CN111942399A - Vehicle speed estimation method and system based on unscented Kalman filtering - Google Patents

Abstract

The present application discloses a vehicle speed estimation method and system based on unscented Kalman filtering, and relates to the technical field of vehicle speed estimation. The vehicle speed estimation method includes the steps of: establishing a high-precision vehicle model, obtaining vehicle parameters; establishing multi-degree-of-freedom vehicle dynamics model, according to the vehicle parameters, determine the state equation and observation equation of the unscented Kalman filter, and determine the input quantity, state quantity and observation quantity of the unscented Kalman filter, the state quantity at least includes longitudinal Speed and lateral speed: Using the vehicle speed estimator, combined with the state quantity at the current moment, and the state equation and observation equation, estimate the longitudinal speed and lateral speed at the next moment. The vehicle speed estimation method and system based on the unscented Kalman filter of the present application can accurately estimate the real-time longitudinal speed and lateral speed of the vehicle during the running process of the vehicle by combining the vehicle speed estimator and the unscented Kalman filter algorithm, so as to improve the Accuracy of vehicle dynamics control.

Description

A method and system for vehicle speed estimation based on unscented Kalman filter

technical field

The present application relates to the technical field of vehicle speed estimation, and in particular to a vehicle speed estimation method and system based on unscented Kalman filtering.

Background technique

At present, in vehicle dynamics control, vehicle speed is a crucial state quantity, it is the basis for calculating wheel slip rate, and slip rate is the main control quantity in vehicle dynamics control. Due to the high cost of installing a sensor that directly measures the speed of the vehicle, in practical research and development, the vehicle speed is generally regarded as a state quantity that cannot be directly obtained, and various estimation algorithms are used to estimate the vehicle speed.

In the related art, the existing estimation algorithms include the maximum wheel speed method, the slope method and the comprehensive method. The above algorithms are simple in process, convenient in implementation, and low in cost, but their estimation accuracy is not high, and they are only suitable for traditional methods that do not require high control accuracy. vehicle. With the development of automobile intelligence and networking, modern automobiles require further improvement in the integrated control accuracy of vehicle dynamics. Therefore, it is urgent to redesign the mathematical model of vehicle speed estimation and the corresponding estimation algorithm.

There are inevitable process errors and observation errors in the process of vehicle speed estimation. Since KF (Kalman Filter) is a linear estimation algorithm, it can only solve the state estimation problem of linear systems under the linear Gaussian model, while practical engineering applications There are always different degrees of nonlinearity in the system in the vehicle system. For example, the tire in the vehicle system is a typical highly nonlinear system. Therefore, the basic Kalman filter algorithm is not suitable to be introduced into the field of vehicle state estimation, and the estimation error is large.

SUMMARY OF THE INVENTION

In view of one of the defects in the prior art, the purpose of the present application is to provide a vehicle speed estimation method and system based on unscented Kalman filtering to solve the problem of low vehicle speed estimation accuracy in the related art.

A first aspect of the present application provides a vehicle speed estimation method based on unscented Kalman filtering, comprising the steps of:

Build high-precision vehicle models and obtain vehicle parameters;

Establish a multi-degree-of-freedom vehicle dynamics model, determine the state equation and observation equation of the unscented Kalman filter according to the above vehicle parameters, and determine the input quantity, state quantity and observation quantity of the above unscented Kalman filter, the above state The quantities include at least longitudinal and lateral velocities;

Using the vehicle speed estimator, combined with the state quantity at the current moment, as well as the above state equation and observation equation, estimate the longitudinal speed and lateral speed at the next moment.

In some embodiments, before obtaining the vehicle parameters, the method further includes:

The vehicle controller outputs the braking pressure of each wheel to the above-mentioned high-precision vehicle model;

The above-mentioned high-precision vehicle model performs real-time simulation on the vehicle according to the braking pressure of the above-mentioned wheels, and obtains the above-mentioned parameters of the whole vehicle;

The above vehicle parameters include longitudinal acceleration, lateral acceleration, yaw rate, wheel speeds of the four wheels, and road adhesion coefficients at the four wheels.

In some embodiments, after obtaining the vehicle parameters, the method further includes:

The above-mentioned high-precision vehicle model outputs the wheel speeds of the four wheels to the above-mentioned vehicle controller.

In some embodiments, the above-mentioned input quantities include the wheel speeds of the four wheels, the above-mentioned state quantities also include longitudinal acceleration, lateral acceleration, yaw rate, and road adhesion coefficients at the four wheels, and the above-mentioned observation quantities include longitudinal acceleration, lateral acceleration, and yaw rate.

In some embodiments, the above-mentioned vehicle speed estimator is used, combined with the state quantity at the current moment, and the state equation and observation equation of the unscented Kalman filter to estimate the longitudinal speed and lateral speed at the next moment, specifically including:

Perform unscented transformation on the state quantity at the current moment to obtain multiple sigma points, and calculate the weight of each sigma point;

According to the above state equation and observation equation, calculate the prior state estimate, prior error covariance and prior observation estimate;

According to the prior state estimate, prior error covariance, and prior observation estimate, the next observation and error covariance are updated.

In some embodiments, the state equation and observation equation of the above-mentioned unscented Kalman filter are:

where x _k+1 is the state vector at time k+1, x _k is the state vector at time k, z _k is the observation vector at time k, uk is the input vector at time _{k, and w k is} the process noise at time k , v _k is the observation noise at time k.

In some embodiments, the above state equation is further:

为k时刻的纵向速度的导数，v(k)为k时刻的侧向速度，a_x(k)为k时刻的纵向加速度，a_y(k)为k时刻的侧向加速度，ω_r(k)为k时刻的横摆角速度，

为k时刻的横摆角速度的导数，μ_fl(k)为k时刻的左前轮路面附着系数，μ_fr(k)为k时刻的右前轮路面附着系数，μ_rl(k)为k时刻的左后轮路面附着系数，μ_rr(k)为k时刻的右后轮路面附着系数，T_s为k到k+1时刻之间的时间步长。where u(k) is the longitudinal velocity at time k,

is the derivative of the longitudinal velocity at time k, v(k) is the lateral velocity at time k, a _x (k) is the longitudinal acceleration at time k, a _y (k) is the lateral acceleration at time k, ω _r (k ) is the yaw rate at time k,

is the derivative of the yaw rate at time k, μ _fl (k) is the road adhesion coefficient of the left front wheel at time k, μ _fr (k) is the road adhesion coefficient of the right front wheel at time k, μ _rl (k) is the time k The left rear wheel road adhesion coefficient of , μ _rr (k) is the right rear wheel road adhesion coefficient at time k, and T _s is the time step between time k and time k+1.

In some embodiments, the multi-degree-of-freedom vehicle dynamics model is a vehicle-speed dependent seven-degree-of-freedom dynamic model.

A second aspect of the present application provides a vehicle speed estimation system based on unscented Kalman filtering, which includes:

a first modeling module, which is used to establish a high-precision vehicle model, and obtain vehicle parameters through the above-mentioned high-precision vehicle model;

The second modeling module is used to establish a multi-degree-of-freedom vehicle dynamics model, and according to the above vehicle parameters, determine the state equation and observation equation of the unscented Kalman filter, and determine the input of the unscented Kalman filter Quantities, state quantities and observation quantities, the above state quantities at least include longitudinal velocity and lateral velocity;

The vehicle speed estimator is used to estimate the longitudinal speed and lateral speed at the next moment by combining the state quantity at the current moment, and the state equation and observation equation of the unscented Kalman filter.

In some embodiments, it also includes:

a vehicle controller, which is used for outputting the braking pressure of each wheel to the above-mentioned first modeling module;

The above-mentioned high-precision vehicle model performs real-time simulation on the vehicle according to the braking pressure of the above-mentioned wheels, and obtains the above-mentioned parameters of the whole vehicle;

The above vehicle parameters include longitudinal acceleration, lateral acceleration, yaw rate, wheel speeds of the four wheels, and road adhesion coefficients at the four wheels.

The beneficial effects brought by the technical solution provided by this application include:

The vehicle speed estimation method and system based on the unscented Kalman filter of the present application performs real-time simulation of the vehicle through a high-precision vehicle model to obtain vehicle parameters, and then uses the multi-degree-of-freedom vehicle dynamics model to determine the vehicle parameters according to the vehicle parameters. The state equation and observation equation of the unscented Kalman filter, and determine the input quantity, state quantity and observation quantity of the unscented Kalman filter, and then use the vehicle speed estimator to combine the state quantity at the current moment, as well as the state equation and observation equation , to realize the estimation of the longitudinal speed and lateral speed at the next moment. Therefore, through the vehicle speed estimator combined with the unscented Kalman filter algorithm, the real-time longitudinal speed and lateral speed of the vehicle can be accurately estimated during the running process of the vehicle, To improve the precision of vehicle dynamics control.

Description of drawings

1 is a flowchart of a method for estimating vehicle speed based on unscented Kalman filtering provided by an embodiment of the present application;

2 is a schematic diagram of vehicle speed estimation based on unscented Kalman filtering in an embodiment of the application;

3 is a flowchart of step S3 in the embodiment of the application;

FIG. 4 is a diagram of a vehicle speed estimation result according to an embodiment of the application;

FIG. 5 is a diagram of a vehicle speed estimation error according to an embodiment of the present application.

Detailed ways

The present application will be further described in detail below with reference to the accompanying drawings and embodiments.

Referring to FIG. 1 , an embodiment of the present application provides a vehicle speed estimation method based on unscented Kalman filtering, which includes the steps:

S1. Establish a high-precision vehicle model and obtain vehicle parameters. Among them, the high-precision vehicle model is a high-precision vehicle model based on the simulation software CarSim, which is used for real-time simulation of real vehicles. The high-precision vehicle model has high degrees of freedom and accurate simulation results.

S2. Establish a multi-degree-of-freedom vehicle dynamics model, determine the state equation and observation equation of the unscented Kalman filter UKF (Unscented Kalman Filter) according to the above vehicle parameters, and determine the input of the above unscented Kalman filter, State quantity and observation quantity, the above state quantity includes at least longitudinal velocity and lateral velocity.

S3. Using the vehicle speed estimator, combined with the state quantity at the current moment, and the above-mentioned state equation and observation equation, estimate the longitudinal speed and lateral speed at the next moment.

The vehicle speed estimation method and system based on the unscented Kalman filter of the present application performs real-time simulation of the vehicle through a high-precision vehicle model to obtain vehicle parameters, and then uses the multi-degree-of-freedom vehicle dynamics model to determine the vehicle parameters according to the vehicle parameters. The state equation and observation equation of the unscented Kalman filter, and determine the input quantity, state quantity and observation quantity of the unscented Kalman filter, and then use the vehicle speed estimator to combine the state quantity at the current moment, as well as the state equation and observation equation , to realize the estimation of the longitudinal speed and lateral speed at the next moment. Therefore, through the vehicle speed estimator combined with the unscented Kalman filter algorithm, the real-time longitudinal speed and lateral speed of the vehicle can be accurately estimated during the running process of the vehicle, To improve the precision of vehicle dynamics control.

In this embodiment, the multi-degree-of-freedom vehicle dynamics model is a medium-precision vehicle model based on MATLAB/simulink, which runs in the vehicle speed estimator and is used for one-step prediction of the state by the vehicle speed estimator.

Preferably, the above-mentioned multi-degree-of-freedom vehicle dynamics model is a seven-degree-of-freedom dynamic model related to vehicle speed. Among them, the seven-degree-of-freedom dynamic model mainly considers the longitudinal motion, lateral motion, yaw motion of the body and the rotational motion of the four wheels, ignoring the pitch motion, roll motion and vertical motion of the vehicle.

Further, before acquiring the vehicle parameters in the above step S1, it also includes:

First, the vehicle controller outputs the braking pressure of each wheel to the above-mentioned high-precision vehicle model.

Then, the above-mentioned high-precision vehicle model simulates the vehicle in real time according to the braking pressure of each wheel to obtain the above-mentioned parameters of the entire vehicle.

The above vehicle parameters include longitudinal acceleration, lateral acceleration, yaw angular velocity, wheel speeds of the four wheels, and road adhesion coefficients at the four wheels during the running of the vehicle.

In this embodiment, after obtaining the vehicle parameters in the above step S1, the method further includes:

The above-mentioned high-precision vehicle model outputs the wheel speeds of the four wheels to the above-mentioned vehicle controller.

Further, the above-mentioned input quantities include the wheel speeds of the four wheels, the above-mentioned state quantities also include longitudinal acceleration, lateral acceleration, yaw rate, and the road adhesion coefficient at the four wheels, and the above-mentioned observations include longitudinal acceleration, lateral acceleration. , and the yaw rate.

Referring to Figure 2, the high-precision vehicle model outputs the vehicle's longitudinal acceleration, lateral acceleration, yaw rate, the wheel speed of the four wheels, and the road adhesion coefficient at the four wheels as the observed values, which are input into the vehicle speed estimator, Based on the multi-degree-of-freedom vehicle dynamics model, the vehicle speed estimator estimates the longitudinal speed and lateral speed at the next moment, and outputs it to the vehicle controller. The vehicle controller outputs the braking pressure of each wheel to the high-precision vehicle. The model forms a closed-loop control to complete the vehicle dynamics control. Optionally, the high precision vehicle model also outputs the steering wheel angle δ into the vehicle speed estimator. In this embodiment, the steering wheel angle δ is a default value.

In this embodiment, the nonlinear state equation and observation equation of the above-mentioned unscented Kalman filter are respectively:

where x _k+1 is the state vector at time k+1, x _k is the state vector at time k, z _k is the observation vector at time k, uk is the input vector at time _{k, and w k is} the process noise at time k , v _k is the observation noise at time k.

Referring to Fig. 3, in the above step S3, the vehicle speed estimator is used, combined with the state quantity at the current moment, and the state equation and observation equation of the unscented Kalman filter, to estimate the longitudinal speed and lateral speed at the next moment, specifically include:

A1. Initialize, obtain the initial state estimate and the initial error covariance matrix.

Since the process noise comes from the inaccuracy of the multi-degree-of-freedom vehicle dynamics model and the observation noise comes from the inaccuracy of the observation, it is assumed in the Kalman filter that the process noise w and the observation noise v are both Gaussian white noises with a mean value of 0. The covariance matrices of are Q and R respectively, that is, E[ww ^T ]=Q, E[vv ^T ]=R, and E[wv ^T ]=0.

A2. Perform unscented transformation on the state quantity at the current moment to obtain multiple sigma points, and calculate the weight of each sigma point.

In this embodiment, the unscented Kalman filter is based on the idea that it is easier to approximate the probability density distribution of the nonlinear function than to approximate the nonlinear function itself. Using unscented transformation UT (Unscented Transformation) to approximate the transfer of probability density, it is necessary to select (2n+1) sampling points, namely Sigma points. The requirements for Sigma point selection are its mean and covariance and the state distribution of the original system. The mean and covariance are equal. Then the Sigma point is brought into the state equation to obtain the sample points after propagation, and finally the statistical properties of the sample points after propagation are used to replace the statistical properties of the nonlinear system after propagation.

First, get 2n+1 Sigma points as follows:

为k-1时刻的状态预测值，

和

分别为k-1时刻第i个和第i+n个Sigma点(对称)，λ为缩放系数，且λ＝α²(n+κ)-n，α，β，κ均根据状态空间的实际状态确定。上述选取方法保证了第1和n+1，2和n+2…n和2n个点都在第0个点(即均值点)的周围，从而满足样本点均值和协方差与原系统状态分布的均值和协方差相等的要求。Among them, P _k-1|k-1 is the error covariance of the system, that is, the state space represented by the state equation at time k-1,

is the state prediction value at time k-1,

and

are the i-th and i+n-th Sigma point (symmetric) at the time k-1, respectively, λ is the scaling factor, and λ=α ² (n+κ)-n, α, β, κ are all based on the actual state space Status OK. The above selection method ensures that the 1st and n+1, 2 and n+2...n and 2n points are all around the 0th point (ie the mean point), so as to satisfy the mean and covariance of the sample points and the state distribution of the original system The mean and covariance are equal.

Then, calculate the weights of (2n+1) Sigma points as follows:

为第0个Sigma点的均值权值，

为第0个Sigma点的方差权值，W_i ^m为第i个Sigma点的均值权值，W_i ^c为第i个Sigma点的方差权值。in,

is the mean weight of the 0th Sigma point,

is the variance weight of the 0th Sigma point, W _i ^m is the mean weight of the i th Sigma point, and W _i ^c is the variance weight of the i th Sigma point.

A3. Time update, according to the above state equation and observation equation, calculate the prior state estimate, prior error covariance and prior observation estimate.

并对Sigma点的先验状态估计值

进行加权求和，得到先验状态估计值均值：First, perform the prediction calculation on the Sigma point to obtain the prior state estimate

and estimate the prior state of the Sigma point

Do a weighted summation to get the mean of the prior state estimates:

为k-1时刻对状态量在k时刻的先验估计值均值。in,

is the mean value of a priori estimates of the state quantity at time k at time k-1.

Then, the prior error covariance is calculated from the prior state estimate and the mean prior state estimate:

Among them, P _k|k-1 is the prior error covariance of the state quantity at time k-1 at time k.

Finally, compute the prior observation estimate:

为k-1时刻对k时刻Sigma点的先验观测值，

为k-1时刻对k时刻的先验观测值均值。in,

is the prior observation of the Sigma point at time k at time k-1,

is the mean of the prior observations at time k-1 to time k.

A4. Update the next observation value and the error covariance according to the above prior state estimate value, prior error covariance and prior observation estimate value.

First, calculate the correction matrix:

为观测协方差，

为预测观测互方差，K_k为卡尔曼增益。in,

is the observation covariance,

To predict the observed cross-variance, K _k is the Kalman gain.

Then, update the observations:

为对状态量在k时刻的观测值，即后验估计值。in,

is the observed value of the state quantity at time k, that is, the posterior estimate.

Finally, update the error covariance:

Among them, P _k|k is the error covariance of the observation value of the state quantity at time k, that is, the posterior error covariance.

中包括估算得到的k时刻的纵向速度和侧向速度。当在k时刻估算k+1时刻的状态量时，令k＝k+1，返回步骤A2，进行下一轮计算，即可估算得到k+1时刻的纵向速度和侧向速度，以此类推循环计算，即可得到每一时刻的纵向速度和侧向速度。At this time, the obtained observation value of the state quantity at time k

includes the estimated longitudinal and lateral velocities at time k. When estimating the state quantity at time k+1 at time k, let k=k+1, return to step A2, and perform the next round of calculation, the longitudinal speed and lateral speed at time k+1 can be estimated, and so on By cyclic calculation, the longitudinal speed and lateral speed at each moment can be obtained.

The state quantity is the state vector x=[uva _x a _y ω _r μ _fl μ _fr μ _rl μ _rr ] ^T , the above observation quantity is the observation vector z=[a _x a _y ω _r ] ^T , and the above input quantity is the input vector u =[ω _fl ω _fr ω _rl ω _rr ] ^T .

where u is the longitudinal velocity, v is the lateral velocity, a _x is the longitudinal acceleration, a _y is the lateral acceleration, ω _r is the yaw rate, ω _fl , ω _fr , ω _rl , and ω _rr are the left front wheel, respectively speed, right front wheel speed, left rear wheel speed, right rear wheel speed, μ _fl , μ _fr , μ _rl , μ _rr are the left front wheel road adhesion coefficient, right front wheel road adhesion coefficient, left rear wheel Road adhesion coefficient, right rear wheel road adhesion coefficient.

In this embodiment, some reasonable assumptions need to be made. The first assumption is that the acceleration will not change abruptly, and the second assumption that the road adhesion coefficient will not change abruptly. The above state equation is further:

为k时刻的纵向速度的导数，v(k)为k时刻的侧向速度，

为k时刻的侧向速度的导数，a_x(k)为k时刻的纵向加速度，a_y(k)为k时刻的侧向加速度，ω_r(k)为k时刻的横摆角速度，

为k时刻的横摆角速度的导数，μ_fl(k)为k时刻的左前轮路面附着系数，μ_fr(k)为k时刻的右前轮路面附着系数，μ_rl(k)为k时刻的左后轮路面附着系数，μ_rr(k)为k时刻的右后轮路面附着系数，T_s为k到k+1之间的时间步长。where u(k) is the longitudinal velocity at time k,

is the derivative of the longitudinal velocity at time k, v(k) is the lateral velocity at time k,

is the derivative of the lateral velocity at time k, a _x (k) is the longitudinal acceleration at time k, a _y (k) is the lateral acceleration at time k, ω _r (k) is the yaw rate at time k,

is the derivative of the yaw rate at time k, μ _fl (k) is the road adhesion coefficient of the left front wheel at time k, μ _fr (k) is the road adhesion coefficient of the right front wheel at time k, μ _rl (k) is the time k The left rear wheel road adhesion coefficient of , μ _rr (k) is the right rear wheel road adhesion coefficient at time k, and T _s is the time step between k and k+1.

In this embodiment, the UKF vehicle speed estimator model and the multi-degree-of-freedom vehicle dynamics model for state estimation are built in MATLAB/simulink, because the multi-degree-of-freedom vehicle dynamics model is only used to verify the estimation algorithm and does not involve the underlying execution Therefore, a first-order inertial system can be used to replace the entire braking system, the vehicle model parameters of the high-precision vehicle model are set in CarSim, and the ABS (Anti-lock Braking System, Anti-lock Braking System, Anti-lock Braking System) that comes with CarSim is used at the same time. System) algorithm for ABS control, select a straight and dry road with a medium adhesion coefficient in the road setting window, set the road adhesion coefficient to 0.5, and set the brake control to sudden braking, that is, suddenly apply a brake of 1.5Mpa at 0.25s. Dynamic pressure; initial braking speed is set to 70km/h.

Referring to Figure 4, taking the longitudinal vehicle speed as an example to verify, the high-precision vehicle model of Carsim outputs the real-time vehicle speed, and the estimated vehicle speed and the actual vehicle speed basically coincide during the simulation process. Referring to Fig. 5, it can be seen from the estimation error curve that the maximum estimation error during the deceleration process is 1.2km/h (the error after the vehicle speed is reduced to 0 is not considered), indicating that the vehicle speed estimation method of this embodiment can It can better estimate the real-time speed of the vehicle and ensure the stability of the ABS control process.

An embodiment of the present application further provides a vehicle speed estimation system based on unscented Kalman filtering, which includes a first modeling module, a second modeling module and a vehicle speed estimator.

The first modeling module is used to establish a high-precision vehicle model, and obtain vehicle parameters through the above-mentioned high-precision vehicle model.

The second modeling module is used to establish a multi-degree-of-freedom vehicle dynamics model, and the second modeling module is also used to determine the state equation and observation of the unscented Kalman filter according to the above vehicle parameters and the multi-degree-of-freedom vehicle dynamics model equation, and determine the input quantity, state quantity and observation quantity of the above-mentioned unscented Kalman filter, and the above-mentioned state quantity at least includes longitudinal velocity and lateral velocity.

The vehicle speed estimator is used to estimate the longitudinal speed and lateral speed at the next moment by combining the state quantity at the current moment, and the state equation and observation equation of the unscented Kalman filter.

Further, the above-mentioned vehicle speed estimation system further includes a vehicle controller.

The vehicle controller is used to output the braking pressure of each wheel to the above-mentioned first modeling module. Then, the high-precision vehicle model can simulate the vehicle in real time according to the braking pressures of the above-mentioned wheels, and obtain the above-mentioned parameters of the whole vehicle.

The above vehicle parameters include longitudinal acceleration, lateral acceleration, yaw rate, wheel speeds of the four wheels, and road adhesion coefficients at the four wheels. Correspondingly, the above-mentioned input quantities include the wheel speeds of the four wheels, the above-mentioned state quantities also include longitudinal acceleration, lateral acceleration, yaw angular velocity, and the road adhesion coefficient at the four wheels, and the above-mentioned observations include longitudinal acceleration, lateral acceleration. , and the yaw rate.

The vehicle speed estimation system of this embodiment is applicable to the above-mentioned vehicle speed estimation methods. In the whole estimation process, only the estimated value at the previous moment and the observed value at the current moment can be stored, and then the calculation can be performed cyclically to obtain the longitudinal speed and Lateral speed, to minimize the performance requirements of the on-board controller.

The present application is not limited to the above-mentioned embodiments. For those skilled in the art, without departing from the principles of the present application, several improvements and modifications can be made, and these improvements and modifications are also regarded as the protection of the present application. within the range.

Claims (10)

1. A vehicle speed estimation method based on unscented Kalman filtering is characterized by comprising the following steps:

establishing a high-precision vehicle model, and acquiring vehicle parameters;

establishing a multi-degree-of-freedom vehicle dynamics model, determining a state equation and an observation equation of an unscented Kalman filter according to the vehicle parameters, and determining input quantity, state quantity and observation quantity of the unscented Kalman filter, wherein the state quantity at least comprises longitudinal speed and lateral speed;

and estimating the longitudinal speed and the lateral speed at the next moment by utilizing a vehicle speed estimator and combining the state quantity at the current moment, the state equation and the observation equation.

2. The unscented kalman filter-based vehicle speed estimation method according to claim 1, wherein before acquiring the vehicle parameters, the method further comprises:

the vehicle control unit outputs the braking pressure of each wheel to the high-precision vehicle model;

the high-precision vehicle model simulates a vehicle in real time according to the brake pressure of each wheel to obtain the whole vehicle parameters;

the whole vehicle parameters comprise longitudinal acceleration, lateral acceleration, yaw rate, wheel speeds of four wheels and road surface adhesion coefficients at the four wheels.

3. The unscented kalman filter-based vehicle speed estimation method according to claim 2, wherein after acquiring the vehicle parameters, the method further comprises:

and the high-precision vehicle model outputs the wheel speeds of four wheels to the whole vehicle controller.

4. The unscented kalman filter-based vehicle speed estimation method according to claim 2, characterized in that:

the input quantities include wheel speeds of four wheels, the state quantities further include longitudinal acceleration, lateral acceleration, yaw rate, and road surface attachment coefficients at the four wheels, and the observed quantities include longitudinal acceleration, lateral acceleration, and yaw rate.

5. The unscented kalman filter-based vehicle speed estimation method according to claim 1, wherein the estimating, by the vehicle speed estimator, the longitudinal speed and the lateral speed at the next time by combining the state quantity at the current time and the state equation and the observation equation of the unscented kalman filter specifically comprises:

carrying out unscented transformation on the state quantity at the current moment to obtain a plurality of sigma points, and calculating the weight of each sigma point;

calculating a prior state estimation value, a prior error covariance and a prior observation estimation value according to the state equation and the observation equation;

and updating the next observed value and the error covariance according to the prior state estimated value, the prior error covariance and the prior observed estimated value.

6. The unscented kalman filter-based vehicle speed estimation method according to claim 1, wherein the state equation and the observation equation of the unscented kalman filter are:

wherein x is_k+1Is the state vector at time k +1, x_kIs the state vector at time k, z_kIs the observation vector at time k, u_kIs the input vector at time k, w_kIs the process noise at time k, v_kIs the observed noise at time k.

7. The unscented kalman filter-based vehicle speed estimation method according to claim 6, wherein the state equation further is:

where u (k) is the longitudinal velocity at time k,

is the derivative of the longitudinal velocity at time k, v (k) is the lateral velocity at time k, a_x(k) Longitudinal acceleration at time k, a_y(k) Lateral acceleration at time k, ω_r(k) Is the yaw rate at the time k,

derivative of yaw rate at time k, mu_fl(k) Road surface adhesion coefficient of the left front wheel at time k, mu_fr(k) Road surface adhesion coefficient of the right front wheel at time k, mu_rl(k) Road surface adhesion coefficient of the left rear wheel at time k, mu_rr(k) Road surface adhesion coefficient of the right rear wheel at time k, T_sThe step of time between k and k + 1.

8. The unscented kalman filter-based vehicle speed estimation method according to claim 1, characterized in that: the multi-degree-of-freedom vehicle dynamic model is a seven-degree-of-freedom dynamic model related to the vehicle speed.

9. A vehicle speed estimation system based on unscented kalman filtering, comprising:

the first modeling module is used for establishing a high-precision vehicle model and acquiring vehicle parameters through the high-precision vehicle model;

the second modeling module is used for establishing a multi-degree-of-freedom vehicle dynamics model, determining a state equation and an observation equation of an unscented Kalman filter according to the vehicle parameters, and determining input quantity, state quantity and observation quantity of the unscented Kalman filter, wherein the state quantity at least comprises longitudinal speed and lateral speed;

and the vehicle speed estimator is used for combining the state quantity at the current moment, and the state equation and the observation equation of the unscented Kalman filter to estimate the longitudinal speed and the lateral speed at the next moment.

10. The unscented kalman filter-based vehicle speed estimation system according to claim 9, further comprising:

the vehicle control unit is used for outputting the braking pressure of each wheel to the first modeling module;

the high-precision vehicle model simulates a vehicle in real time according to the brake pressure of each wheel to obtain the whole vehicle parameters;

the whole vehicle parameters comprise longitudinal acceleration, lateral acceleration, yaw rate, wheel speeds of four wheels and road surface adhesion coefficients at the four wheels.

Images