JP7361242B2 - Model adaptive lateral velocity estimation method based on multi-sensor information fusion - Google Patents

Description

The present invention relates to the technical field of estimating lateral velocity, and specifically relates to a model adaptive lateral velocity estimation method based on multi-sensor information fusion.

Vehicle running safety has been a theme from the beginning to the end of the automobile industry, and in order to ensure vehicle running safety, stability control is the first issue. The vehicle stability control system uses state information such as the vehicle's lateral and longitudinal speed, acceleration, yaw rate, and center-of-gravity slip angle to control and distribute drive or brake torque on both sides of the vehicle to ensure stable vehicle running. . Among them, the vehicle's lateral and longitudinal acceleration and yaw rate can be measured using existing on-board sensors, but other dynamic state parameters need to be estimated using state estimation algorithms. Since variations in sensor noise cannot be ignored in sensor measurement information, it is necessary to correct sensor signals in conjunction with a dynamic model. Since the dynamic model has large model errors under nonlinear vehicle conditions, it is necessary to combine sensor information to ensure the convergence and accuracy of the estimation algorithm. Additionally, since the vehicle's lateral and longitudinal speed and acceleration are affected by the vehicle's own pitch, roll, and yaw angles, the accuracy of estimating the vehicle's center-of-gravity slip angle (lateral speed) always depends on the various factors mentioned above. limited.

Compared to the longitudinal speed, there is less information about the vehicle's lateral speed and the estimation difficulty is high, and the accuracy and robustness of the primitive Kalman filter estimation method cannot meet the requirements, so a new adaptive estimation method is used. It is necessary to estimate the lateral velocity.

The purpose of the present invention is to first design the adaptive process noise matrix and measurement noise matrix of the SR-UKF algorithm by referring to information such as the vehicle's lateral acceleration, yaw rate, and front wheel steering angle; Based on the kinematic model, an adaptive term is added to fuse the kinematic model, the weight ratio of both models is adjusted by the coefficient of the adaptive term, and finally, the adaptive noise matrix and the adaptive model are substituted into the SR-UKF algorithm. An object of the present invention is to provide a model-adaptive lateral velocity estimation method based on multi-sensor information fusion, which performs lateral velocity estimation based on multi-sensor information fusion. Here, by defining the basic probability function of the reliability of both sensors by the deviation between the lateral acceleration and yaw rate sensor values and the dynamic model calculation value, and fusing the information of both sensors according to the Dempster-Shafer evidence theory, The accuracy of the dynamic model and the uncertainty of the sensor are quantitatively calculated and evaluated from the observed values of both sensors, and the coefficient values of the adaptation term in the estimation method model are obtained, and finally the adaptation of the model is realized.

In order to solve the above technical problem, the present invention provides the following embodiments.

A model adaptive lateral velocity estimation method based on multi-sensor information fusion, comprising:
Step 1: Input sensor data including vehicle lateral acceleration, yaw rate, longitudinal speed, steering wheel deflection angle, and lateral acceleration sensor measurement results to create a vehicle dynamics model and a kinematics model. by adopting the Magic Formula tire model to calculate the tire longitudinal force and lateral force;
Step 3: Create an observation system model by merging the vehicle dynamics model and kinematics model, and realize observation system model adaptation using the observation system model weight adaptation term. The weight ratio between the dynamic model and the kinematics model is adjusted by adjusting the weight adaptation term coefficient, and the specific value of the observation system model weight adaptation term coefficient is determined by Dempster-Shafer with reference to the lateral acceleration and yaw rate sensor information. It is calculated according to the evidence theory, and here, as a method for determining the observation system model weight adaptation term, the measurement information of the vehicle's lateral acceleration sensor and the dynamics are calculated from the measurement information of the vehicle's lateral acceleration sensor and the yaw rate sensor. The deviation value from the lateral force in the model is calculated, and the deviation value between the measurement information of the yaw rate sensor and the yaw rate in the dynamic model is calculated, and the two deviation values are used as error determination statistics for the dynamic model. By defining the basic probability function of the vehicle's lateral acceleration sensor and yaw rate sensor reliability using decision statistics, and by fusing the information values of the lateral acceleration sensor and yaw rate sensor according to the Dempster-Shafer evidence theory, The accuracy of the dynamic model and the uncertainty of the lateral acceleration sensor and yaw rate sensor are quantitatively calculated and evaluated from the observed values of the acceleration sensor and yaw rate sensor, and the coefficient values of the weight adaptation term coefficients in the observation system model are obtained, and the final The objective is to realize adaptive adjustment of the observation system model in a
Step 4: Substitute the noise matrix and observation system model into the SR-UKF algorithm to construct an adaptive SR-UKF algorithm (ASR-UKF) to perform lateral velocity estimation; -UKF) processes include weight calculation and Sigma point generation, Sigma point propagation, state value and covariance value update/time update, estimated value and measured value update/measurement update, Kalman filter gain update including the process;
including methods.

...(5.11)
Creating the kinematics model is, that is, equation (4.27),
The tire longitudinal force and lateral force are calculated by adopting the magic formula, and the magic formula tire model uses a combination of trigonometric functions to fit the tire experimental data and uses one formula in the same format. The operating conditions of the tire's longitudinal force, lateral force and return torque can be completely described, and it can be applied in the same way even when the tire's lateral and longitudinal forces act together, and it has fitting accuracy, so the magic A tire model is created using the tire formula, and the calculation formula for longitudinal force and lateral force is Equation 2.8.
Here, B is a stiffness factor, which determines the slope at the origin of the function curve and can also be approximated to the stiffness coefficient of a linear tire model, and C is a curve shape factor, which determines the shape of the function curve, D is the peak factor and determines the maximum value of the function curve, E is the curve curvature factor and indicates the curve shape near the maximum value of the function curve, is the longitudinal slip rate of the wheel, and is the tire side slip angle When the vehicle is turning, there is a certain coupling relationship between the longitudinal force and the lateral force of the tire, and in that case, the Magic Tire Formula calculation results are corrected as follows.
Equation 2.13

In the above estimation method, the specific method of noise adaptation in step 2 is such that the error determination statistic of the dynamic model constituted by the lateral acceleration is shown by equation 5.12,

...(5.16)
Equation 5.17 above shows the effect of the coefficient more intuitively, and when the adaptive term coefficient is zero, the second term in Equation 5.16 is a dynamic model completely based on the magic tire formula, On the other hand, if the coefficient is 1, the second term in Equation 5.16 becomes a kinematic model completely based on the acceleration sensor, and therefore the deviation value between the kinematic model and the sensor, as shown in Equation 5.18. By referring to , and calculating the specific value of the adaptive term coefficient, weight adaptive adjustment of the dynamic model and kinematic model can be realized.
Equation 5.18
Similarly, the basic probability distribution function of the dynamic model accuracy observed by the vehicle body yaw rate sensor is Equation 5.29,
The combined mass function of sensor information accuracy k is Equation 5.32,
Here μ is the orientation parameter and is defined by simulation and actual experiment.

Compared with the prior art, the beneficial effects achieved by the present invention are that the present invention improves vehicle performance in three ways: algorithm model adaptive adjustment, noise matrix adaptive adjustment and application of square root no-track Kalman filter (SR-UKF) algorithm. Achieving adaptation of lateral velocity estimation. First, referring to the vehicle's lateral acceleration and yaw rate sensor information, the process noise matrix and measurement noise matrix of the SR-UKF algorithm are designed, and the estimation method realizes noise adaptation in various driving conditions. After that, by adding an adaptive term to the estimation method model and calculating the numerical value of the adaptive term parameter according to the DS evidence theory, the weight ratio between the dynamic model and the kinematic model is quantitatively distributed, and the estimation Achieving method model adaptation. Finally, the adaptive noise matrix and the adaptive model are substituted into the SR-UKF algorithm to perform lateral velocity estimation.

The accompanying drawings are used to provide a further understanding of the invention, form part of the specification, and together with the examples thereof are used to explain the invention and are not intended to limit it. .
FIG. 2 is a schematic configuration diagram of a lateral velocity adaptive estimation method in an embodiment of the present invention. FIG. 3 is a schematic diagram in which the estimation method of the present invention verifies the lateral speed estimation result under a double flow line driving condition on a highly adhesive road surface. FIG. 2 is a schematic diagram in which the estimation method of the present invention verifies the lateral speed estimation result under a dual flow line driving condition on a low-adhesion road surface.

Hereinafter, embodiments of the present invention will be clearly and completely described with reference to the accompanying drawings of the present invention, but the described embodiments are only some embodiments of the present invention, Obviously, not all implementations are possible. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without any creative efforts fall within the scope of protection of the present invention.

Example 1

Considering the errors of the vehicle model and tire model in the case of low adhesion road surface as well as the cumulative error of the sensor in the case of colored noise, the present invention develops a model adaptive lateral velocity estimation method based on multi-sensor information fusion. provide. First, by referring to information such as the vehicle's lateral acceleration, yaw rate, and front wheel steering angle, and designing the process noise matrix and measurement noise matrix of the SR-UKF algorithm, the algorithm noise can be adapted to various vehicle driving conditions. Realize. Then, based on the original estimation method dynamic model, an adaptive term is added to fuse the kinematic models, and the weight ratio of both models is adjusted by the coefficient of the adaptive term. Since the longitudinal force information of an in-wheel motor-driven vehicle is relatively accurate, the yaw rate information of the dynamic model basically reflects the degree of accuracy of the lateral force in the model, and also uses it as an adaptation term. Estimated calculations can be made by inputting the data. Based on the measurement information of the lateral acceleration sensor and yaw rate sensor in the vehicle ESP, the deviation values between the lateral force and the yaw rate in the dynamic model are calculated and used as error determination statistics for the dynamic model. The basic probability function of the reliability of both sensors is defined using the statistics, and the information values of both sensors are fused according to the Dempster-Shafer evidence theory, and the accuracy of the dynamic model and the uncertainty of the sensor are calculated from the observed values of both sensors. We quantitatively calculate and evaluate the coefficients of the adaptive term in the estimation method model, and finally realize model adaptation and improve the robustness of the lateral velocity estimation method.

The present invention refers to information such as the lateral acceleration, yaw rate, and front wheel steering angle of the vehicle, and estimates the lateral velocity of the vehicle using the square root no-track Kalman filter SR-UKF method. Thereby, the adaptation of the algorithm noise to various driving conditions of the vehicle is realized. In other words, the core of the present invention is to perform estimation using the SR-UKF algorithm, and before estimation, calculate the noise matrix and model using sensor data for orientation. The model consists of two parts: a dynamic model and a kinematics model, and the weights (occupied ratios) of the two models are calculated according to the DS evidence theory. After getting the weights, we get the adaptive model. A detailed outline thereof is shown in FIG. In order to improve the robustness and estimation accuracy in the lateral velocity estimation process, an adaptive law design is performed for the estimation method. On the one hand, a noise matrix is defined by the lateral acceleration and yaw rate to realize noise adaptation in various driving conditions, and on the other hand, an adaptation term is added to the estimation method model, and the dynamics model and kinematics are adjusted by adjusting the adaptation term coefficient. The weight ratio with the model is adjusted, and the specific numerical value of the adaptive term coefficient is calculated according to the Dempster-Shafer evidence theory with reference to the lateral acceleration and yaw rate sensor information.

The present invention provides a model adaptive lateral velocity estimation method based on multi-sensor information fusion, including the following steps.

Step 1: Create a vehicle dynamics model and kinematics model by inputting sensor data including vehicle lateral acceleration, yaw rate, longitudinal speed, steering wheel deflection angle, etc.

Step 3: Create an observation system model by merging the dynamic model and kinematics model, and realize model adaptation using the model weight adaptation term. As a model adaptation method, an adaptive term is added to the estimation method model, and the weight ratio between the dynamic model and the kinematic model is adjusted by adjusting the adaptive term coefficient.The specific value of the adaptive term coefficient is It is calculated based on the Dempster-Shafer evidence theory with reference to the yaw rate sensor information. Here, as a method for determining the model adaptation term (i.e., the weight ratio between the dynamic model and the kinematic model), the lateral force and yaw rate in the dynamic model are determined using the vehicle's lateral acceleration sensor and yaw rate sensor measurement information, respectively. The deviation value is calculated and used as the error determination statistic for the dynamic model. The basic probability function of the reliability of both sensors is defined using the statistics, and the information values of both sensors are fused according to the Dempster-Shafer evidence theory, and the accuracy of the dynamic model and the uncertainty of the sensor are calculated from the observed values of both sensors. We quantitatively calculate and evaluate the performance, obtain the coefficient values of the adaptation term in the estimation method model, and finally realize the adaptation of the model.

Step 4: Substitute the adaptive noise matrix and adaptive model into the SR-UKF algorithm to perform lateral velocity estimation. The SR-UKF algorithm steps include weight calculation and Sigma point generation, Sigma point propagation, state value and covariance value update/time update, estimated value and measured value update/measurement update, Kalman filter gain update. Includes processes such as updates.

In step 1 above, creating the kinematic model is, in other words, equation (4.27).

In step 1 above, the tire longitudinal force and lateral force are calculated by adopting the magic formula.

The Magic Formula tire model uses a combination of trigonometric functions to fit tire experimental data and uses one equation in the same format to fully describe the driving conditions of the tire's longitudinal force, lateral force and return torque. It can be applied in the same way even when the tire's lateral and longitudinal forces act together, and it has fitting accuracy. Therefore, a tire model is created using the Magic Tire Formula. The calculation formula for the longitudinal force and the lateral force is Equation 2.8.

While the vehicle is turning, a certain coupling relationship exists between the longitudinal force and the lateral force of the tire. In that case, the Magic Tire Formula calculation result is corrected as follows and becomes Equation 2.13.

In step 2 above, the estimation method noise adaptation is specifically as follows.

here,

The significance of the dynamic model adaptation term is that the weight values of the kinematic model and the dynamic model can be adjusted adaptively. Expanding equation 5.16, we get

The above equation 5.17 shows the effect of the coefficient more intuitively. If the adaptive term coefficient is zero, the second term in equation 5.16 is a dynamic model completely based on the magic tire formula. On the other hand, when the coefficient is 1, the second term in equation 5.16 becomes a kinematic model completely based on the acceleration sensor, as shown in equation 5.18. Therefore, by referring to the deviation value between the dynamic model and the sensor and calculating a specific numerical value of the adaptive term coefficient, adaptive weight adjustment of the dynamic model and the kinematic model can be realized.

UKF estimation methods based on model adaptation compensate for errors due to model inaccuracy by adjusting noise with adaptation factors. By adding an adaptive term based on lateral acceleration sensor information to the dynamic model, the influence of model parameters and accelerometer noise accumulation is reduced. At the same time, instead of the original covariance, a square root algorithm is used repeatedly to solve the problem that the covariance loses positive definiteness due to step error and cutoff error, making filtering impossible.

By introducing the Dempster-Shafer evidence theory to process the uncertainty information of the lateral acceleration sensor, body yaw rate sensor, and vehicle dynamics model, the probability value of each information source is calculated, and finally the dynamics Obtain specific values for the model adaptation term coefficients.

Calculation of model adaptation term coefficients for the above estimation method

In estimating the vehicle center-of-gravity side slip angle, the present invention uses the lateral acceleration sensor and the vehicle yaw rate sensor as observables, and iterates the estimated center-of-gravity side slip angle and lateral and longitudinal velocity values into the magic formula tire model. Calculate the lateral force on each tire by: The seven-degree-of-freedom dynamic model of the vehicle creates a dynamic relational expression during turning motion of the vehicle based on the lateral force calculated by the magic tire formula. In order to evaluate the accuracy of the dynamics model, the deviation value between the calculation result of the lateral force and the measurement result of the lateral acceleration sensor is used as the error determination statistic, and the lateral acceleration sensor information dynamics model is referred to. We suggest making some corrections. However, in actual cases, the lateral acceleration sensor may be interfered with by certain external factors, so the deviation of its measurement value may be relatively large, and in this case, the power of lateral movement The degree of accuracy of the academic model cannot be accurately corrected and evaluated. As can be seen from Equation 5.17, the yaw rate dynamics model accuracy depends significantly on the longitudinal force and lateral force accuracy of the four wheels, and the longitudinal force of the in-wheel motor driven vehicle is known precisely. Therefore, the yaw rate calculation value can similarly reflect the accuracy of the dynamic model lateral force. In order to improve the robustness of the algorithm, the present invention introduces yaw rate sensor measurement information, creates a new yaw rate dynamics model error determination statistic, cross-validates it with the lateral acceleration deviation statistic, and calculates D - Calculate the final dynamical model adaptation term coefficient value by estimating the degree of accuracy of the dynamical model in combination with the S evidence theory.

The yaw rate of the vehicle body is the same as the integral of the ratio of the yaw moment to the z-axis rotational moment of inertia of the vehicle body, as shown in equation 5.26 below.

Similarly, the basic probability distribution function of the dynamic model accuracy observed by the vehicle body yaw rate sensor is Equation 5.29.

First, the normalization constant K is calculated as shown in Equation 5.30.

After that, the combined mass function of the dynamic model accuracy d is calculated as shown in Equation 5.31.

The combined mass function of the accuracy k of sensor information is Equation 5.32.

In this case, for the combined mass functions of d and k, the mass function values obtained by combining the confidence function and the likelihood function value are all equivalent. That is, Equation 5.33.

The dynamic model adaptation term coefficient can be calculated according to the following equation, as shown in equation 5.34.

Here μ is the orientation parameter and is defined by simulation and actual experiment.

In step 4 above, the SR-UKF algorithm is used to estimate the lateral velocity by substituting the adaptive noise matrix and the adaptive model. The SR-UKF algorithm steps include weight calculation and Sigma point generation, Sigma point propagation, state value and covariance value update/time update, estimated value and measured value update/measurement update, Kalman filter gain update. Includes processes such as updates.

The contents included in the SR-UKF algorithm process are weight calculation and Sigma point generation, Sigma point propagation, state value and covariance value update/time update, estimated value and measured value update/measurement update. , the Kalman filter gain updating process are all conventional technology processes.

The present invention realizes the adaptation of wheel lateral speed estimation in three aspects: system noise matrix, algorithm model, and SR-UKF algorithm. First, the RLS algorithm is used to mark the vehicle's mass and center of gravity location, respectively, to reduce dynamic model errors due to loading changes. After that, the process noise matrix and measurement noise matrix of the SR-UKF algorithm are designed with reference to the vehicle's lateral acceleration and yaw rate sensor information, and the noise adaptation of the estimation method to various driving conditions is realized. Finally, by adding an adaptive term to the estimation method model and calculating the numerical value of the adaptive term parameter according to the DS evidence theory, the weight ratio between the dynamic model and the kinematic model is quantitatively distributed, Achieve adaptation of the estimation method model.

In order to verify the estimation effect of the estimation method of the present invention, verification was performed using the following verification method.

1. The estimation method of the present invention was verified under dual flow line driving conditions on a highly adhesive road surface, and verification effects were obtained.

The road surface adhesion coefficient was set to 0.85, the standard double flow line driving condition, and the target speed to 60 km/h. The estimation error of the lateral speed under the dual flow line driving condition on the high adhesion road surface is shown in Table 1 below.

As can be seen from Table 1 and FIG. 2, in the method of the present invention, when the vehicle turns in a stable driving condition on a high adhesion road surface, both the SR-UKF algorithm and the SR-UKF algorithm based on model adaptation was able to follow the changes well. However, when the vehicle is moving steadily straight, ie, the steering angle is relatively small, the accuracy of the ASR-UKF algorithm is better than the SR-UKF, ie, the adaptation method of the algorithm is effective.

2. The estimation method of the present invention was verified under a dual flow line driving condition on a low-adhesion road surface, and verification effects were obtained.

In CarSim, the initial speed of the vehicle was set to 48 km/h, the road surface condition was set to be an icy road surface, the road surface adhesion coefficient was 0.4, and the test was conducted in a dual flow line driving state. These are the front wheel rotation angle, the wheel speed of each wheel, a lateral acceleration signal with Gaussian white noise added, and a yaw rate signal.

The error analysis is shown in Table 2 below.

As can be seen from Table 2 and Figure 3, when the vehicle runs normally, the difference between ASR-UKF and SR-UKF is not large, and both track relatively accurately and estimate the value of lateral speed. I was able to do that. However, when the vehicle enters the extreme nonlinear region where the steering angle is the largest, the original SR-UKF algorithm shows a constant deviation, and the error increasing over time is accumulated. On the other hand, the SR-UKF algorithm based on model adaptation proposed by the present invention was still able to estimate the value of lateral velocity with good tracking. That is, the adaptive rules of the algorithm in the method of the present invention were able to successfully handle the problem of estimating the dynamic state of the vehicle in the extreme driving state.

It should be noted that the above description is only a preferred embodiment of the present invention, and does not limit the present invention. Although the present invention has been explained in detail with reference to the above-mentioned embodiment, those skilled in the art will understand that the above-mentioned The technical means described in each embodiment can be modified or some of its technical features can be equivalently replaced. Any modification, equivalent replacement, improvement, etc. within the spirit and principle of the present invention shall fall within the protection scope of the present invention.

Claims (4)

A model adaptive lateral velocity estimation method based on multi-sensor information fusion, comprising:
Step 1: Input sensor data including vehicle lateral acceleration, yaw rate, longitudinal speed, steering wheel deflection angle, and lateral acceleration sensor measurement results to create a vehicle dynamics model and a kinematics model. by adopting the Magic Formula tire model to calculate the tire longitudinal force and lateral force;
Step 3: Create an observation system model by merging the vehicle dynamics model and kinematics model, and realize observation system model adaptation using the observation system model weight adaptation term. The weight ratio between the dynamic model and the kinematics model is adjusted by setting the weight adaptation term coefficient to 1 , and the specific value of the observation system model weight adaptation term coefficient is determined by Dempster-Shafer with reference to the lateral acceleration and yaw rate sensor information. It is calculated according to the evidence theory, and here, as a method for determining the observation system model weight adaptation term, the measurement information of the vehicle's lateral acceleration sensor and the dynamics are calculated from the measurement information of the vehicle's lateral acceleration sensor and the yaw rate sensor. The deviation value from the lateral force in the model is calculated, and the deviation value between the measurement information of the yaw rate sensor and the yaw rate in the dynamic model is calculated, and the two deviation values are used as error determination statistics for the dynamic model. By defining the basic probability function of the vehicle's lateral acceleration sensor and yaw rate sensor reliability using decision statistics, and by fusing the information values of the lateral acceleration sensor and yaw rate sensor according to the Dempster-Shafer evidence theory, The accuracy of the dynamic model and the uncertainty of the lateral acceleration sensor and yaw rate sensor are quantitatively calculated and evaluated from the observed values of the acceleration sensor and yaw rate sensor, and the coefficient values of the weight adaptation term coefficients in the observation system model are obtained, and the final The objective is to realize adaptive adjustment of the observation system model in a
Step 4: Substitute the noise matrix and observation system model into the SR-UKF algorithm to construct an adaptive SR-UKF algorithm (ASR-UKF) to perform lateral velocity estimation; -UKF) processes include weight calculation and Sigma point generation, Sigma point propagation, state value and covariance value update/time update, estimated value and measured value update/measurement update, Kalman filter gain update including the process;
An estimation method comprising: The estimation method according to claim 1,
In step 1, creating the vehicle dynamics model includes:
The tire longitudinal force and lateral force are calculated by adopting a magic formula,
The Magic Formula tire model uses a combination of trigonometric functions to fit tire experimental data and uses one equation in the same format to fully describe the driving conditions of the tire's longitudinal force, lateral force and return torque. It can be applied in the same way even when the tire's lateral and longitudinal forces act together, and it has fitting accuracy. Therefore, the magic tire formula is adopted to create a tire model and the longitudinal and lateral forces are combined. The calculation formula is formula 2.8,
Here, B is a stiffness factor, which determines the slope at the origin of the function curve and can also be approximated to the stiffness coefficient of a linear tire model, and C is a curve shape factor, which determines the shape of the function curve, D is the peak factor and determines the maximum value of the function curve, E is the curve curvature factor and indicates the curve shape near the maximum value of the function curve, is the longitudinal slip rate of the wheel, and is the tire side slip angle and
While the vehicle is turning, there is a certain coupling relationship between the longitudinal force and the lateral force of the tire. In this case, the Magic Tire Formula calculation result is corrected as follows, as shown in Equation 2.13. ,
An estimation method characterized by: The estimation method according to claim 1,
The specific method of noise adaptation in step 2 is as follows:
The error determination statistic of the dynamic model composed of lateral acceleration is shown by Equation 5.12,
The estimation method according to claim 1,
The specific method of step 3 is as follows:
Equation 5.17 above shows the effect of the coefficient more intuitively, and the adaptive term coefficient is 1 , the second term in Equation 5.16 is a kinematic model completely based on the acceleration sensor, and Equation 5.18, the weight adaptive adjustment of the dynamic model and kinematics model can be realized by referring to the deviation value of the dynamic model and the sensor and calculating the specific value of the adaptive term coefficient. and as shown in equation 5.18,
The yaw rate of the vehicle body is equal to the integral of the ratio of the yaw moment to the z-axis rotational moment of inertia of the vehicle body, as shown in equation 5.26 below,
Similarly, the basic probability distribution function of the dynamic model accuracy observed by the vehicle body yaw rate sensor is Equation 5.29,
Then, calculate the combined mass function of the dynamic model accuracy d,
Equation 5.31
The combined mass function of sensor information accuracy k is Equation 5.32,
In this case, for the combined mass functions of d and k, the mass function values after combining the confidence function and the likelihood function value are all equivalent , that is, Equation 5.33,
The dynamic model adaptation term coefficient is calculated according to the following formula,
Equation 5.34
Here, μ is the orientation parameter, which is determined by simulation and actual experiment.
An estimation method characterized by: