Vehicle state parameter joint estimation method based on fuzzy event triggering mechanism

Abstract

The invention relates to a vehicle state parameter joint estimation method based on a fuzzy event triggering mechanism. Comprising the following steps: step 1, establishing a four-wheel steering vehicle kinematics and dynamics model; 2, combining a joint estimation method of a four-wheel steering vehicle kinematics and dynamics model, and constructing a vehicle state and parameter joint estimator based on a three-degree-of-freedom equation; and step 3, designing an event trigger mechanism based on a fuzzy rule, and dynamically adjusting a trigger threshold of state estimation. According to the method, the vehicle state and the tire cornering stiffness are estimated in parallel in the estimation process, the influence of different to-be-estimated quantities on the estimation of the other party is reduced, the estimation precision of the vehicle side slip angle is effectively improved, the updating frequency can be dynamically adjusted and corrected according to the change of the observation residual error and the measurement error covariance, and the estimation precision of the vehicle side slip angle is improved. Unnecessary updating is reduced to avoid accumulative errors caused by integration, meanwhile, the use of computing resources is reduced, and a new solution is provided for estimation of the side slip angle of the vehicle.

Description

Technical Field

[0001] The present invention belongs to the technical field of autonomous driving vehicles, and specifically is a vehicle state parameter joint estimation method based on a fuzzy event triggering mechanism. Background Art

[0002] The rapid development of intelligent vehicles has effectively alleviated key challenges such as traffic congestion, driving safety, and energy efficiency. The development of advanced chassis control technologies, such as electronic stability systems and drive anti-skid systems, can significantly enhance the safety performance of intelligent vehicles. Four-wheel steering (FWS) is currently recognized as one of the most important chassis control systems affecting vehicle handling and stability. This technology significantly optimizes vehicle steering characteristics by actively adjusting the rear wheel steering angle. However, the implementation of various intelligent vehicle chassis control systems relies on accurate vehicle state parameters. Among these, the vehicle's slip angle (CSA), the angle between the vehicle's longitudinal axis and the velocity vector, is considered a key variable that significantly influences the dynamic performance of intelligent vehicles. Direct measurement of the CSA requires expensive, specialized test sensors such as optical or GPS sensors, making it unsuitable for mass-produced vehicles. Therefore, it is crucial to achieve low-cost, highly reliable CSA estimation for four-wheel steering vehicles using sensors available in mass-produced vehicles.

[0003] Currently, vehicle center of mass slip angle estimation has been extensively studied, primarily encompassing methods based on kinematic models and dynamic models. Kinematic model-based methods are independent of specific vehicle parameters, tire model, or road friction coefficient. They utilize continuous integration of vehicle kinematic signals through methods such as vehicle kinematic geometry estimation, self-correcting data fusion, lateral acceleration deviation compensation, and fuzzy logic fusion estimation to achieve simple and reliable estimation. However, kinematic estimation methods are highly sensitive to interference from low-cost sensor signals and measurement noise, and the accumulated errors from continuous integration can lead to a certain degree of drift in estimation accuracy. Dynamic model-based estimation methods combine simplified or variant vehicle and tire models with advanced state observers to achieve more robust and accurate estimation of vehicle center of mass slip angles under various operating conditions. However, most dynamic model-based methods assume that the vehicle longitudinal velocity can be accurately measured and ignore the impact of uncertainty in tire cornering stiffness on the center of mass slip angle estimation under conditions with high vehicle lateral angular velocity. This, in turn, reduces the reliability and accuracy of the vehicle center of mass slip angle estimation results.

[0004] Therefore, a joint estimation method for the vehicle's slip angle and vehicle model parameters can be used to address the uncertainty of vehicle dynamics model parameters. For example, the vehicle slip angle is estimated in conjunction with vehicle mass, center of mass position, tire forces, and wheel cornering stiffness. These joint estimation methods can reliably help vehicles achieve good estimation performance in a variety of operating conditions. However, the complex observer structure also significantly increases the computational burden of engineering applications.

[0005] In recent years, event-triggered mechanisms have reduced the frequency of redundant data updates, lowering the communication and computational burden of systems and improving resource efficiency. They have been widely applied in networked control systems, wireless sensor networks, and multi-agent vehicle dynamics control. Introducing event-triggered mechanisms into vehicle state estimation can eliminate accumulated errors. While event-triggered mechanisms can effectively reduce the computational burden of vehicle sideslip angle estimation, selecting the triggering events and setting the thresholds remain challenging. Summary of the Invention

[0006] To solve the above technical problems, the present invention provides a joint estimation method for vehicle state parameters based on a fuzzy event triggering mechanism. During the estimation process, the vehicle state and tire lateral stiffness are estimated in parallel, reducing the influence of different estimated quantities on the estimation of the other party, effectively improving the estimation accuracy of the vehicle center of mass sideslip angle, and dynamically adjusting the correction update frequency according to the changes in the observation residual and the measurement error covariance, reducing unnecessary updates to avoid the cumulative error caused by integration while reducing the use of computing resources, providing a new solution for vehicle center of mass sideslip angle estimation.

[0007] The technical solution of the present invention is described as follows in conjunction with the accompanying drawings:

[0008] A vehicle state parameter joint estimation method based on a fuzzy event triggering mechanism includes the following steps:

[0009] Step 1: Establish the kinematic and dynamic model of the four-wheel steering vehicle;

[0010] Step 2: Combine the joint estimation method of the four-wheel steering vehicle kinematics and dynamics model to construct a joint estimator of vehicle state and parameters based on the three-degree-of-freedom equation;

[0011] Step 3: Design an event trigger mechanism based on fuzzy rules to dynamically adjust the trigger threshold of state estimation.

[0012] Furthermore, the specific method of step one is as follows:

[0013] 11) Assuming the wheel angles on the front and rear axles are equal, a three-degree-of-freedom model of the vehicle is established by combining the vehicle kinematics and dynamics equations:

[0014]

[0015]

[0016]

[0017] Where V x is the longitudinal velocity; V y is the lateral velocity; ω r is the yaw angular velocity; a x is the longitudinal acceleration; a y is the lateral acceleration; I z is the moment of inertia of the vehicle around the z-axis; a and b are the distances from the front and rear axles to the center of mass of the vehicle, respectively; δ f , δ r are the wheel angles of the front and rear axles respectively; F yf 、F yr are the cornering forces of the front and rear axle wheel tires respectively;

[0018] 12) The relationship between tire lateral force and tire slip angle is as follows:

[0019]

[0020] Where, F yf 、F yr are the cornering forces of the front and rear axle tires respectively; C f 、C r is the lateral stiffness of the front and rear axle wheels; α f , α r is the side slip angle of the tires on the front and rear axles of the vehicle;

[0021] 13) The side slip angle of the front and rear axle tires of the vehicle is calculated by the following formula:

[0022]

[0023] Where, α f , α r V is the side slip angle of the front and rear axle wheels of the vehicle; x is the longitudinal velocity; V y is the lateral velocity; ω r is the yaw angular velocity; a and b are the distances from the front and rear axles to the center of mass of the vehicle respectively; δ f , δ r are the wheel angles of the front and rear axles respectively;

[0024] 14) Combining equations (3), (4), and (5), we can get the kinematic-dynamic model equation of the four-wheel steering vehicle:

[0025]

[0026] Where V x is the longitudinal velocity; V y is the lateral velocity; ω r is the yaw angular velocity; a x is the longitudinal acceleration; a y is the lateral acceleration; I z is the moment of inertia of the vehicle around the z-axis; a and b are the distances from the front and rear axles to the center of mass of the vehicle, respectively; δ f , δ r are the front and rear axle wheel angles respectively; C f 、C r is the wheel cornering stiffness of the front and rear axles;

[0027] 15) The continuous-time vehicle model is discretized using the forward Euler method to obtain the following discrete state space equations:

[0028]

[0029] Where x k is the discrete system state variable, x k =[V x,k ,V y,k ,ω r,k ] T ;z k is the discrete system measurement variable, z k =ω r,k ; f(·) is the state transfer equation of the discrete system; h(·) is the measurement equation of the discrete system.

[0030] Furthermore, the specific method of step 2 is as follows:

[0031] 21) Adaptive Kalman filtering method is used to estimate tire cornering stiffness;

[0032] 22) Adaptive cubature Kalman filtering method is used to estimate vehicle state.

[0033] Furthermore, the specific method of step 21) is as follows:

[0034] The state variable of the tire cornering stiffness estimator is selected as x p =[C f ,C r ] T , the lateral acceleration and yaw angular acceleration measured directly from the sensor are selected as the measurement variables The required input for the tire cornering stiffness estimator is u p =[V x ,V y ,ω r ,δf ,δ r ];

[0035] The state prediction equation for tire cornering stiffness estimation is:

[0036]

[0037] The tire cornering stiffness estimation measurement equation is:

[0038]

[0039] Where C f,k 、C r,k C is the wheel cornering stiffness of the front and rear axles at time k; f,k+1 、C r,k+1 is the wheel cornering stiffness of the front and rear axles at time k+1; is the yaw angular acceleration at time k+1; a y,k+1 is the lateral acceleration at time k+1; δ f,k , δ r,k are the front and rear axle wheel angles at time k+1; α f,k , α r,k I is the side slip angle of the front and rear axle tires of the vehicle at time k+1; z is the moment of inertia of the vehicle around the z-axis; a and b are the distances from the front and rear axles to the center of mass of the vehicle, respectively; and m is the mass of the vehicle.

[0040] Furthermore, the specific method of step 22) is as follows:

[0041] The state variable of the vehicle state estimator is selected as x s =[V x ,V y ,ω r ] T The measured variable is selected as the yaw angular velocity z directly obtained from the vehicle inertial sensor s =ω r , the required input of the vehicle state estimator is u s =[a x ,a y ,δ f ,δ r ,C f ,C r ] T ;

[0042] The vehicle longitudinal speed V x,0 , lateral velocity V y,0 and yaw rate ω r,0 The state vector x s,0 As the initial value, initialize the error covariance matrix P s,0, the initial covariance Q of the process noise ω s,0 and the initial covariance R of the measurement noise v s,0 ;

[0043] initialization:

[0044]

[0045] At the initial moment, there is x s,k =x s,0 ,P s,k =P s,0 ;

[0046] The covariance matrix P is decomposed into s,k Convert to lower triangular matrix S k :

[0047]

[0048] Let the state dimension be n and construct 2n volume points:

[0049]

[0050] Using unit orthogonal vector ξ i Construct a volume point set:

[0051]

[0052] Where i is the volume point sequence number; n is the dimension of the state variable; [1] i is the i-th column of the n-dimensional identity matrix; the volume points corresponding to the vehicle state estimation method are determined by the symmetric sampling rule of the three-dimensional state space:

[0053]

[0054] In the prediction stage, each volume point propagates the vehicle state prediction equation (15) through the nonlinear state equation, which takes into account the coupling effect of lateral and longitudinal acceleration and yaw motion:

[0055]

[0056] Where, T s is the sampling time; [V x,k+1 ,V y,k+1 ,ω r,k+1 ] T is the value of the state variable at the next sampling moment.

[0057]

[0058] Where, is the predicted value of the propagation volume point at time k+1 at time k.

[0059] The propagated volume points are weighted averaged to obtain the state prediction value x s,k+1|k , and update the prediction covariance matrix P s,k+1|k , the process explicitly includes process noise Q s,k Impact:

[0060]

[0061] After entering the measurement update phase, the algorithm regenerates volume points:

[0062]

[0063] Since the yaw angular velocity ω r Direct measurement, the vehicle state measurement equation is simplified to the identity mapping:

[0064] ω r,k+1 =ω r,k+1 (twenty two)

[0065] Volumetric Point Propagation:

[0066]

[0067] The measured volume points after propagation calculate the predicted value of the measured variable z s,k+1 , and obtain the measurement variable covariance matrix and the measured variable-predictor variable cross-covariance matrix for:

[0068]

[0069] Then, the state estimate x is updated by fusing the prediction and measurement information through the Kalman gain formula s,k+1 and the covariance matrix P s,k+1 :

[0070]

[0071]

[0072] Where, is the actual measurement value at the k+1 sampling time;

[0073] The exponential weighting method is used to dynamically adjust the measurement noise covariance matrix R s,k+1 , the adjustment range depends on the current observation residual e k+1 :

[0074]

[0075]

[0076] Where H is the observation matrix of the measured variables;

[0077] Finally, using the updated lateral / longitudinal velocity estimates, the vehicle's sideslip angle is calculated from the geometric relationship:

[0078]

[0079] Furthermore, the specific method of step three is as follows:

[0080] 31) The vehicle state estimator uses the observation residual as the trigger event, and the preset trigger threshold is adaptively adjusted based on the changes in the observation residual and the measurement error covariance;

[0081] 32) The parameter estimator uses the change in front wheel angle as a trigger event, and the preset trigger threshold is related to the vehicle speed.

[0082] Furthermore, the specific method of step 31) is as follows:

[0083] The difference in the measurement error covariance matrix can represent the impact of changes in measurement noise on the system. The difference in the measurement error covariance matrix is:

[0084]

[0085] The TS fuzzy system is used to adaptively adjust the trigger threshold; the rules of the TS fuzzy system are:

[0086] R j :

[0087] If x1 is x2 is ……,x n yes Then y j =f j (X),j=1,2,…,M,X=(x1,x2,…,x n ) T , where x i is the input variable, i=1,2,……,n, It is a fuzzy subset, and the input variables are x1=e,x2=ΔP z In order to realize the fuzzy inference system, the input variables are divided into three fuzzy subsets S = Small, M = Middle, B = Big, N is the total number of fuzzy rules, f j (·) is the rule R j The consequent function, y j is the output, and the activation strength of the j-th rule is given by:

[0088]

[0089] Where, is x i exist The degree of membership in ;

[0090] The output Y is calculated by aggregating the contributions of each rule:

[0091]

[0092] The fuzzy system has two input variables and three fuzzy sets, and the following nine rules are defined:

[0093]

[0094] Finally, the output ξ s for:

[0095]

[0096] Furthermore, the specific method of step 32) is as follows:

[0097] The vehicle front wheel angle difference is selected as the trigger event for tire cornering stiffness estimation:

[0098] Δδ f,k =δ f,k -δ f,k-1 (37)

[0099] Event trigger threshold ξ of tire cornering stiffness estimator p It should be able to adjust adaptively with the change of vehicle longitudinal speed, ξ p Calculated by the following formula:

[0100]

[0101] Where k is the cornering stiffness estimation trigger threshold coefficient.

[0102] When the vehicle speed is high, p Decrease to increase the tire cornering stiffness estimation update frequency, otherwise increase ξ p .

[0103] The beneficial effects of the present invention are:

[0104] 1) The present invention adopts a dual Kalman filter parallel estimation mechanism and designs an architecture in which the state estimator and parameter estimator interact in parallel. Adaptive cubature Kalman filtering is used for state estimation and adaptive Kalman filtering is used for parameter estimation, respectively, to reduce the mutual interference between different estimation variables and improve estimation accuracy and system stability.

[0105] 2) This invention proposes an event triggering mechanism based on fuzzy logic, introduces Takagi-Sugeno fuzzy rules, and dynamically adjusts the event triggering threshold based on changes in observation residuals and measurement covariance. This solves the problem that traditional fixed thresholds are difficult to adapt to changing working conditions. It has strong adaptability and practicality, and effectively controls error accumulation.

[0106] 3) The present invention achieves high-precision estimation at a low computational cost. The proposed method reduces the update frequency through event triggering, thereby reducing computing resource consumption without sacrificing estimation performance. It is suitable for intelligent vehicle control systems with high real-time requirements, and its accuracy is better than that of traditional methods in simulation verification. BRIEF DESCRIPTION OF THE DRAWINGS

[0107] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for use in the embodiments. It should be understood that the following drawings only illustrate certain embodiments of the present invention and therefore should not be regarded as limiting the scope. For ordinary technicians in this field, other relevant drawings can be obtained based on these drawings without paying any creative work.

[0108] Figure 1 This is an architecture diagram of the vehicle state and parameter joint estimation method based on fuzzy event-triggered dual Kalman filtering;

[0109] Figure 2 This is a schematic diagram of a single-track four-wheel steering vehicle model;

[0110] Figure 3 Schematic diagram of the joint estimator of vehicle state and parameters;

[0111] Figure 4 Schematic diagram of TS fuzzy rule framework;

[0112] Figure 5 This is a schematic diagram of the steering wheel angle under sinusoidal steering conditions;

[0113] Figure 6a This is a comparison diagram of vehicle longitudinal velocity estimation under sinusoidal steering conditions;

[0114] Figure 6b This is a comparison diagram of vehicle lateral velocity estimation under sinusoidal steering conditions;

[0115] Figure 6c This is a comparison diagram of the vehicle center of mass sideslip angle estimation under sinusoidal steering conditions;

[0116] Figure 7a Comparison chart of front axle tire cornering stiffness estimation under sinusoidal steering conditions;

[0117] Figure 7b This is a comparison chart of the estimated cornering stiffness of the rear axle tire under sinusoidal steering conditions;

[0118] Figure 8a This is a schematic diagram of the trigger flag for vehicle state estimation under sinusoidal steering conditions;

[0119] Figure 8b Schematic diagram of the trigger flag for tire cornering stiffness estimation under sinusoidal steering conditions. DETAILED DESCRIPTION

[0120] The present invention will be further described in detail below with reference to the accompanying drawings and examples. It will be understood that the specific embodiments described herein are intended only to illustrate the present invention and are not intended to limit the present invention. It should also be noted that, for ease of description, the accompanying drawings only illustrate portions relevant to the present invention, not all structures.

[0121] Example 1

[0122] See Figure 1 In this embodiment, a vehicle state and parameter joint estimation method based on fuzzy event-triggered dual Kalman filtering includes the following steps:

[0123] Step 1: Establish the kinematic and dynamic model of the four-wheel steering vehicle, as follows:

[0124] 11) This paper focuses on the longitudinal, lateral and yaw motion of the vehicle based on the single-track four-wheel steering vehicle model. Assuming that the wheel angles on the front and rear axles are the same, the vehicle kinematics and dynamics equations are combined to establish the following Figure 2 The three-degree-of-freedom model of the vehicle is shown:

[0125]

[0126]

[0127]

[0128] Where V x is the longitudinal velocity; V y is the lateral velocity; ω r is the yaw angular velocity; a x is the longitudinal acceleration; a y is the lateral acceleration; I z is the moment of inertia of the vehicle around the z-axis; a and b are the distances from the front and rear axles to the center of mass of the vehicle, respectively; δ f , δ r are the wheel angles of the front and rear axles respectively; F yf 、F yr are the cornering forces of the front and rear axle wheel tires respectively;

[0129] 12) Under normal driving conditions, the relationship between tire lateral force and tire slip angle is considered to be as follows:

[0130]

[0131] Where, F yf 、F yr are the cornering forces of the front and rear axle tires respectively; C f 、C r is the lateral stiffness of the front and rear axle wheels; α f , α r is the side slip angle of the tires on the front and rear axles of the vehicle;

[0132] 13) The side slip angle of the front and rear axle tires of the vehicle is calculated by the following formula:

[0133]

[0134] Where, α f , α r V is the side slip angle of the front and rear axle wheels of the vehicle; x is the longitudinal velocity; V y is the lateral velocity; ω r is the yaw angular velocity; a and b are the distances from the front and rear axles to the center of mass of the vehicle respectively; δ f , δ r are the wheel angles of the front and rear axles respectively;

[0135] 14) Combining equations (3), (4), and (5), we can get the kinematic-dynamic model equation of the four-wheel steering vehicle:

[0136]

[0137] Where V x is the longitudinal velocity; V y is the lateral velocity; ω r is the yaw angular velocity; a x is the longitudinal acceleration; a y is the lateral acceleration; I z is the moment of inertia of the vehicle around the z-axis; a and b are the distances from the front and rear axles to the center of mass of the vehicle, respectively; δ f , δ r are the front and rear axle wheel angles respectively; C f 、C r is the wheel cornering stiffness of the front and rear axles;

[0138] 15) In actual estimation systems, a discrete-time model is often required for numerical calculations and real-time processing. To achieve this, the continuous-time vehicle model is discretized using the Euler method, resulting in a discrete state-space equation of the following form, as shown below:

[0139]

[0140] Where x kis the discrete system state variable, x k =[V x,k ,V y,k ,ω r,k ] T ;z k is the measured variable of the discrete system, z k =ω r,k ; f(·) is the state transfer equation of the discrete system; h(·) is the measurement equation of the discrete system.

[0141] Step 2: Combine the joint estimation method of the four-wheel steering vehicle kinematics and dynamics models to construct a joint estimator of vehicle state and parameters based on the three-degree-of-freedom equation, as follows:

[0142] For vehicle state V x 、V y 、ω r and tire cornering stiffness C f 、C r Simultaneously make estimates such as Figure 3 The two estimators are used in parallel to reduce the influence of different estimators. The vehicle state estimator and the tire cornering stiffness estimator transmit estimation results in real time and exchange information.

[0143] 21) For the estimation of tire cornering stiffness, the applied equation has a linear form, so the adaptive Kalman filter method is used to estimate it;

[0144] As the core parameter connecting the tire slip angle and lateral force, tire cornering stiffness is of great significance in describing the lateral dynamic characteristics of the vehicle. Therefore, the state variable of the tire cornering stiffness estimator is selected as x p =[C f ,C r ] T , the lateral acceleration and yaw angular acceleration that can be directly measured from the sensor are selected as the measurement variables The required input for the tire cornering stiffness estimator is u p =[V x ,V y ,ω r ,δ f ,δ r ].

[0145] The state prediction equation for tire cornering stiffness estimation is:

[0146]

[0147] The tire cornering stiffness estimation measurement equation is:

[0148]

[0149] Where C f,k 、C r,k C is the wheel cornering stiffness of the front and rear axles at time k; f,k+1 、C r,k+1 is the wheel cornering stiffness of the front and rear axles at time k+1; is the yaw angular acceleration at time k+1; a y,k+1 is the lateral acceleration at time k+1; δ f,k , δ r,k are the front and rear axle wheel angles at time k+1; α f,k , α r,k I is the side slip angle of the front and rear axle tires of the vehicle at time k+1; z is the moment of inertia of the vehicle around the z-axis; a and b are the distances from the front and rear axles to the center of mass of the vehicle, respectively; and m is the mass of the vehicle.

[0150] 22) For vehicle state estimation, the present invention adopts an adaptive cubature Kalman filter method;

[0151] The state variable of the vehicle state estimator is selected as x s =[V x ,V y ,ω r ] T The measured variable is selected as the yaw rate z directly obtained from the vehicle inertial sensor s =ω r , the required input of the vehicle state estimator is u s =[a x ,a y ,δ f ,δ r ,C f ,C r ] T ;

[0152] The cubature Kalman filter generates a set of "cubature points" and then propagates and updates at these points to improve the filtering accuracy. The iterative steps of the adaptive cubature Kalman filter algorithm are as follows.

[0153] First, the implementation of the adaptive volumetric Kalman filter algorithm requires a preset initial state. x,0 , lateral velocity V y,0 and yaw rate ω r,0 The state vector x s,0 As the initial value, initialize the error covariance matrix P s,0 , the initial covariance Q of the process noise ω s,0 and the initial covariance R of the measurement noise v s,0 .

[0154] initialization:

[0155]

[0156] At the initial moment, there is x s,k =x s,0 ,P s,k =P s,0 ;

[0157] The covariance matrix P is decomposed into s,k Convert to lower triangular matrix S k :

[0158]

[0159] Let the state dimension be n and construct 2n volume points:

[0160]

[0161] Using unit orthogonal vector ξ i Construct a volume point set:

[0162]

[0163] Where i is the volume point sequence number; n is the dimension of the state variable; [1] i is the i-th column of the n-dimensional identity matrix; the volume points corresponding to the vehicle state estimation method are determined by the symmetric sampling rule of the three-dimensional state space:

[0164]

[0165] In the prediction stage, each volume point propagates the vehicle state prediction equation (15) through the nonlinear state equation, which takes into account the coupling effect of lateral and longitudinal acceleration and yaw motion:

[0166]

[0167] Where, T s is the sampling time; [V x,k+1 ,V y,k+1 ,ω r,k+1 ] T is the value of the state variable at the next sampling moment;

[0168]

[0169] Where, is the predicted value of the propagation volume point at time k+1 at time k.

[0170] The propagated volume points are weighted averaged to obtain the state prediction value x s,k+1|k , and update the prediction covariance matrix P s,k+1|k, the process explicitly includes process noise Q s,k Impact:

[0171]

[0172] After entering the measurement update phase, the algorithm regenerates volume points:

[0173]

[0174] Since the yaw angular velocity ω r Direct measurement, the vehicle state measurement equation is simplified to the identity mapping:

[0175] ω r,+1 =ω r,k+1 (twenty two)

[0176] Volumetric Point Propagation:

[0177]

[0178] The measured volume points after propagation calculate the predicted value of the measured variable z s,k+1 , and obtain the measurement variable covariance matrix and the measured variable-predictor variable cross-covariance matrix for:

[0179]

[0180] Then, the state estimate x is updated by fusing the prediction and measurement information through the Kalman gain formula s,k+1 and the covariance matrix P s,k+1 :

[0181]

[0182]

[0183]

[0184] Where, is the actual measurement value at the k+1 sampling time;

[0185] In order to improve the adaptability to complex working conditions, the algorithm uses the exponential weighting method to dynamically adjust the measurement noise covariance matrix R s,k+1 , the adjustment range depends on the current observation residual e k+1 :

[0186]

[0187]

[0188] Where H is the observation matrix of the measured variables;

[0189] Finally, using the updated lateral / longitudinal velocity estimates, the vehicle's sideslip angle is calculated from the geometric relationship:

[0190]

[0191] Step 3: Design an event trigger mechanism based on fuzzy rules to dynamically adjust the trigger threshold of state estimation, as follows:

[0192] 31) In the state estimation process based on the vehicle kinematic equation, the integration operation may amplify the sensor measurement noise and introduce cumulative errors, thereby affecting the estimation accuracy.

[0193] In order to balance computational efficiency and estimation accuracy, traditional methods often use a fixed threshold event trigger mechanism, that is, when the observation residual e expressed by formula (28) k When the modulus exceeds the preset threshold, the status update is triggered:

[0194]

[0195] Where, ξ s The preset trigger threshold.

[0196] However, fixed threshold triggering mechanisms have certain shortcomings and are difficult to adapt to different driving conditions. Therefore, this paper proposes an adaptive event triggering mechanism based on fuzzy rules, which aims to dynamically adjust the state estimation update frequency to reduce the computational burden and mitigate the impact of cumulative errors.

[0197] The trigger threshold has limitations if it relies solely on the empirical value of the observation residual for triggering, because the increase in the observation residual may be caused by different types of noise, but it is impossible to directly distinguish whether it is prediction process noise or measurement noise.

[0198] In the volumetric Kalman estimation method, the measurement error covariance matrix is ​​calculated by propagating the measurement noise through the volume point, and the difference in the measurement error covariance matrix can represent the impact of the change in measurement noise on the system. The difference in the measurement error covariance matrix is:

[0199]

[0200] When the difference in the measurement error covariance matrix is ​​large, it means that the measurement noise has a greater impact on the measurement prediction value.

[0201] The modulus of the observed residual |e k The modulus of the difference between | and the measurement error covariance matrix Comprehensively determining the event trigger threshold can further determine which type of noise mainly causes the unreliability of the predicted state, and thus trigger the update of the state estimation more reasonably.

[0202] Adding fuzzy rules can map multivariate inputs to set trigger events and enhance the adaptability of the estimation system.

[0203] The present invention adopts Figure 4 The TS fuzzy system is shown to adaptively adjust the trigger threshold; the rules of the TS fuzzy system are:

[0204] R j :

[0205] If x1 is x2 is ……,x n yes Then y j =f j (X),j=1,2,…,M,X=(x1,x2,…,x n ) T , where x i is the input variable, i=1,2,……,n, It is a fuzzy subset, and the input variables are x1=e,x2=ΔP z In order to realize the fuzzy inference system, the input variables are divided into three fuzzy subsets S = Small, M = Middle, B = Big, N is the total number of fuzzy rules, f j (·) is the rule R j The consequent function, y j is the output, and the activation strength of the j-th rule is given by:

[0206]

[0207] Where, is x i exist The degree of membership in ;

[0208] The output Y is calculated by aggregating the contributions of each rule:

[0209]

[0210] The fuzzy system has two input variables and three fuzzy sets, and the following nine rules are defined:

[0211]

[0212] Finally, the output ξ s for:

[0213]

[0214] 32) The present invention also designs an adaptive event triggering mechanism based on the vehicle's lateral dynamic changes in the tire cornering stiffness estimator. Obtaining tire cornering stiffness is particularly important when the vehicle experiences severe lateral motion, which is closely related to wheel angle. The present invention selects the difference in front wheel angle as the triggering event for tire cornering stiffness estimation:

[0215] Δδ f,k =δ f,k -δ f,k-1 (37)

[0216] Fuzzy logic is used to dynamically adjust the trigger threshold by integrating the effects of observation residuals and measurement error covariance, thereby enhancing the robustness and adaptability of the estimation process. This not only enables more precise control of the trigger frequency but also more effectively suppresses error accumulation, achieving a better balance between computational cost and estimation accuracy.

[0217] Considering the coupling effect of the longitudinal and lateral motion of the vehicle, under the same wheel angle difference, the greater the longitudinal speed of the vehicle, the more severe the lateral motion of the vehicle. Therefore, the event trigger threshold ξ of the tire cornering stiffness estimator is p It should be able to adjust adaptively with the change of vehicle longitudinal speed, ξ p Calculated by the following formula:

[0218] Event trigger threshold ξ of tire cornering stiffness estimator p It should be able to adjust adaptively with the change of vehicle longitudinal speed, ξ p Calculated by the following formula:

[0219]

[0220] Where k is the cornering stiffness estimation trigger threshold coefficient.

[0221] When the vehicle speed is high, p Decrease to increase the tire cornering stiffness estimation update frequency, otherwise increase ξ p .

[0222] Example 2

[0223] This embodiment uses a joint simulation platform based on MATLAB / Simulink and vehicle dynamics software CarSim to test the vehicle state and parameter joint estimation method based on fuzzy event-triggered dual Kalman filter designed by the present invention. In order to compare and verify the effect of the proposed estimator, a cubature Kalman filter estimator and an event-triggered cubature Kalman filter estimator based on a fixed threshold are established. The three estimators use the same parameters. A simulation verification is carried out under a sinusoidal steering condition with a relatively violent steering operation. The steering wheel angle set in the simulation is as follows: Figure 5As shown in the figure. In the simulation, the road condition is set to an adhesion coefficient of 0.85 and the vehicle speed is constant, set to 72 km / h. In the fixed threshold-based event-triggered VCKF estimator, the fixed trigger thresholds of the state estimator and the cornering stiffness estimator are 1×10 -3 and 1×10 -5 ,In the three estimation methods, the initial longitudinal vehicle speed is set to 20m / s, and the initial lateral vehicle speed and yaw angular velocity are set to 0. Figure 6a 、 Figure 6b and Figure 6c The dynamic changes of the vehicle state estimation values ​​are shown. It can be seen that due to the frequent and drastic changes in the longitudinal vehicle speed, the estimation method using the event trigger mechanism is more effective in the estimation of the longitudinal vehicle speed and can converge well to the reference value. Since the method adopts an adaptive method to determine the trigger threshold, it is more sensitive to external changes and can achieve better estimation results. In the estimation of lateral speed and center of mass slip angle, it can be seen that the three estimation methods all produce different degrees of deviation. Among them, the estimation result of the method at the peak position is closer to the reference value. The root mean square error and range of the center of mass slip angle estimation of the method are 0.0125 and 0.0282 respectively. The estimation accuracy is 33.2% and 4.6% higher than that of the cubature Kalman filter estimator and the event-triggered cubature Kalman filter estimator based on a fixed threshold. In the lateral stiffness estimator, the initial front and rear axle lateral stiffness are set to -1×10 5 N / rad, -5×10 4 N / rad, given by Figure 7a and Figure 7b It can be seen that due to the large lateral excitation during the entire sinusoidal steering process, the cornering stiffness estimator in the proposed method can quickly converge to the actual cornering stiffness and track the cornering stiffness of the front and rear axle wheels in real time. Figure 8a and Figure 8b It can be seen that the vehicle state estimator and lateral stiffness estimator of the described method can be triggered when specific conditions are met. During the sinusoidal steering condition, due to the drastic state changes, the estimator is frequently turned on to achieve rapid tracking of the estimated target; however, compared with the estimator that is turned on throughout the process, the estimator based on the event trigger mechanism can still reduce the update frequency to a certain extent, thereby reducing the cumulative error and the use of computing resources.

[0224] While embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions, and variations may be made to these embodiments without departing from the principles and spirit of the invention, and that the scope of the invention is defined by the appended claims and their equivalents.

Claims

1. A joint estimation method for vehicle state parameters based on fuzzy event triggering mechanism, characterized in that: The following steps are involved: Step 1: Establish the kinematic and dynamic model of the four-wheel steering vehicle; Step 2: Combine the joint estimation method of the four-wheel steering vehicle kinematic and dynamic models to construct a joint estimator of vehicle state and parameters based on the three-degree-of-freedom equation; Step 3: Design an event trigger mechanism based on fuzzy rules to dynamically adjust the trigger threshold of state estimation.

2. The vehicle state parameter joint estimation method based on fuzzy event triggering mechanism according to claim 1 is characterized in that: The specific method of step one is as follows: 11) Assuming the wheel angles on the front and rear axles are equal, a three-degree-of-freedom model of the vehicle is established by combining the vehicle kinematics and dynamics equations: Where V x is the longitudinal velocity; V y is the lateral velocity; ω r is the yaw angular velocity; a x is the longitudinal acceleration; a y is the lateral acceleration; I z is the moment of inertia of the vehicle around the z-axis; a and b are the distances from the front and rear axles to the center of mass of the vehicle, respectively; δ f , δ r are the wheel angles of the front and rear axles respectively; F yf 、F yr are the cornering forces of the front and rear axle wheel tires respectively; 12) The relationship between tire lateral force and tire slip angle is as follows: Where, F yf 、F yr are the cornering forces of the front and rear axle tires respectively; C f 、C r is the lateral stiffness of the front and rear axle wheels; α f , α r is the side slip angle of the tires on the front and rear axles of the vehicle; 13) The side slip angle of the front and rear axle tires of the vehicle is calculated by the following formula: Where, α f , α r V is the side slip angle of the front and rear axle wheels of the vehicle; x is the longitudinal velocity; V y is the lateral velocity; ω r is the yaw angular velocity; a and b are the distances from the front and rear axles to the center of mass of the vehicle respectively; δ f , δ r are the wheel angles of the front and rear axles respectively; 14) Combining equations (3), (4), and (5), we can get the kinematic-dynamic model equation of the four-wheel steering vehicle: Where V x is the longitudinal velocity; V y is the lateral velocity; ω r is the yaw angular velocity; a x is the longitudinal acceleration; a y is the lateral acceleration; I z is the moment of inertia of the vehicle around the z-axis; a and b are the distances from the front and rear axles to the center of mass of the vehicle, respectively; δ f , δ r are the front and rear axle wheel angles respectively; C f 、C r is the wheel cornering stiffness of the front and rear axles; 15) The continuous-time vehicle model is discretized using the forward Euler method to obtain the following discrete state space equations: Where x k is the discrete system state variable, x k =[V x,k ,V y,k ,ω r,k ] T ;z k is the measured variable of the discrete system, z k =ω r,k ; f(·) is the state transfer equation of the discrete system; h(·) is the measurement equation of the discrete system.

3. The vehicle state parameter joint estimation method based on fuzzy event triggering mechanism according to claim 1 is characterized in that: The specific method of step 2 is as follows: 21) Adaptive Kalman filtering method is used to estimate tire cornering stiffness; 22) Adaptive cubature Kalman filtering method is used to estimate vehicle state.

4. The method according to claim 3, wherein: The specific method of step 21) is as follows: The state variable of the tire cornering stiffness estimator is selected as x p =[C f ,C r ] T , the lateral acceleration and yaw angular acceleration measured directly from the sensor are selected as the measurement variables The required input for the tire cornering stiffness estimator is u p =[V x ,V y ,ω r ,δ f ,δ r ]; The state prediction equation for tire cornering stiffness estimation is: The tire cornering stiffness estimation measurement equation is: Where C f,k 、C r,k C is the wheel cornering stiffness of the front and rear axles at time k; f,k+1 、C r,k+1 is the wheel cornering stiffness of the front and rear axles at time k+1; is the yaw angular acceleration at time k+1; a y,k+1 is the lateral acceleration at time k+1; δ f,k , δ r,k are the front and rear axle wheel angles at time k+1; α f,k , α r,k I is the side slip angle of the front and rear axle tires of the vehicle at time k+1; z is the moment of inertia of the vehicle around the z-axis; a and b are the distances from the front and rear axles to the center of mass of the vehicle, respectively; and m is the mass of the vehicle.

5. The vehicle state parameter joint estimation method based on fuzzy event triggering mechanism according to claim 3 is characterized in that: The specific method of step 22) is as follows: The state variable of the vehicle state estimator is selected as x s =[V x ,V y ,ω r ] T The measured variable is selected as the yaw rate z directly obtained from the vehicle inertial sensor s =ω r , the required input of the vehicle state estimator is u s =[a x ,a y ,δ f ,δ r ,C f ,C r ] T ; The vehicle longitudinal speed V x,0 , lateral velocity V y,0 and yaw angular velocity ω r,0 The state vector x s,0 As the initial value, initialize the error covariance matrix P s,0 , the initial covariance Q of the process noise ω s,0 and the initial covariance R of the measurement noise v s,0 ; initialization: At the initial moment, there is x s,k =x s,0 ,P s,k =P s,0 ; The covariance matrix P is decomposed into s,k Convert to lower triangular matrix S k : Let the state dimension be n and construct 2n volume points: Using unit orthogonal vector ξ i Construct a volume point set: Where i is the volume point sequence number; n is the dimension of the state variable; [1] i is the i-th column of the n-dimensional identity matrix; the volume points corresponding to the vehicle state estimation method are determined by the symmetric sampling rule of the three-dimensional state space: In the prediction stage, each volume point propagates the vehicle state prediction equation (15) through the nonlinear state equation, which takes into account the coupling effect of lateral and longitudinal acceleration and yaw motion: Where, T s is the sampling time; [V x,k+1 ,V y,k+1 ,ω r,k+1 ] T is the value of the state variable at the next sampling moment; Where, is the predicted value of the propagation volume point at time k+1 at time k. The propagated volume points are weighted averaged to obtain the state prediction value x s,k+1|k , and update the prediction covariance matrix P s,k+1|k , the process explicitly includes process noise Q s,k Impact: After entering the measurement update phase, the algorithm regenerates volume points: Since the yaw angular velocity ω r Direct measurement, the vehicle state measurement equation is simplified to the identity mapping: oh r,k+1 =ω r,k+1 (22) Volumetric Point Propagation: The measured volume points after propagation calculate the predicted value of the measured variable z s,k+1 , and obtain the measurement variable covariance matrix and the measured variable-predictor variable cross-covariance matrix for: Then, the state estimate x is updated by fusing the prediction and measurement information through the Kalman gain formula s,k+1 and the covariance matrix P s,k+1 : Where, is the actual measurement value at the k+1 sampling time; The exponential weighting method is used to dynamically adjust the measurement noise covariance matrix R s,k+1 , the adjustment range depends on the current observation residual e k+1 : Where H is the observation matrix of the measured variables; Finally, using the updated lateral / longitudinal velocity estimates, the vehicle's sideslip angle is calculated from the geometric relationship:

6. The vehicle state parameter joint estimation method based on fuzzy event triggering mechanism according to claim 1 is characterized in that: The specific method of step three is as follows: 31) The vehicle state estimator uses the observation residual as the trigger event, and the preset trigger threshold is adaptively adjusted based on the changes in the observation residual and the measurement error covariance; 32) The parameter estimator uses the change in front wheel angle as a trigger event, and the preset trigger threshold is related to the vehicle speed.

7. The vehicle state parameter joint estimation method based on fuzzy event triggering mechanism according to claim 6 is characterized in that: The specific method of step 31) is as follows: The difference in the measurement error covariance matrix can represent the impact of changes in measurement noise on the system. The difference in the measurement error covariance matrix is: The TS fuzzy system is used to adaptively adjust the trigger threshold; the rules of the TS fuzzy system are: R j ： If x1 is x2 is yes Then y j =f j (X),j=1,2,…,M,X=(x1,x2,…,x n ) T , where x i is the input variable, It is a fuzzy subset, and the input variables are x1=e,x2=ΔP z In order to realize the fuzzy inference system, the input variables are divided into three fuzzy subsets S = Small, M = Middle, B = Big, N is the total number of fuzzy rules, f j (·) is the rule R j The consequent function, y j is the output, and the activation strength of the j-th rule is given by: Where, is x i exist The degree of membership in ; The output Y is calculated by aggregating the contributions of each rule: The fuzzy system has two input variables and three fuzzy sets, and the following nine rules are defined: R1:If e=S andΔP z =S,then y1=0.0005; R2:If e=S andΔP z =M,then y2=0.0025; R9:If e=B andΔP z =B,then y9=0; Finally, the output ξ s for:

8. The vehicle state parameter joint estimation method based on fuzzy event triggering mechanism according to claim 6 is characterized in that: The specific method of step 32) is as follows: The vehicle front wheel angle difference is selected as the trigger event for tire cornering stiffness estimation: Dd f,k =d f,k -d f,k-1 (37) Event trigger threshold ξ of tire cornering stiffness estimator p It should be able to adjust adaptively with the change of vehicle longitudinal speed, ξ p Calculated by the following formula: Where k is the cornering stiffness estimation trigger threshold coefficient. When the vehicle speed is high, p Decrease to increase the tire cornering stiffness estimation update frequency, otherwise increase ξ p .

Images