A real-time dynamic modeling method for heavy vehicles

Abstract

The present invention relates to a method for real-time dynamic modeling of heavy vehicles. The method includes the following steps: establishing a three-degree-of-freedom vehicle dynamics model and a tire model to obtain vehicle states and chassis dynamics states; establishing a road surface state equation to obtain road surface states; judging the cargo state, and combining the vehicle states, chassis dynamics states and road surface states to establish a vehicle-road-load integrated nonlinear state equation including state inputs and scheduling parameters; obtaining the mapping relationship between vehicle configurations, road surface states and cargo states and the observable state space; constructing a kernel function based on linearized feature extraction according to the vehicle-road-load integrated nonlinear state equation; based on the kernel function, in the data integration recursive moving window, the initial state of the window is discarded in real time and the data at the k+1 moment is incorporated to obtain an updated vehicle-road-load integrated linear state equation. The present invention is used for heavy vehicle dynamics modeling and realizes real-time state update.

Description

Technical Field

[0001] The present invention belongs to the fields of vehicle engineering and dynamic system modeling, and specifically relates to a real-time dynamic modeling method for heavy vehicles. Background Art

[0002] With the rapid development of the logistics industry and autonomous driving technology, heavy vehicles have strong loading capacity and are widely used. However, the vehicle state variables involved are numerous and complex, and road conditions and cargo states also need to be considered. In order to accurately control heavy vehicles and carry out a series of engineering applications, there is an urgent need to propose a real-time dynamic modeling method for heavy vehicles that is widely applicable, highly versatile, and can achieve real-time update of vehicle states. First, although the traditional method based on physical modeling has strong interpretability, it depends on complex experimental parameters and premise assumptions, is difficult to accurately reflect the dynamic behavior under extreme conditions, and cannot achieve real-time update of vehicle data. Due to the higher requirements for vehicle dynamic modeling in the performance optimization and control algorithm design of heavy vehicles, the real-time modeling method based on physical-data update has emerged, but it still cannot break through the fixed scenario limitation of "specific vehicle for specific use", resulting in inaccurate vehicle data update. Therefore, there are drawbacks in dealing with variable road environments and versatility. To sum up, the current heavy vehicle dynamic modeling methods still have the following deficiencies:

[0003] Specific vehicle for specific use: Model design often targets specific vehicles and is only applicable to specific application scenarios, resulting in a long vehicle development cycle and inability to flexibly respond to various task requirements.

[0004] Fixed scenario limitation: Most existing models assume specific road conditions (such as flat or specific friction coefficient) and cargo configurations, and cannot adapt to complex road and cargo changes.

[0005] Insufficient real-time performance: Traditional models are usually constructed offline and are difficult to update in real time according to the dynamic changes of vehicle states and the environment, thus affecting control accuracy. Summary of the Invention

[0006] The purpose of the present invention is to provide a real-time dynamic modeling method for heavy vehicles, which breaks through the limitations of "specific vehicle for specific use and fixed scenario" of traditional dynamic models, and realizes the construction of a unified dynamic model for heavy vehicles and real-time state update.

[0007] The technical solution to achieve the purpose of the present invention is: A real-time dynamic modeling method for heavy vehicles, comprising the following steps:

[0008] S1: Establish a three-degree-of-freedom vehicle dynamic model and a tire model, and obtain vehicle states and chassis dynamic states;

[0009] S2: Establish a road surface state equation and obtain road surface states;

[0010] S3: Determine the cargo status, and combine the vehicle status, chassis dynamics status, and road surface status to establish an integrated vehicle-road-load nonlinear state equation that includes state inputs and scheduling parameters; obtain the mapping relationship between the vehicle configuration, road surface status, and cargo status and the observable state space;

[0011] S4: According to the integrated vehicle-road-load nonlinear state equation, construct a Koopman kernel function based on linearized feature extraction;

[0012] S5: Based on the Koopman kernel function, in the data integration recursive moving window, discard the initial state of the window in real time and incorporate the data at the (k + 1)th moment to obtain an updated integrated vehicle-road-load linear state equation.

[0013] Furthermore, the three-degree-of-freedom vehicle dynamics model established in S1 is as follows:

[0014] (1)

[0015] In the formula, m is the vehicle's total mass, F f is the air resistance, F w is the rolling resistance, is the longitudinal acceleration, is the lateral acceleration, is the wheel steering angle, is the track width, j = l, r represent the front and rear wheels of the vehicle respectively, i = f, m, l represent the front axle, middle axle, and rear axle of the vehicle respectively, L i is the distance from the front and rear axles of the vehicle to the center of mass, and are the yaw angular velocity and yaw angular acceleration of the vehicle respectively, is the moment of inertia of the vehicle body about the Z1 axis, I x is the moment of inertia of the vehicle body about the X1 axis, and are the longitudinal tire force and lateral tire force of the th wheel respectively; when the th wheel is in front of the vehicle's center of gravity, " " is positive, and when it is behind the vehicle's center of gravity, it is negative; when the th wheel is on the left side of the vehicle, " " is positive, and when it is on the right side of the vehicle, it is negative;

[0016] The tire model is:

[0017] (2)

[0018] In the formula, Y is the lateral force, X is the sideslip angle, D is the peak factor, B is the stiffness factor, C is the curve shape factor, E is the curve curvature factor, S hFor the horizontal drift of the curve, S v For the vertical drift of the curve.

[0019] Furthermore, the road surface state equation established in S2 is:

[0020] (3)

[0021] where m is the vehicle mass, is the sideslip angle, is the vehicle speed, is the vehicle acceleration, g is the gravitational acceleration, is the air resistance coefficient, is the contact area between the wheel and the ground, is the air density, is the engine torque, is the engine transmission ratio, is the transmission ratio of the gearbox, r is the wheel radius, is the transmission efficiency, f is the ground friction force.

[0022] Furthermore, "judging the cargo state" in S3 is to establish a cargo state equation to judge the rigid-flexible state of the cargo; specifically:

[0023] S31: When the vehicle is stationary, the vertical force balance equation of the vehicle is obtained from the moment balance, force balance and geometric relationship of the suspension spring as follows:

[0024] (4)

[0025] In the formula, is the vehicle unloaded mass, 、 、 are the spring stiffnesses of the front, middle and rear suspensions respectively, taking = = ; 、 、 are the suspension forces of the front, middle and rear axles respectively, l f 、l m 、l r are the distances from the front axle to the center of mass, the distance from the middle axle to the center of mass, and the distance from the rear axle to the center of mass respectively;

[0026] S32: The calculation and analysis of the static vertical load of the wheel are as follows:

[0027] (5)

[0028] In the formula, 、 are the static vertical loads of the right and left front axles respectively, and are the static vertical loads of the right and left middle axles respectively, and are the static vertical loads of the right and left rear axles respectively;

[0029] S33: Calculate and analyze the dynamic vertical load of the wheels as follows:

[0030] (6)

[0031] where is the vertical displacement of the center of gravity of each wheel, is the displacement of each suspension, is the vertical displacement of the vehicle body, is the height difference between the inner and outer sides of the road surface; is the sideslip angle of the vehicle, is the yaw angle of the vehicle; is the spring stiffness of each wheel, is the spring stiffness of each suspension, is the tire damping of each wheel, is the dynamic vertical load of each wheel; when the wheel is in front of the vehicle's center of gravity, " " is taken as positive and the cargo is flexible; when it is behind the vehicle's center of gravity, it is taken as negative and the cargo is rigid; when the wheel is on the left side of the vehicle, " " is taken as positive and the cargo is rigid; when it is on the right side of the vehicle, it is taken as negative and the cargo is flexible.

[0032] Furthermore, in S3, "Combining the vehicle state, chassis dynamics state, and road surface state, establish a vehicle-road-load integrated non-linear state equation including state inputs and scheduling parameters" is specifically

[0033] S34: Transform Equation (1) into an explicit equation about , , , and set up a general heavy vehicle model in the form of a non-linear state space by combining the vehicle dynamics equation, road surface state equation, and cargo state equation:

[0034] (7)

[0035] where h s is the height of the vehicle's center of mass from the ground, m s is the full load mass, F x , F y are the lateral force and longitudinal force respectively, M z is the torque of the vehicle about the z-axis;

[0036] S35: The vehicle-road integrated vehicle state space form equation obtained after linearization is:

[0037] (8)

[0038] (9)

[0039] (10)

[0040] In the formula, F c is the control force, L c is the control gain, A is the system matrix, X b is the state vector, B is the input matrix, U is the input vector, C is the bias vector, A F is the additional system matrix, is the increment of the input vector; F is the final control input, is the gain coefficient; is the time derivative of the state vector, B F is the force input matrix;

[0041] The general non-linear state equation expression is as follows:

[0042] (11)

[0043] In the formula is the time derivative of the state vector, is the non-linear function describing the system dynamics, is the predicted control force;

[0044] Define the state vector of the system , and obtain:

[0045] ,

[0046]

[0047] Furthermore, S4 "Construct the Koopman kernel function based on linearized feature extraction" is specifically:

[0048] (12)

[0049] In the formula , is the scheduling parameter; , , ;

[0050] S42: Discretize Equation (12) and represent it as:

[0051] (13)

[0052] In the formula

[0053] S43: By introducing the Kronecker product, Equation (13) is written as:

[0054] (14)

[0055] where A jk (j = 0, 1, 2, 3) is the lumped state - related matrix, B jk (j = 0, 1, 2, 3) is the lumped input - related matrix, is the Kronecker product operator, 、 、 are data trajectories;

[0056] S44: The equation at the previous moment of Equation (13) is expressed as:

[0057] (15)

[0058] Through the inverse operation of the matrix, the system matrix is solved as:

[0059] (16)

[0060] S45: Obtain the Koopman kernel function based on the Koopman operator and with a moving window to recursively update the vehicle state data, and its formula is expressed as:

[0061] (17)

[0062] where is both the Koopman operator and the system transformation matrix, 、 、 are the data trajectories of the input variable, state variable, and scheduling parameter variable respectively, 、 are the transfer variable matrix and the update variable matrix respectively.

[0063] Furthermore, S5 includes:

[0064] S51: 、 、 are the data trajectories of the input variable, state variable, and scheduling parameter variable respectively. The data is updated in real - time using the moving - window method, and the window length is n d data points. Define a moving window and express it as:

[0065] (18)

[0066] In is a data snapshot matrix, which contains combinations of inputs, states, and scheduling parameters, and is a dataset for a moving window in real-time modeling. is an input matrix that contains to time system input u, X k is a state matrix that contains from to time system state x, X Fk is a future state matrix that contains from to time future system state; and are respectively the Kronecker inner product matrices of the input state and the scheduling parameter ρ. is a scheduling parameter vector, n d The number of data points in the moving window; is a Kronecker inner product operator used to generate combinations between inputs, states, and scheduling parameters;

[0067] S52: Rewrite the moving window based on Equations (17) and (18) as:

[0068] (19)

[0069] S53: Using the vehicle state data collected by Equation (18), obtain the system transformation matrix expressed as:

[0070] (20)

[0071] S54: The recursive update formula is expressed as:

[0072] (21)

[0073] S55: Assume the relationship between the system state x at time steps k and k + 1 is:

[0074] (22)

[0075] S56: Use the data points collected at time k + 1 , , to update the system transformation matrix , and realize real-time update of the data for the integrated vehicle-road-load modeling of heavy vehicles. The updated integrated state space is expressed as:

[0076] (23)

[0077] In the formula is the state matrix at time k after update, is the longitudinal velocity at time k after update, is the lateral velocity at time k after update, and are the yaw angular velocity and yaw angular acceleration of the vehicle at time k after update respectively.

[0078] Compared with the prior art, the remarkable advantages of the present invention are as follows:

[0079] The present invention establishes an integrated general model to perform real-time update of vehicle data. Based on the vehicle physical model, the present invention uses a moving window to iteratively update the vehicle state. Compared with the traditional physical modeling method, the present invention breaks through the usage limitation of one vehicle for one use, and the integrated modeling can capture dynamic changes in real time in a complex road environment; compared with the physical-data modeling method, the present invention can not only comprehensively consider the impacts of the three elements of vehicle, road and cargo on vehicle control, but also the established moving window can be updated in real time according to the dynamic changes of the vehicle state and the environment, and the data of the previous moment can be discarded in real time to reduce the model calculation amount and improve the model accuracy. Description of the Drawings

[0080] Figure 1 is the flowchart of the real-time modeling method of the present invention.

[0081] Figure 2 is the calculation flowchart of the real-time state update of the recursive moving window of the present invention. Detailed Embodiments

[0082] The present invention discloses a comprehensive physical big data modeling method based on invariant physical laws and time-varying information data, which is specifically used for heavy vehicle dynamics modeling and real-time state update. Aiming at the problems of poor parameter update ability of traditional vehicle dynamics models and the fact that big data modeling methods such as neural networks are divorced from the actual physical state, the mapping relationship between the system dynamics and the observable state space including vehicle configuration, road state and cargo specificity is deeply analyzed, a Kronecker product vehicle nonlinear state matrix including state input and scheduling parameters is constructed, and a Koopman kernel function based on linearized eigenvalue extraction is constructed to describe the physical state of the vehicle; a real-time update method for the vehicle state of a moving window based on data integration recursion is designed, the state data at the beginning of the window is discarded in real time under the selected window length and the latest data at time k + 1 is incorporated, and the real-time state of "vehicle-road-load" is iteratively calculated and solved based on the time series prediction theory. Finally, a physical-information comprehensive vehicle dynamics modeling method with both physical interpretability and adaptive update ability is established.

[0083] As Figure 1-2 shown, a real-time modeling method for the dynamics of a heavy vehicle includes the following steps:

[0084] S1: Establish a three-degree-of-freedom vehicle dynamics model and a tire model, and obtain the vehicle state and the chassis dynamics state;

[0085] S2: Establish a road surface state equation and obtain the road surface state;

[0086] S3: Judge the cargo state, combine the vehicle state, the chassis dynamics state and the road surface state, and establish an integrated vehicle-road-load nonlinear state equation including state inputs and scheduling parameters; obtain the mapping relationship between the vehicle configuration, the road surface state and the cargo state and the observable state space;

[0087] S4: According to the integrated vehicle-road-load nonlinear state equation, construct a Koopman kernel function based on linearized feature extraction;

[0088] S5: Based on the Koopman kernel function, in the data integration recursive moving window, discard the initial state of the window in real time and incorporate the data at the k+1 moment to obtain an updated integrated vehicle-road-load linear state equation.

[0089] Further, the three-degree-of-freedom vehicle dynamics model established in S1 is:

[0090] (1)

[0091] In the formula, m is the total vehicle mass, F f is the air resistance, F w is the rolling resistance, is the longitudinal acceleration, is the lateral acceleration, is the wheel angle, is the track width, j = l, r represent the front and rear wheels of the vehicle respectively, i = f, m, l represent the front axle, the middle axle and the rear axle of the vehicle respectively, L i is the distance from the front and rear axles of the vehicle to the center of mass, and are the yaw angular velocity and the yaw angular acceleration of the vehicle respectively, is the moment of inertia of the vehicle body about the Z1 axis, I x is the moment of inertia of the vehicle body about the X1 axis, and are respectively the longitudinal tire force and the lateral tire force of the wheel; when the wheel is in front of the vehicle center of gravity, " " is positive, and when it is behind the vehicle center of gravity, it is negative; when the wheel is on the left side of the vehicle, "

[0092] " is positive, and when it is on the right side of the vehicle, it is negative;

[0093] (2)

[0094] Where Y is the lateral force, X is the sideslip angle, D is the peak factor, B is the stiffness factor, C is the curve shape factor, E is the curve curvature factor, S h is the horizontal drift of the curve, S v is the vertical drift of the curve.

[0095] Furthermore, the road surface state equation established in S2 is:

[0096] (3)

[0097] where m is the total vehicle mass, is the sideslip angle, is the vehicle speed, is the vehicle acceleration, g is the gravitational acceleration, is the air resistance coefficient, is the contact area between the wheel and the ground, is the air density, is the engine torque, is the engine transmission ratio, is the transmission ratio of the gearbox, r is the wheel radius, is the transmission efficiency, f is the ground friction force.

[0098] Furthermore, "judging the cargo state" in S3 is to establish a cargo state equation to judge the rigid-flexible state of the cargo; specifically:

[0099] S31: When the vehicle is stationary, the vehicle vertical force balance equation is obtained from the moment balance, force balance and geometric relationship of the suspension spring as follows:

[0100] (4)

[0101] In the formula, is the total vehicle unloaded mass, , , are the spring stiffnesses of the front, middle and rear suspensions respectively, taking = = ; , , are the suspension forces of the front, middle and rear axles respectively, l f , l m , l r are the distances from the front axle to the center of mass, the distance from the middle axle to the center of mass, and the distance from the rear axle to the center of mass respectively;

[0102] S32: The calculation and analysis of the static vertical load of the wheel are as follows:

[0103] (5)

[0104] In the formula, and are the static vertical loads of the right and left front axles respectively, and are the static vertical loads of the right and left middle axles respectively, and are the static vertical loads of the right and left rear axles respectively;

[0105] S33: The calculation and analysis of the dynamic vertical load of the wheels are as follows:

[0106] (6)

[0107] In the formula is the vertical displacement of the center of gravity of each wheel, is the displacement of each suspension, is the vertical displacement of the vehicle body, is the height difference between the inner and outer sides of the road surface; is the sideslip angle of the vehicle, is the yaw angle of the vehicle; is the spring stiffness of each wheel, is the spring stiffness of each suspension, is the tire damping of each wheel, is the dynamic vertical load of each wheel; when the wheel is in front of the vehicle's center of gravity, " " takes a positive value and the cargo is flexible; when it is behind the vehicle's center of gravity, it takes a negative value and the cargo is rigid; when the wheel is on the left side of the vehicle, " " takes a positive value and the cargo is rigid; when it is on the right side of the vehicle, it takes a negative value and the cargo is flexible.

[0108] Furthermore, in S3, "Combining the vehicle state, chassis dynamics state, and road surface state, establish a vehicle-road-load integrated non-linear state equation including state inputs and scheduling parameters" is specifically

[0109] S34: Transform Equation (1) into an explicit equation about 、 、 and set up a general heavy vehicle model in the form of a non-linearized state space by combining the vehicle dynamics equation, road surface state equation, and cargo state equation:

[0110] (7)

[0111] In the formula, h s is the height of the vehicle's center of mass from the ground, ms is the full load mass, F x and F y are the lateral force and longitudinal force respectively, M z is the torque of the vehicle about the z-axis;

[0112] S35: After linearization, the vehicle-road load integrated vehicle state space form equation is obtained:

[0113] (8)

[0114] (9)

[0115] (10)

[0116] where F c is the control force, L c is the control gain, A is the system matrix, X b is the state vector, B is the input matrix, U is the input vector, C is the bias vector, A F is the additional system matrix, is the input vector increment; F is the final control input, is the gain coefficient; is the time derivative of the state vector, B F is the force input matrix;

[0117] The general non-linear state equation expression is as follows:

[0118] (11)

[0119] where is the time derivative of the state vector, is the non-linear function describing the system dynamics, is the predicted control force;

[0120] Define the state vector of the system , and obtain:

[0121] ,

[0122]

[0123] Furthermore, S4 "Construct the Koopman kernel function based on linearized feature extraction" is specifically as follows:

[0124] S41: Express equation (1) as a continuous-time linear parameter varying system, and the expression is as follows:

[0125] (12)

[0126] where , is a scheduling parameter; , , ;

[0127] S42: Discretize Equation (12) and represent it as:

[0128] (13)

[0129] In the equation

[0130] S43: Introduce the Kronecker product, and rewrite Equation (13) as:

[0131] (14)

[0132] where A jk (j = 0, 1, 2, 3) is the lumped state-related matrix, B jk (j = 0, 1, 2, 3) is the lumped input-related matrix, is the Kronecker product operator, , , are data trajectories;

[0133] S44: The equation at the previous moment of Equation (13) is expressed as:

[0134] (15)

[0135] Through the inverse operation of the matrix, solve for the system matrix :

[0136] (16)

[0137] S45: Obtain a Koopman kernel function based on the Koopman operator and with a moving window to recursively update the vehicle state data, and its formula is expressed as:

[0138] (17)

[0139] where is both the Koopman operator and the system transformation matrix, , , are the data trajectories of the input variable, state variable, and scheduling parameter variable respectively, , are the transfer variable matrix and update variable matrix respectively.

[0140] Furthermore, S5 includes:

[0141] S51: , , are the data trajectories of the input variable, the state variable, and the scheduling parameter variable respectively. The data is updated in real time using a moving window method. The window length is n d data points. Define a moving window and represent it as:

[0142] (18)

[0143] in is the data snapshot matrix, which contains combinations of inputs, states, and scheduling parameters and is the dataset of the moving window in real-time modeling, is the input matrix, which contains the system input u from to time, X k is the state matrix, which contains the system state x from to time, X Fk is the future state matrix, which contains the future system state from to time; and are the Krokhmal inner product matrices of the input state and the scheduling parameter ρ respectively, is the scheduling parameter vector, n d the number of data points in the moving window; is the Krokhmal inner product operator, which is used to generate combinations between inputs, states, and scheduling parameters;

[0144] S52: Rewrite the moving window based on equations (17) and (18) as:

[0145] (19)

[0146] S53: Using the vehicle state data collected by equation (18), obtain the system transformation matrix represented as:

[0147] (20)

[0148] S54: The recursive update formula is represented as:

[0149] (21)

[0150] S55: Assume the relationship between the system state x at time steps k and k + 1 is:

[0151] (22)

[0152] S56: Using the data points collected at time k + 1 , , Update the system transformation matrix , and realize the real-time update of the integrated vehicle-road-load modeling data of heavy vehicles. The updated integrated state space is expressed as:

[0153] (23)

[0154] In the formula is the state matrix at the updated time k, is the longitudinal speed at the updated time k, is the lateral speed at the updated time k, and are the yaw angular velocity and yaw angular acceleration of the vehicle at the updated time k, respectively.

[0155] The present invention breaks through the limitations of the traditional dynamic model of "special vehicle for special use and fixed scenario", and can not only comprehensively consider the three elements of vehicle, road and cargo, but also capture dynamic changes in real time to meet the modeling and control requirements of heavy vehicles in complex environments.

Claims

1. A real-time dynamic modeling method for heavy vehicles, characterized in that It includes the following steps: S1: Establish a three-degree-of-freedom vehicle dynamics model and a tire model, and obtain the vehicle state and chassis dynamics state; S2: Establish a road surface state equation and obtain the road surface state; S3: Judge the cargo state, and combine the vehicle state, chassis dynamics state and road surface state to establish an integrated vehicle-road-load nonlinear state equation including state inputs and scheduling parameters; obtain the mapping relationship between the vehicle configuration, road surface state and cargo state and the observable state space; S4: According to the integrated vehicle-road-load nonlinear state equation, construct a Koopman kernel function based on linearized feature extraction; S5: Based on the Koopman kernel function, in the data integration recursive moving window, discard the initial state of the window in real time and incorporate the data at the k+1 moment to obtain an updated integrated vehicle-road-load linear state equation.

2. The method according to claim 1, wherein The three-degree-of-freedom vehicle dynamics model established in S1 is: where m is the vehicle mass, F f is the air resistance, F w is the rolling resistance, is the longitudinal acceleration, is the lateral acceleration, δ i is the wheel angle, B i is the track width, j = l, r represent the left and right wheels of the vehicle respectively, i = f, m, l represent the front axle, middle axle, and rear axle of the vehicle respectively, L i is the distance from the front and rear axles of the vehicle to the center of mass, ω z and are the yaw angular velocity and yaw angular acceleration of the vehicle respectively, I z is the moment of inertia of the vehicle body about the Z1 axis, I x is the moment of inertia of the vehicle body about the X1 axis, F xij and F yij are the longitudinal tire force and lateral tire force of the ith wheel respectively; When the ij-th wheel is in front of the vehicle's center of gravity, "F" xij is positive, and negative when it is behind the vehicle's center of gravity; when the ij-th wheel is on the left side of the vehicle, "±F" yij is positive, and negative when it is on the right side of the vehicle. The tire model is: Where Y is the lateral force, X is the slip angle, D is the peak factor, B is the stiffness factor, C is the curve shape factor, E is the curve curvature factor, S h is the horizontal drift of the curve, S v is the vertical drift of the curve.

3. The method according to claim 2, wherein The road surface state equation established in S2 is: where m is the vehicle mass, α is the sideslip angle, u is the vehicle speed, g is the acceleration due to gravity, C d is the air resistance coefficient, A f is the contact area between the wheel and the ground, ρ a is the air density, T tq is the engine torque, i g is the engine transmission ratio, i0 is the transmission ratio of the gearbox, r is the wheel radius, η γ is the transmission efficiency, and f is the ground friction force.

4. The method according to claim 3, wherein In S3, "judging the cargo state" is to establish a cargo state equation to judge the rigid-flexible state of the cargo; specifically: S31: When the vehicle is stationary, the vehicle vertical force balance equation is obtained from the moment balance, force balance and geometric relationship of the suspension spring as follows: Where m b is the unladen mass of the whole vehicle, and K f , K m , and K r are the spring stiffnesses of the front, middle, and rear suspensions respectively. Let K f = K m = K r ; F zsf , F zsm , and F zsr are the suspension forces of the front, middle, and rear axles respectively. l f , l m , and l r are the distances from the front axle to the center of mass, from the middle axle to the center of mass, and from the rear axle to the center of mass respectively; S32: The calculation and analysis of the static vertical load of the wheel are as follows: where F zsfr1 and F zfgl1 are the static vertical loads of the right and left front axles respectively, F zsmr1 and F zsml1 are the static vertical loads of the right and left middle axles respectively, F zsrr1 and F zsrl1 are the static vertical loads of the right and left rear axles respectively; S33: The calculation and analysis of the dynamic vertical load of the wheel are as follows: where z wij is the vertical displacement of the center of gravity of each wheel, z sij is the displacement of each suspension, z b is the vertical displacement of the vehicle body, z road is the height difference between the inner and outer sides of the road surface; is the sideslip angle of the vehicle, is the yaw angle of the vehicle; K wij is the spring stiffness of each wheel, K sij is the spring stiffness of each suspension, C sij is the tire damping of each wheel, F zsji2 is the dynamic vertical load of each wheel; when the ij-th wheel is in front of the vehicle's center of gravity take positive, the cargo is flexible; when it is behind the vehicle's center of gravity, take negative, the cargo is rigid; when the ij-th wheel is on the left side of the vehicle take positive, the cargo is rigid; when it is on the right side of the vehicle, take negative, the cargo is flexible.

5. The method according to claim 4, characterized in that, In S3, "combining the vehicle state, chassis dynamics state and road surface state to establish an integrated vehicle-road-load nonlinear state equation including state inputs and scheduling parameters" is specifically S34: Transform equation (1) into a display equation with respect to v x 、v y 、ω z , and set up a general heavy vehicle model in a non-linearized state space form by combining the vehicle dynamics equation, the road surface state equation, and the cargo state equation: where h s is the height of the vehicle's center of mass from the ground, in m s is the full load mass, F x and F y are the lateral force and longitudinal force respectively, and M z is the torque of the vehicle about the z-axis; S35: After linearization, the integrated vehicle-road-load vehicle state space form equation is obtained: F c = L w (Ax b + BU + C + Bλ w ΔU) (8) F = L c λ c F c (9) where F c is the control force, L c is the control gain, A is the system matrix, x b is the state vector, B is the input matrix, U is the input vector, C is the bias vector, A F is the additional system matrix, ΔU is the input vector increment; F is the final control input, λ c is the gain coefficient; is the time derivative of the state vector, B F is the force input matrix; The expression of the general nonlinear state equation is as follows: where is the time derivative of the state vector, is the nonlinear function describing the system dynamics, and F CG is the predicted control force; Define the state vector of the system Obtain:

6. The method according to claim 5, wherein In S4, "constructing a Koopman kernel function based on linearized feature extraction" is specifically: S41: Express Equation (1) as a continuous-time linear parameter-varying system, and the expression is as follows: wherein A(ρ) = A0 + A1ρ1 + A2ρ2 + A3ρ3, B(ρ) = B0 + B1ρ1 + B2ρ2 + B3ρ3, ρ is the scheduling parameter; S42: Discretize Equation (12) and express it as: wherein S43: Introduce the Kroenecker inner product, and Equation (13) is written as: where A jk , j = 0, 1, 2, 3 are lumped state-related matrices, B jk , j = 0, 1, 2, 3 are lumped input-related matrices, is the Kronecker inner product operator, x k , u k , ρ k are data trajectories; The equation at the previous moment of Equation (13) is expressed as: Through the inverse operation of the matrix, solve the system matrix [BA]: S45: Obtain a Koopman kernel function based on the Koopman operator and with a moving window to recursively update the vehicle state data, and its formula is expressed as: T in k is the Koopman operator and also the system transformation matrix, x k , u k , ρ k are the data trajectories of the input variable, state variable, and scheduling parameter variable respectively. B k , A k are the transfer variable matrix and the update variable matrix respectively.

7. The method according to claim 6, wherein S5 It includes: S51: x k , u k , ρ k are the data traces of the input variable, the state variable, and the scheduling parameter variable respectively. The data is updated in real time using a moving window method, and the window length is n d data points. Define a moving window and represent it as: where D k is a data snapshot matrix, which contains combinations of inputs, states, and scheduling parameters, and is the data set of the moving window in real-time modeling. U k is the input matrix, which contains the system input u from time k - n d to k - 1. X k is the state matrix, which contains the system state x from time k - n d to k - 1. X Fk is the future state matrix, which contains the future system state from time k - n d+1 to k. U(ρ k ) and X(ρ k ) are the Klochner inner product matrices of the input state and the scheduling parameter ρ respectively. ρ k is the scheduling parameter vector, and n d is the number of data points in the moving window; is the Klochner inner product operator, which is used to generate combinations between the input and the state and the scheduling parameter; S52: Rewrite the moving window based on Equations (17) and (18) as: X Fk = T k D k (19) S53: Using the vehicle state data collected by Equation (18), obtain the system transformation matrix expressed as: The recursive update formula is expressed as: T k+1 = T k + ΔT k (21) Let the relationship between the system state x at time steps k and k+1 be: x k+1 = F(x k ) (22) S56: Update the system transformation matrix T using the data points x k+1 , u k+1 , ρ k+1 at time k + 1, to achieve real-time update of the integrated vehicle-road-load modeling data. The updated integrated state space is represented as: k ​ In the formula is the state matrix at the updated time k, v y,t is the longitudinal velocity at the updated time k, v x,t is the lateral velocity at the updated time k, ω z,k and are respectively the vehicle yaw angular velocity and yaw angular acceleration at the updated time k.

Images