragments that meet the continuous excitation conditions under typical working conditions and perform data memory;

Based on the memory data fragments, the particle swarm optimization is used to identify the tire model parameters, and the relationship between the tire model parameters and the vehicle speed is obtained by fitting the optimal tire model parameters under different vehicle speeds;

The relationship between the tire model parameters and the vehicle speed is substituted into the vehicle state estimation scheme based on the unscented Kalman filter UKF, and the real-time estimation of the vehicle state is obtained.

Further, according to the distribution of the vertical load at each wheel position, the measured longitudinal vehicle speed and the calculated wheel center speed, the tire empirical model obtains the relationship between the vertical force and the longitudinal force and lateral force on the tire. Relationship:

Among them, r _w is the rolling radius of the wheel; ω _i is the speed of the i-th wheel; v _x,i and v _y,i are the longitudinal velocity and lateral velocity of the i-th wheel center respectively; s _i,i ,s _{y ,i are longitudinal slip} rate and lateral slip _{rate ;} tire model parameters; F _z,i is the vertical load distribution vector of the i-th wheel.

Further, the vehicle load transfer model obtains the relationship between the acceleration of the vehicle and the vertical force of each wheel according to the collected vehicle acceleration information and the measured intrinsic parameter information of the vehicle:

F _z ＝(θ ^T ·(θ·θ ^T ) ^-1 )·[a _x ，h _g -a _y ·h _g g] ^T m where, F _z is the vertical load distribution vector of the wheel, θ is the vehicle Size parameter matrix, a _x is the longitudinal acceleration of the vehicle, a _y is the lateral acceleration of the vehicle, h _g is the height of the center of mass of the vehicle, g is the acceleration of gravity, and m is the mass of the vehicle.

Further, the wheel center speed includes the longitudinal speed and lateral speed of each wheel center,

The longitudinal velocity and lateral velocity of each wheel center are obtained by calculating the measured vehicle front wheel angle, vehicle yaw rate, vehicle longitudinal velocity, lateral velocity and vehicle size parameters.

Further, the establishment of the vehicle state estimation scheme based on the Unscented Kalman Filter UKF, the specific steps are as follows:

Carry out the definition of UKF parameters, the parameters include the vehicle's state vector, control vector, measurement vector, algorithm hyperparameters and noise;

Establish the UKF vehicle state space equation;

According to the UKF parameters defined above and the established UKF vehicle state space equation, the UKF-based vehicle state estimation is performed, and the estimated value of the longitudinal velocity and the lateral velocity of the vehicle are output.

Further, the establishment of the UKF vehicle state space equation, the specific steps are as follows:

Based on the vehicle load and the measured vehicle speed information, combined with the inherent parameters of the vehicle, calculate the air resistance F _a,x and the rolling resistance F _f,i of the wheels during the running of the vehicle:

F _a,x ＝0.5·ρ _a ·c _D ·A·v _x ²

F _f,i = f·F _z,i

Among them, ρ _a is the air density; c _D is the drag coefficient; A is the frontal projected area of the vehicle; _f is the rolling resistance coefficient; 4 represents left front wheel, right front wheel, left rear wheel, right rear wheel;

According to the measured vehicle front wheel rotation angle δ, the wheel driving moment M _M,i , the wheel braking moment M _B,i , the inherent parameters of the vehicle and the calculated vehicle load, the differential equation for the vehicle state vector is obtained:

Among them, F _{x, i} , F _{y, i} are the longitudinal force and lateral force on the tire respectively, and i respectively take 1, 2, 3, 4 to represent the left front wheel, right front wheel, left rear wheel, and right rear wheel; δ is the front wheel rotation angle of the vehicle; v _x , v _y are the longitudinal velocity and lateral velocity of the vehicle respectively; w _z is the yaw rate of the vehicle; a is the distance between the center of mass of the vehicle and the front axle; b is the center of mass of the vehicle The distance from the rear axle; B is half the distance between the left and right wheels of the vehicle; M _M,i is the driving moment of the wheel; I _z is the moment of inertia of the vehicle; J _ω is the moment of inertia of the wheel.

Further, when the positioning accuracy of the high-precision navigation system is good, use the high-precision navigation system to correct the tire model parameters in the vehicle speed estimation algorithm. The specific steps are as follows:

According to the defined vehicle state vector and UKF vehicle state estimation, the state estimation bias function J _c is evaluated:

where _cy is the lateral velocity estimation error weight; are respectively the estimated value of the longitudinal velocity and the estimated value of the lateral velocity of the vehicle estimated by the UKF vehicle state; /> are the longitudinal velocity and lateral velocity of the vehicle output by the high-precision inertial navigation system, respectively.

Further, the screening of data fragments that meet the continuous excitation conditions under typical working conditions and performing data memory, the specific process is as follows:

Select the constant speed lane change condition, divide the vehicle speed into segments at set speed intervals, and set the tire model parameters in the segments as constant values;

Obtain the distance from the vehicle to the lane line;

Judging the time point T _chang when the vehicle crosses the lane line according to the change of the lateral distance;

Use T _chang to search forward for the time T _start when the absolute value of the lateral acceleration is less than the set value, and the time T _start is the starting point of the lane change segment;

Use T _change to search backward for the time T _end when the absolute value of the lateral acceleration is less than the set threshold, and the time T _end is the end point of the lane change segment;

Extract the data of the time segment from T _start to T _end in the historical data;

According to the extracted lane-changing data segment, calculate the average longitudinal velocity, the absolute value of the maximum lateral acceleration and the time T _chang when the vehicle crosses the lane line, and calculate the priority of each data segment;

Save the set number of data fragments with the highest priority in each speed section for parameter identification.

Further, the particle swarm optimization is used to identify the tire model parameters, and the relationship between the tire model parameters and the vehicle speed is obtained by fitting the optimal tire model parameters under different vehicle speeds. The specific steps are as follows:

The tire model parameters and the distance from the center of mass to the front axle are used as the optimized particles of PSO;

Set the total number of initial particles and the maximum number of iterations;

Initialize the initial position and velocity of each particle through the initialization function;

In the process of each iteration, the fitness function is evaluated according to the current position of the particle;

In each round of iteration, by calculating the fitness function value of each particle, the historical optimal position of the individual particle and the global optimal position of the particle swarm for the current iteration number are obtained. According to the gradient change direction of the fitness function, the current The latest velocity and latest position of the particle position change for the number of iterations:

When updating the speed in each iteration, the variable weight coefficient method is used to linearly change the weight coefficient;

According to the above steps, when the number of iterations reaches the maximum number of iterations, the optimal particle position under the current working condition is obtained, that is, the optimal model parameter under this working condition;

According to the optimal model parameters under different working conditions as the identification point, the relationship between tire model parameters and longitudinal velocity is obtained by using polynomial fitting method.

Further, when evaluating the fitness function, parameter matching and calculation are performed according to the current particle state and the UKF vehicle state estimation algorithm, and the vehicle state deviation estimation function is used as the cost item for tire model parameter optimization:

J ⁱ (j)＝p ₀ J _c +p ₁ ||a ⁱ (j)-a _reference || ²

Among them, J ⁱ (j) is the fitness function value of the i-th particle in the j-th iteration; p ₀ and p ₁ are the corresponding weight coefficients; J _c is the accuracy cost of parameter estimation based on UKF; a ⁱ ( j) is the distance from the center of mass to the front axis of the j-th iteration of the i-th particle; a _reference is the reference value of the distance from the center of mass to the front axis.

Compared with the prior art, the present invention has the following beneficial effects:

1) When designing the tire model parameter identification algorithm, fully consider the nonlinearity of the vehicle, the coupling relationship between transverse and longitudinal dynamics, and the global search capability of the optimization algorithm, etc., so that more reliable and accurate identification of vehicle dynamics parameters can be achieved.

2) The identification of vehicle dynamics parameters in the present invention is mainly used for the estimation of the vehicle state, and is oriented to the state estimation requirements in the actual driving scene of the vehicle. Using multiple sensors to realize the evaluation of the identification of dynamic parameters can greatly save manpower and material resources.

3) The present invention corrects the tire model parameters in the vehicle speed estimation algorithm when the positioning accuracy of the high-precision navigation system is good, thereby ensuring that the vehicle speed estimation algorithm can also provide a certain speed estimation accuracy when the high-precision navigation system fails.

Description of drawings

Fig. 1 is a structural schematic diagram of the present invention.

Fig. 2 is the relationship figure of tire model parameter C ₂ and longitudinal velocity;

Fig. 3 is the effect diagram of lateral vehicle speed estimation;

Fig. 4 is a schematic diagram of the change process of the tire model during the simulation process.

Detailed ways

The present invention will be described in detail below in conjunction with the accompanying drawings and specific embodiments. This embodiment is carried out on the premise of the technical solution of the present invention, and detailed implementation and specific operation process are given, but the protection scope of the present invention is not limited to the following embodiments.

Example 1

The invention combines the particle swarm optimization algorithm and the unscented Kalman filter algorithm, aims at the vehicle response state information under typical working conditions, and aims at the estimation accuracy of the optimal vehicle state information to realize the dynamic identification of vehicle dynamic parameters, thereby providing a vehicle model The establishment and operation of the vehicle control system provide more reliable vehicle dynamic parameters. It mainly includes the following steps:

Step 1: Collect real vehicle data

Firstly, the inherent parameters of the vehicle are measured, including size parameters, mass, center of mass height, etc. Aiming at the typical working conditions of the vehicle, multi-sensors are used to measure the vehicle's three-axis velocity, angular velocity, acceleration (longitudinal acceleration a _x , lateral acceleration a _y ), longitudinal velocity v _x , lateral velocity v _y , yaw rate w _z , wheel Wheel speed, motor driving torque, braking torque and other data are collected to provide data support for subsequent research on the relationship between tire model parameters and vehicle speed.

Step 2: Establish a vehicle state estimation module based on UKF

Based on the real vehicle data in step 1, the tire empirical model is established by modeling the vehicle load transfer, wheel center velocity calculation, etc. Referring to the unscented Kalman filter estimation algorithm, the vehicle state estimation schemes of the vehicle's state vector x, control vector u, and measurement vector y are respectively established.

2.1 Establish vehicle load transfer module

According to the vehicle acceleration information collected in step 1 and the measured intrinsic parameter information of the vehicle, the relationship between the acceleration of the vehicle and the vertical force of each wheel is deduced.

F _z ＝[F _z,1 F _z,2 F _z,3 F _z,4 ] ^T

θ·F _z ＝[a _x ·h _g -a _y ·h _g g] ^T ·m

F _z ＝(θ ^T ·(θ·θ ^T ) ^-1 )·[a _x ·h _g -a _y ·h _g g] ^T ·m

Among them, F _z is the vertical load distribution vector of the wheel, F _{z, i} is the vertical load of the i-th wheel, and i takes 1, 2, 3, 4 to represent the left front wheel, right front wheel, left rear wheel, Right rear wheel; θ is the vehicle size parameter matrix; a is the distance between the center of mass of the vehicle and the front axle; b is the distance between the center of mass of the vehicle and the rear axle; B is half the distance between the left and right wheels of the vehicle _; The height of the center of mass; g is the acceleration of gravity; m is the mass of the vehicle.

2.2 Calculation of wheel center speed

According to the vehicle's front wheel rotation angle δ measured in step 1, the vehicle's yaw rate w _z , the vehicle's longitudinal velocity v _x , lateral velocity v _y , and the size parameters applied in step 2.1, calculate the center of each wheel Longitudinal velocity v _x,i and transverse velocity v _y,i .

Among them, v _{x, i} is the longitudinal velocity of the center of the wheel, and i respectively take 1, 2, 3, 4 to represent the left front wheel, the right front wheel, the left rear wheel, and the right rear wheel; v _{y, i} is the lateral velocity of the wheel;

2.3 Establish tire empirical model

According to the distribution of the vertical load at each wheel position obtained in step 2.1, the longitudinal vehicle speed obtained in the test in step 1, and the speed information of the wheel calculated in step 2.2, the vertical force and longitudinal force on the tire are obtained, Relationship between lateral forces.

Among them, r _w is the rolling radius of the wheel; ω _i is the speed of the i-th wheel; s _{x, i} , s _{y, i} are the longitudinal slip rate and lateral slip rate respectively; s _i is the comprehensive slip rate; F _{x, i} , F _{y, i} are the longitudinal force and lateral force on the tire respectively; c ₁ , c ₂ are the tire model parameters, which are mainly related to the driving conditions and are the vehicle dynamic parameters to be optimized in the present invention.

2.4 Definition of UKF parameters

Establish the vehicle's state vector x, control vector u, and measurement vector y:

x＝[v _x v _y w _z ω ₁ ω ₂ ω ₃ ω ₄ ]

u=[δ M _{M, 1} M _{M, 2} M _{M, 3} M _{M, 4} F _{z, 1} F _{z, 2} F _{z, 3} F _{z, 4} ]

y＝[a _x a _y w _z ω ₁ ω ₂ ω ₃ ω ₄ ]

Among them, M _M,i is the motor torque of the wheel.

Set the hyperparameters α=0.7, κ=3, β=2 for the sigma point assignment; and calculate the hyperparameters accordingly

λ=α·α(n _x +κ)-n _x

where n _x is the dimension of the state vector. The weight coefficient of the sigma point and the weight coefficient of the distribution covariance of the sigma point are then calculated based on the above hyperparameters.

Set the noise of the state transition and the noise of the measurement process:

R＝diag([0.01 0.01 0.01 0.1 0.1 0.1 0.1]) ²

Q＝diag([0.00001 0.0001 0.0001 0.0001 0.0001 0.0001 0.0001]) ²

2.5 UKF Vehicle State Space Equation

First, based on the wheel vertical load obtained in step 2.1 and the vehicle speed information measured in step 1, combined with the inherent parameters of the vehicle model in step 1, calculate the air resistance F _a,x and the rolling resistance F _f of the wheel during driving _{, i} .

F _a,x ＝0.5·ρ _a ·c _D ·A·v _x ²

F _f,i = f·F _z,i

Among them, ρ _a is the air density; c _D is the drag coefficient; A is the projected area of the front of the car; f is the rolling resistance coefficient.

According to the vehicle front wheel rotation angle δ obtained in step 1, the driving torque M _{M,i of} the wheel; the braking torque M _B,i of the wheel; the inherent parameters of the vehicle, the force on the wheel calculated in step 2.3 can be used to calculate the vehicle state vector The differential equation is calculated:

Among them, I _z is the moment of inertia of the vehicle; J _ω is the moment of inertia of the wheel.

2.6 UKF vehicle state estimation algorithm

According to the state space equation and UKF algorithm parameters in the above steps, the vehicle state estimation based on UKF is performed, and the estimated value of the longitudinal velocity of the vehicle is output and estimates of lateral velocity/>

2.7 Evaluation of State Estimation Bias

According to the state vector of the vehicle defined in step 2.4, the UKF vehicle state estimation in step 2.6 is used to evaluate the state estimation bias function _Jc :

Among them, c _y is the weight of the lateral velocity estimation error, because the variation ranges of the longitudinal velocity and the lateral velocity are different in order of magnitude, the weight is taken as 10000; Vehicle longitudinal velocity and lateral velocity output by high-precision inertial navigation system.

Step 3: Establish a typical working condition screening and data memory module

Due to different working conditions, the optimal values of vehicle parameters are different. In order to obtain the change rule of the tire model parameters under different working conditions, this embodiment takes the influence of vehicle speed on the tire model parameters as an example to illustrate that this method can realize the identification of the optimal parameters of the tire model under different working conditions, and establish the working condition and Variation relationship of the tire model.

Typical parameter identification methods require an external input excitation signal, and then perform a parameter identification algorithm according to the variation law of the input and output signals. Considering the safety and comfort of the vehicle during driving, it is impossible to add an excitation signal to the vehicle system. At the same time, in order to meet the continuous excitation conditions, the present invention designs a typical working condition screening and data memory algorithm. Working condition fragments, and store them in the data memory module. The data memory module has designed a data update mechanism, comprehensively sorts data fragments according to vehicle performance and data recording time, and retains the three data fragments with the highest priority in each speed range. In this embodiment, the influence of the vehicle speed on the tire model parameter _c2 is studied in a targeted manner, and the data segments satisfying the continuous excitation condition are screened out according to the actual driving process of the vehicle. The specific process is as follows:

3.1 Longitudinal speed division

In order to fit the optimal value of the tire model parameters at different speeds, try to choose the constant speed lane change condition, but the speed cannot be guaranteed to maintain a constant value during the driving process. Therefore, in this embodiment, the vehicle speed is segmented at 10 km/h speed intervals. And assume that the tire model parameters in the segment are constant. Therefore, lane change segments at different speeds are divided into different longitudinal speed segments.

3.2 Segment Extraction of Lane Changing Conditions

First, the distance between the vehicle and the lane line is obtained based on the sensor system, and the time point T change when the vehicle crosses the lane line can be judged according to the change of the lateral distance; then, the time point T _change when the absolute value of the lateral acceleration is ^less than 0.2m/s is searched forward by T _{change start} , T _start is the starting point of the lane change segment; then use T _change to search backward when the absolute value of the lateral acceleration is less than 0.2m/s ² T _end , T _end is the end point of the lane change segment.

Secondly, extract the data of the time segment from T _start to T _end in the historical data.

3.3 Lane changing data memory module

According to the lane-changing data segment extracted in step 3.2, calculate the average longitudinal velocity v _x,mean of the segment, the absolute value of the maximum lateral acceleration ||a _y || _max and T _change when the vehicle crosses the lane line, etc.

Due to the limited storage space, all data cannot be stored, so this embodiment will comprehensively consider the absolute value of the maximum lateral acceleration and the moment when the vehicle crosses the lane line, and filter out the three data segments with the highest priority for each speed segment.

Among them, R is the priority value of each data segment, S ₁ is the weight of the maximum lateral velocity, and S ₂ is the weight of the lane change time of the data segment. The priority of each data segment can be calculated according to the above formula, and only the 3 data segments with the highest priority in each speed segment are saved for parameter identification.

Step 4: Identification of tire model parameters based on particle swarm optimization

During the actual driving process of the vehicle, through the third step, the data fragments satisfying the continuous excitation conditions under different working conditions are accumulated. The tire model parameters are used as the particles of particle swarm optimization, and the changes of the tire model parameters are represented by the position and speed of the particles. The reference values of the vehicle dynamics parameters are used to establish the fitness function. In each round of iteration, it is necessary to calculate the fitness function of each particle, and then obtain the historical optimal value of individual particles and the global optimal value of the particle swarm. According to the gradient change direction of the fitness function, calculate the The latest speed and latest position of particle position changes, realizing round-by-round iterations. When the maximum number of iterations is reached, the identification of the optimal dynamic parameters for this working condition is completed. Then, according to the optimal tire model parameters under different vehicle speeds, the relationship between the tire model parameters and the vehicle speed is obtained by fitting.

4.1 Particle Swarm Optimization (PSO) Initialization

When the road conditions remain unchanged, the tire model parameter c ₁ is basically a constant value, and the vehicle speed has a greater impact on the tire model parameter c ₂ . At the same time, considering that the position of the center of mass of the vehicle may change during motion, the present invention uses the tire model parameter c ₂ and the distance a from the center of mass to the front axle as the optimization particles of PSO, that is, x _p =[c ₂ a]. The total number of initial particles is set to N, the maximum number of iterations is set to j _max , and the initial position and velocity of each particle are initialized through the initialization function to obtain is the initial position of the i-th particle; /> is the initial velocity of the i-th particle.

4.2 Calculation of fitness function

During each iteration, the fitness function needs to be evaluated according to the current position of the particle. When evaluating the fitness function of tire model parameter optimization, the parameters are matched and calculated according to the current particle state and the UKF vehicle state estimation algorithm in step 2.6, and then based on the vehicle state deviation estimation function J _c in step 2.7, as The cost term for tire model parameter optimization:

J ⁱ (j)＝p ₀ J _c +p ₁ ||a ⁱ (j)-a _reference || ²

Among them, J ⁱ (j) is the fitness function value of the i-th particle in the j-th iteration; p ₀ and p ₁ are the corresponding weight coefficients; J _c is the accuracy cost of parameter estimation based on UKF; a ⁱ ( j) is the distance from the center of mass to the front axis of the j-th iteration of the i-th particle; a _reference is the reference value of the distance from the center of mass to the front axis.

4.3 Update particle position and velocity

In each round of iteration, by calculating the fitness function value of each particle, the optimal position of the individual particle history of the current iteration number can be obtained The global optimal position of the particle swarm G _best (j), according to the gradient change direction of the fitness function, calculates the latest speed and position of the particle position change for the current number of iterations:

in, is the position of the i-th particle in round j, j+1; v ⁱ (j), v ⁱ (j+1) is the velocity of the i-th particle in round j, j+1; p(j) is Inertia factor; k ₁ , k ₂ are learning factors; r ₁ , r ₂ are random numbers between (0,1).

4.4 Linear variable weight coefficient

When updating the speed in each iteration, in order to solve the premature problem of the algorithm, the variable weight coefficient method is used to linearly change the weight coefficient:

Among them, p(j) represents the weight coefficient in the j-th iteration; p _max and p _min are the maximum and minimum values of the weight coefficient; j _max is the maximum number of iterations.

4.5 In the iterative process, perform the cycle according to steps 4.2-4.4. When the number of iterations j reaches the maximum number of iterations j _max , the optimal particle position under the current working condition is obtained, that is, the optimal model parameter x _{p under this working condition, best} = [c ₂ a].

4.6 As shown in Figure 2, according to the optimal model parameters under different working conditions are the identification points in the figure, and then use the polynomial fitting method to obtain the relationship between the tire model parameter c ₂ and the longitudinal speed.

Step 5: Real-time estimation of vehicle state

Using the relationship between the tire model parameters obtained in step 4 and the change of vehicle speed, it is substituted into step 2 to improve the accuracy of real-time estimation of the vehicle state. The change process of the tire model parameters during the simulation process is shown in Figure 4. In order to compare the advantages of the method proposed by the present invention, two UKF algorithms with fixed vehicle model parameters are selected, wherein the tire model parameter C _of UKF1 is 12, and the tire model parameter _C of UKF2 is 30, as shown in the simulation effect figure 3, from It can be seen from the figure that the lateral vehicle speed estimation accuracy of the method of the present invention is obviously better than the other two estimation algorithms.

The preferred specific embodiments of the present invention have been described in detail above. It should be understood that those skilled in the art can make many modifications and changes according to the concept of the present invention without creative efforts. Therefore, all technical solutions that can be obtained by those skilled in the art based on the concept of the present invention through logical analysis, reasoning or limited experiments on the basis of the prior art shall be within the scope of protection defined by the claims.

Claims (10)

1. A method for estimating a vehicle state based on tire model parameter adaptation, the method comprising:

collecting real vehicle data, including measuring intrinsic parameters of the vehicle and collecting vehicle motion parameters for typical working conditions of vehicle running;

building a tire experience model by calculating a vehicle load transfer model and a wheel center speed; establishing a vehicle state estimation scheme based on unscented Kalman filtering UKF;

screening data fragments meeting the continuous excitation condition under the typical working condition and memorizing the data;

based on the memorized data segments, performing tire model parameter identification by adopting particle swarm optimization, and fitting according to the optimal tire model parameters in different vehicle speed sections to obtain the relationship of the tire model parameters along with the change of the vehicle speed;

and substituting the relation of the tire model parameters along with the change of the vehicle speed into a vehicle state estimation scheme based on unscented Kalman filtering UKF to obtain real-time vehicle state estimation.

2. The vehicle state estimation method based on the tire model parameter adaptation according to claim 1, wherein the tire empirical model obtains the relationship between the vertical force and the longitudinal force, the lateral force of the tire according to the distribution of the vertical load of each wheel position, the measured longitudinal vehicle speed and the calculated wheel center speed:

wherein r is _w Is the rolling radius of the wheel; omega _i Is the rotation speed of the ith wheel; v _x,i And v _y,i The longitudinal speed and the transverse speed of the ith wheel center are respectively; s is(s) _x,i ,s _y,i The longitudinal slip rate and the transverse slip rate are respectively; s is(s) _i Is the comprehensive slip rate; f (F) _x,i ,F _y,i The longitudinal force and the transverse force respectively applied to the tire; c ₁ ,c ₂ Respectively tire model parameters; f (F) _z,i Is the vertical load distribution vector of the ith wheel.

3. The vehicle state estimation method based on tire model parameter adaptation according to claim 2, wherein the vehicle load transfer model obtains the relationship between the acceleration of the vehicle and the vertical force of each wheel according to the acquired information of the acceleration of the vehicle and the measured intrinsic parameter information of the vehicle:

F _z ＝(θ ^T ·(θ·θ ^T )-1)·[a _x ·h _g -a _y ·h _g g] ^T ·m

wherein F is _z Is the vertical load distribution vector of the wheel, theta is the vehicle size parameter matrix, a _x For vehicle longitudinal acceleration, a _y For vehicle lateral acceleration, h _g G is gravitational acceleration, and m is the mass of the vehicle.

4. A vehicle condition estimation method based on tire model parameter adaptation as in claim 2, wherein the wheel center speeds include a longitudinal speed and a lateral speed of each wheel center,

the longitudinal speed and the transverse speed of each wheel center are calculated by measuring the front wheel rotation angle of the vehicle, the yaw rate of the vehicle, the longitudinal speed and the transverse speed of the vehicle and the vehicle size parameters.

5. The vehicle state estimation method based on tire model parameter adaptation according to claim 1, wherein the building of the vehicle state estimation scheme based on unscented kalman filter UKF comprises the following specific steps:

performing UKF parameter definition, wherein the parameters comprise a state vector, a control vector, a measurement vector, an algorithm super-parameter and noise of the vehicle;

establishing a UKF vehicle state space equation;

and according to the defined UKF parameters and the established UKF vehicle state space equation, estimating the vehicle state based on the UKF, and outputting an estimated value of the longitudinal speed and an estimated value of the transverse speed of the vehicle.

6. The method for estimating a vehicle state based on tire model parameter adaptation according to claim 5, wherein the building of the UKF vehicle state space equation comprises the following specific steps:

based on the vehicle load and the measured vehicle speed information, the air resistance F of the vehicle in the running process is calculated by combining the inherent parameters of the vehicle _a,x And rolling resistance F of the wheel _f,i ：

F _a,x ＝0.5·ρ _a ·c _D ·A·v _x ²

F _f,i ＝f·F _z,i

Wherein ρ is _a Is air density; c _D Is the wind resistance coefficient; a is the front projection area of the locomotive; f is the rolling resistance coefficient; f (F) _z,i Taking 1,2,3 and 4 as vertical loads of the ith wheel respectively to represent a left front wheel, a right front wheel, a left rear wheel and a right rear wheel;

based on the measured front wheel angle delta, the driving moment M of the wheels _M,i Braking moment M of wheel _B,i And obtaining a differential equation of the vehicle state vector by the vehicle intrinsic parameters and the calculated vehicle load:

wherein F is _x,i ,F _y,i Respectively taking 1,2,3 and 4 as the left front wheel, the right front wheel, the left rear wheel and the right rear wheel; delta is the front wheel corner of the vehicle; v _x ，v _y Longitudinal speed and transverse speed of the vehicle respectively; w (w) _z Is the yaw rate of the vehicle; a is the distance from the center of mass of the vehicle to the front axle; b is the distance from the center of mass of the vehicle to the rear axle; b is half of the distance between the wheels on the left side and the right side of the vehicle; m is M _M,i Is the driving torque of the wheels; i _z Is the rotational inertia of the vehicle; j (J) _ω Is the moment of inertia of the wheel.

7. The method for estimating a vehicle state based on adaptive tire model parameters according to claim 5, wherein when the positioning accuracy of the high-accuracy navigation system is good, the tire model parameters in the vehicle speed estimation algorithm are corrected by the high-accuracy navigation system, comprising the following steps:

based on the defined vehicle state vector and UKF vehicle state estimation, performing state estimation deviation function J _c Is determined by the following steps:

wherein c _y Estimating an error weight for the lateral velocity;an estimated value of a longitudinal speed and an estimated value of a lateral speed of the vehicle, which are respectively estimated from the UKF vehicle state; />The longitudinal speed and the lateral speed of the vehicle are respectively output by the high-precision inertial navigation system.

8. The vehicle state estimation method based on tire model parameter adaptation according to claim 1, wherein the screening of the data segments meeting the continuous excitation condition under the typical working condition and the data memorization are performed comprises the following specific processes:

selecting a uniform speed lane change working condition, segmenting the vehicle speed at set speed intervals, and setting tire model parameters in the segments as fixed values;

obtaining the distance between a vehicle and a lane line;

judging the time point T of the vehicle crossing the lane line according to the lateral distance change _change ；

By T _chang Time T when the absolute value of the forward search lateral acceleration is smaller than the set value _start ，T _start The moment is the starting point of the channel changing segment;

by T _change Searching backward for a time T when the absolute value of the lateral acceleration is smaller than the set threshold _end ，T _end The moment is the termination point of the lane change segment;

extracting T from historical data _start ～T _end Data of the time slice;

according to the extracted lane change data segment, calculating the average longitudinal speed, the maximum lateral acceleration absolute value and the moment T of the vehicle crossing the lane line of the segment _change And calculating the priority of each data segment;

and storing the data fragments with the highest priority in each vehicle speed section for parameter identification.

9. The vehicle state estimation method based on tire model parameter adaptation according to claim 7, wherein the tire model parameter identification is performed by adopting particle swarm optimization, and the relationship of the tire model parameter along with the change of the vehicle speed is obtained according to the optimal tire model parameter fitting in different vehicle speed sections, and the specific steps are as follows:

taking tire model parameters and the distance from the mass center to the front axle as PSO optimization particles;

setting the total number of initial particles and the maximum iteration number;

initializing the initial position and speed of each particle through an initialization function;

in the process of each iteration, evaluating the fitness function according to the current position of the particle;

in each iteration, calculating an fitness function value of each particle to obtain a particle individual history optimal position and a particle swarm global optimal position of the current iteration number, and calculating the latest speed and the latest position of the particle position change of the current iteration number according to the gradient change direction of the fitness function:

when the speed is updated in each iteration, a method of variable weight coefficients is adopted to linearly change the weight coefficients;

the steps are circulated according to the steps, and when the iteration times reach the maximum iteration times, the optimal particle positions under the current working condition, namely the optimal model parameters under the working condition, are obtained;

and obtaining a relation of the tire model parameters along with the longitudinal speed by using a polynomial fitting method according to the optimal model parameters under different working conditions as identification points.

10. A vehicle state estimation method based on tire model parameter adaptation as in claim 9,

when the fitness function is evaluated, parameter matching and calculation are carried out according to the current particle state and the UKF vehicle state estimation algorithm, and the vehicle state deviation estimation function is used as a cost item for optimizing the tire model parameters:

J ⁱ (j)＝p ₀ ·J _c +p ₁ ||a ⁱ (j)-a _reference || ²

wherein J is ⁱ (j) The fitness function value of the jth iteration of the ith particle is the fitness function value of the jth iteration of the ith particle; p is p ₀ 、p ₁ Respectively corresponding weight coefficients; j (J) _c The cost of accuracy for parameter estimation based on UKF; a, a ⁱ (j) The centroid-to-front axis distance value for the ith particle jth round of iteration; a, a _reference Is a reference value for the centroid to front axis distance.