Adaptive UKF road adhesion coefficient estimation method based on event trigger mechanism

Abstract

The invention discloses a self-adaptive UKF (Unscented Kalman Filter) road adhesion coefficient estimation method based on an event trigger mechanism. Aiming at the problem that the traditional UKF algorithm depends on an accurate noise model and is sensitive to sensor data loss, the invention provides an improved scheme which comprises the following steps: firstly, establishing a seven-degree-of-freedom vehicle dynamics model, and deducing longitudinal, transverse and yaw motion equations; secondly, a Dugoff tire model is established, tire force is decomposed into normalized parts irrelevant to adhesion coefficients, normalized tire force is obtained, and vertical loads, slip angles and slip rates of four wheels are deduced and calculated; and finally, designing an adaptive UKF algorithm, dynamically adjusting a measurement noise covariance through a suboptimal unbiased MAP noise estimator, introducing an event triggering mechanism to judge a data loss state, and selectively executing measurement updating. A simulation experiment shows that under the condition of sensor noise abrupt change or data loss, compared with a traditional UKF algorithm, the method has the advantage that the reliability of stability control of the new energy automobile is remarkably improved.

Description

Technical Field

[0001] The present invention relates to the technical field of vehicles. Specifically, an Unscented Kalman Filter (UKF) method combining adaptive noise estimation and event-triggering mechanism is proposed to realize real-time estimation of road adhesion coefficient, which is applicable to the stability control of new energy vehicles for researchers to use. Background Art

[0002] The road adhesion coefficient is a core parameter for vehicle dynamics control, which directly affects tire force distribution, braking efficiency and steering stability. However, the road adhesion coefficient cannot be directly measured, or the measurement cost is too high. In this case, an estimation method needs to be used to indirectly obtain the road adhesion coefficient.

[0003] In recent years, the UKF algorithm has been widely studied in the estimation of road adhesion coefficient. This algorithm is a sub-optimal Kalman filter algorithm with non-linear varying mean and covariance, which does not require the calculation of the Jacobian matrix and can achieve the accuracy of the second-order Taylor expansion of the non-linear function, and its accuracy is better than that of the Extended Kalman Filter.

[0004] However, UKF requires a relatively accurate system mathematical model and noise estimation characteristics. Since the sensor noise varies dynamically with the environment and vehicle state and data loss is likely to occur, if the noise characteristics are not accurately estimated, it may lead to estimation errors in the actual application of UKF. Summary of the Invention

[0005] The purpose of the present invention is to adaptively adjust the measurement noise covariance through a sub-optimal unbiased MAP noise estimator, and judge the data loss situation by means of an event-triggering mechanism, so as to improve the robustness of the original road adhesion coefficient estimation based on UKF.

[0006] An event-triggering adaptive UKF road adhesion coefficient estimation method includes the following steps:

[0007] Step 1: Establish a seven-degree-of-freedom vehicle model and derive the motion equations of the vehicle in the longitudinal, lateral and yaw directions:

[0008] Step 2: Establish a Dugoff tire model, decompose the tire force into a normalized part independent of the adhesion coefficient to obtain the normalized tire force, as well as the vertical loads, sideslip angles and slip ratios of the four wheels;

[0009] Step 3: Adopt an adaptive unscented Kalman filter algorithm, dynamically adjust the measurement noise covariance through a sub-optimal unbiased MAP noise estimator, and introduce an event-triggering mechanism to judge the data loss state, and selectively perform measurement update to estimate the road adhesion coefficient.

[0010] In the said Step 1:

[0011] The seven - degree - of - freedom vehicle model is

[0012]

[0013]

[0014]

[0015] wherein, a x , a y respectively represent the longitudinal and lateral accelerations of the vehicle, F xi (i = 1, 2, 3, 4) represent the longitudinal tire forces of the four wheels, F yi (i = 1, 2, 3, 4) represent the lateral tire forces of the four wheels, represents the front - wheel steering angle, m represents the vehicle mass, ω r represents the angular velocity of the vehicle rotating about the vertical axis, I z represents the moment of force for the vehicle to rotate about the vertical axis, a and b respectively represent the distances from the vehicle's center of mass to the front axle and the rear axle, T f and T r represent the wheelbases of the front and rear wheels.

[0016] In the second step:

[0017] The Dugoff tire model is:

[0018]

[0019]

[0020] wherein, represents the normalized longitudinal tire force, represents the normalized lateral tire force, represents the road adhesion coefficient of the tire, respectively represent the longitudinal slip stiffness and the cornering stiffness of the tire, represents the tire slip ratio, is a non - linear function,

[0021]

[0022]

[0023] The vertical loads of the four wheels are respectively:

[0024]

[0025]

[0026]

[0027]

[0028] Among them, and represent the longitudinal and lateral accelerations, g represents the acceleration due to gravity, represents the height from the center of mass to the ground;

[0029] The calculation formula for the slip ratio is:

[0030]

[0031] Among them, represents the sign function, represents the longitudinal speed of the vehicle, R is the effective rolling radius of the wheel, is the rotational speed of the four wheels.

[0032] The calculation formula for the tire sideslip angle is:

[0033]

[0034]

[0035]

[0036]

[0037] Among them, are the tire sideslip angles of the four wheels, respectively represent the longitudinal and lateral speeds of the vehicle.

[0038] In the third step:

[0039] Set the state variables , respectively represent the road adhesion coefficients of the front left wheel, front right wheel, rear left wheel, and rear right wheel, the measurement variable , the control variable , where and are the sensor noises, and the corresponding state equations are respectively:

[0040]

[0041] The measurement equation is:

[0042]

[0043] Wherein,

[0044]

[0045]

[0046]

[0047]

[0048] The noise statistical characteristics are

[0049]

[0050]

[0051]

[0052]

[0053] Wherein, k and j represent the time step indices; E represents the mathematical expectation; and are respectively the mean and covariance matrix of the process noise; and are respectively the mean and covariance matrix of the observation noise; is the Kronecker function.

[0054] Furthermore, the event-triggered adaptive UKF algorithm in step three is described:

[0055] (1) Set the initial filter estimate and the initial covariance matrix , that is:

[0056]

[0057]

[0058] (2) Based on the state - estimated mean and covariance matrix of the n - dimensional random variable , generate a set of Sigma points that cover the probability distribution of the nonlinear system. The set of generated Sigma points is:

[0059]

[0060]

[0061] where is the scaling factor.

[0062] Weight calculation: Calculate the mean weight and covariance weight corresponding to the i - th Sigma point. The calculation formulas are as follows:

[0063]

[0065]

[0066] where is the scale parameter, taken as 0.001 here, is the higher - order moment containing the prior distribution, taken as 2 here.

[0067] (3) The time update process is as follows:

[0068]

[0069] The predicted mean is:

[0070]

[0071] The predicted covariance is:

[0072]

[0073] (4) The event - trigger condition judgment equation is:

[0074]

[0075] where Let [[[ID=]]] be the trigger threshold. If the trigger condition is 1, it is determined that data is lost and the measurement update is skipped. If it is 0, a complete measurement update is performed.

[0076] (5) The process of measurement update is as follows:

[0077] Substitute the predicted Sigma points into the measurement equation to obtain the measurement Sigma points:

[0078]

[0079] Calculate the predicted measurement mean and covariance:

[0080]

[0081]

[0082]

[0083] Among them, represents the innovation variance matrix, is the estimated measurement covariance matrix.

[0084] Calculate the Kalman gain:

[0085]

[0086] Among them, represents the inverse matrix of.

[0087] State and covariance update:

[0088]

[0089]

[0090] Among them, represents the inverse matrix of.

[0091] (6) The process of estimating the measurement noise statistical characteristics using an unbiased MAP noise estimator is as follows:

[0092]

[0093] Build a vehicle dynamics model in Carsim software and establish a road adhesion coefficient estimation method in Matlab / Simulink. Through co-simulation, first establish a seven-degree-of-freedom vehicle model; secondly, establish a Dugoff tire model to obtain the normalized tire force of the tire; finally, use the event-triggered adaptive UKF algorithm to estimate the road adhesion coefficient.

[0094] Advantages of the present invention:

[0095] Specifically, the present invention adaptively adjusts the measurement noise covariance through a sub-optimal unbiased MAP noise estimator. At the same time, to avoid data loss caused by sensors during data transmission, an event-triggering condition is introduced to determine whether to perform measurement updates through a threshold. After adding the noise estimator, under the condition of sudden increase in sensor noise, the estimation error of the adhesion coefficient is significantly reduced. At the same time, after introducing the event-triggering mechanism, the accuracy of estimation in the case of data loss is significantly improved. Description of the drawings

[0096] Figure 1 It is the flowchart of the event-triggered adaptive UKF algorithm proposed by the present invention;

[0097] Figure 2 It is the seven-degree-of-freedom vehicle model diagram adopted by the present invention;

[0098] Figure 3 It is the force and motion coordinate system diagram of the Dugoff tire model adopted by the present invention;

[0099] Figure 4 It is the comparison diagram of estimating the road adhesion coefficient with and without using the adaptive UKF under the sudden change of the adhesion coefficient;

[0100] Figure 5 It is the comparison diagram of estimating the road adhesion coefficient with and without using the adaptive UKF under the sudden change of noise;

[0101] Figure 6 It is the schematic diagram of the event-triggering mechanism;

[0102] Figure 7 It is the comparison diagram of estimating the road adhesion coefficient with and without using the event-triggering mechanism under the condition of data loss. Specific implementation manners

[0103] The proposed method will be further elaborated and explained below in conjunction with the drawings.

[0104] As Figure 1 shown, the present invention proposes an event-triggered adaptive UKF road adhesion coefficient estimation method, including the following steps:

[0105] Step 1: Establish as Figure 2The seven-degree-of-freedom vehicle model shown is used to obtain the longitudinal, lateral, and yaw motion equations of the vehicle based on vehicle dynamics relationships.

[0106] Step 2: Establish the Dugoff tire model as shown in Figure 3 to obtain the normalized tire force and derive the calculation formulas for the vertical loads, sideslip angles, and slip ratios of the four wheels.

[0107] Step 3: Use an adaptive unscented Kalman filter algorithm combined with an event-triggering mechanism to estimate the road adhesion coefficient.

[0108] The simulation experiment data of the technical solution provided by the present invention will be described in detail below.

[0109] The simulation environment of this experiment is a co-simulation platform built using CarSim and Simulink software. When the vehicle is driving on the road surface at a constant speed, there may be a sudden change in the adhesion coefficient. Under this condition, the estimation of the adhesion coefficient by the UKF algorithm and the adaptive UKF algorithm is as shown in Figure 4 At the same time, considering the sudden change in sensor noise, the estimation of the adhesion coefficient by the UKF algorithm and the adaptive UKF algorithm under this condition is as shown in Figure 5 The adaptive UKF significantly reduces the estimation error. To prevent the loss of sensor data, the present invention introduces an event-triggering mechanism to improve the estimation accuracy in the case of data loss. The schematic diagram of the event-triggering mechanism is as shown in Figure 6 The road adhesion coefficient estimation diagrams with and without the event-triggering mechanism are as shown in Figure 7 shown.

Claims

1. A method for estimating road adhesion coefficient based on event-triggered adaptive UKF, characterized in that: The change method includes the following steps: Step 1: Establish a seven-degree-of-freedom vehicle model and derive the vehicle's motion equations in the longitudinal, lateral and yaw directions: Step 2: Establish the Dugoff tire model and decompose the tire force into normalized parts that are independent of the adhesion coefficient to obtain the normalized tire force, as well as the vertical load, sideslip angle and slip rate of the four wheels; Step 3: Adopting the adaptive unscented Kalman filter algorithm, the measurement noise covariance is dynamically adjusted through the suboptimal unbiased MAP noise estimator, and an event trigger mechanism is introduced to determine the data loss state, and selectively perform measurement updates to estimate the road adhesion coefficient.

2. The method for estimating the road adhesion coefficient based on event-triggered adaptive UKF according to claim 1, characterized in that: The seven-degree-of-freedom vehicle model in step 1 is Among them, a x , a y Denote the longitudinal and lateral acceleration of the vehicle, respectively, F xi (i=1,2,3,4) represents the longitudinal force of the tires of the four wheels, F yi (i=1,2,3,4) represents the lateral force of the tires of the four wheels, represents the front wheel turning angle, m represents the vehicle mass, ω r represents the angular velocity of the vehicle around the vertical axis, I z represents the rotational torque of the vehicle around the vertical axis, a and b represent the distance from the center of mass of the vehicle to the front axle and to the rear axle respectively, T f and T r Indicates the front and rear wheelbase.

3. The method for estimating the road adhesion coefficient based on event-triggered adaptive UKF according to claim 2, characterized in that: The Dugoff tire model in step 2 is: in, represents the normalized tire longitudinal force, represents the normalized tire lateral force, Represents the tire's road adhesion coefficient, They represent the longitudinal slip stiffness and cornering stiffness of the tire respectively. is the tire slip rate, is a nonlinear function, 。 4. The method for estimating the road adhesion coefficient based on event-triggered adaptive UKF according to claim 3, characterized in that: The vertical loads of the four wheels in step 2 are: in, and represents longitudinal and lateral acceleration, g represents gravitational acceleration, Indicates the height from the centroid to the ground.

5. The method for estimating the road adhesion coefficient based on event-triggered adaptive UKF according to claim 4, characterized in that: The calculation formula for the slip rate in step 2 is: in, represents the symbolic function, represents the longitudinal speed of the vehicle, R is the effective rolling radius of the wheel, The four wheel speeds.

6. The method for estimating the road adhesion coefficient based on event-triggered adaptive UKF according to claim 5, characterized in that: The calculation formula for the tire slip angle in step 2 is: in, is the tire slip angle of the four wheels, represent the vehicle longitudinal and lateral velocities respectively.

7. The method for estimating the road adhesion coefficient based on event-triggered adaptive UKF according to claim 6, characterized in that: In step 3, we set the state variables , Respectively represent the road adhesion coefficients of the front left wheel, front right wheel, rear left wheel, and rear right wheel, and the measured variable , control amount ,in and is the sensor noise, and the corresponding state equations are: The measurement equation is: in, The noise statistics are Where k, j represent time step index; E represents mathematical expectation; and are the mean and covariance matrices of the process noise, respectively; and are the mean and covariance matrices of the observation noise, respectively; is the Kronecker function.

8. The method for estimating the road adhesion coefficient based on event-triggered adaptive UKF according to claim 7, characterized in that: The steps of the adaptive UKF algorithm triggered by the event in step 3 are as follows: (1) Set the initial filter estimate and the initial covariance matrix ,Right now: (2) Based on the state estimation mean and the covariance matrix n-dimensional random variable , generate a set of Sigma points, covering the probability distribution of nonlinear systems, and the set of generated Sigma points is: in, is the scaling factor; Weight calculation, calculate the mean weight corresponding to the i-th Sigma point With covariance weights , the calculation formula is as follows: in, is the scale parameter, here it is 0.001, is the higher-order moment containing the prior distribution, which is taken as 2 here; (3) The time update process is: The predicted mean is: The forecast covariance is: (4) The event trigger condition judgment equation is: in, is the trigger threshold. If the trigger condition is 1, the data is determined to be lost and the measurement update is skipped. If it is 0, a complete measurement update is performed. (5) The measurement update process is: Substitute the predicted Sigma point into the measurement equation to obtain the measured Sigma point: Compute the forecast measure mean and covariance: in, represents the innovation variance matrix, is the estimated measurement covariance matrix; Calculate the Kalman gain: in, express The inverse matrix of State and covariance updates: in, express The inverse matrix of (6) The process of estimating the statistical characteristics of measurement noise using the unbiased MAP noise estimator is: 。 9. The method for estimating the road adhesion coefficient based on event-triggered adaptive UKF according to claim 8, characterized in that: In step three, the vehicle dynamics model is built in Carsim software, and the road adhesion coefficient estimation method is built in Matlab / Simulink. Through joint simulation, firstly, a seven-degree-of-freedom model of the vehicle is established; secondly, a Dugoff tire model is established to obtain the normalized tire force of the tire; finally, the event-triggered adaptive UKF algorithm is used to estimate the road adhesion coefficient.

Images