An automated parameter calibration method for a handling stability controller of an electric vehicle

Abstract

The application relates to an automatic parameter calibration method of a handling stability controller for an electric automobile, which comprises the following steps: constructing a vehicle dynamics model, performing a discretization operation on the vehicle dynamics model by adopting an Euler discretization method to obtain a discrete nonlinear state equation; constructing a control target function; constructing a calibration evaluation function, obtaining a relaxation factor by subtracting an actual calibration evaluation index from an expected calibration evaluation index; designing an updating rule of a calibration parameter; calculating the relaxation factor and a calibration parameter control increment by adopting an unscented Kalman filtering to obtain the calibration parameter control increment; and substituting the calibration parameter control increment into the updating rule of the calibration parameter to obtain a control weight parameter at each moment, so that automatic calibration of the control weight parameter is completed. Compared with the prior art, the application has the advantages of not requiring a large amount of manpower and parameter calibration experience, greatly reducing artificial cost, improving parameter calibration efficiency and the like.

Description

Technical Field

[0001] This invention relates to the field of electric vehicle chassis control, and in particular to an automated parameter calibration method for a handling stability controller for electric vehicles. Background Technology

[0002] With the rapid development of electronic and electrical technologies, electric vehicles, due to their unique advantages such as energy saving and environmental protection, are gradually replacing gasoline vehicles and becoming the mainstream mode of transportation. Handling stability is a crucial performance characteristic related to the safe and comfortable driving of electric vehicles. Various types of controllers have been developed to improve the handling stability performance of distributed drive electric vehicles, such as Model Predictive Control (MPC) and Sliding Mode Control (SMC). Controller parameter calibration is a vital step in the design and development of control systems. By designing reasonable controller parameters, different control requirements can be met, and control performance can be improved. However, existing controller parameter calibration methods have the following problems:

[0003] 1. Existing controller calibration methods primarily rely on manual parameter calibration. This involves data collection, offline analysis, and parameter selection to create parameter tables for control parameters applicable to different operating conditions. Specific control parameters are then selected based on the specific operating conditions. However, offline parameter calibration is time-consuming and cannot account for changes in the controlled system over its lifespan. This makes it difficult to meet the constantly evolving driving conditions and control performance requirements. Therefore, an automated parameter calibration method for electric vehicle handling stability controllers that adapts to different driving conditions should be developed.

[0004] 2. Existing automated parameter calibration methods, such as Bayesian optimization and inverse reinforcement learning, can achieve automatic tuning of controller parameters. However, Bayesian optimization learns a black-box mapping from control parameters to specified control performance through trial and error search. This method is limited by the construction of a trial-and-error interactive environment, and the trial-and-error process is sensitive to external interference. Inverse reinforcement learning uses human demonstrations to learn the objective function, but it is limited by specific controller types and requires a high level of human knowledge base construction. Therefore, the above methods are not suitable for automated parameter calibration of electric vehicle handling stability controllers for diverse driving environments. To address this, this invention provides an automated parameter calibration method for handling stability controllers in electric vehicles. Summary of the Invention

[0005] The purpose of this invention is to overcome the shortcomings of the prior art and provide an automated parameter calibration method for a handling stability controller for electric vehicles.

[0006] The objective of this invention can be achieved through the following technical solutions:

[0007] In a first aspect, the present invention provides an automated parameter calibration method for a handling stability controller for electric vehicles, the method comprising the following steps:

[0008] Step S1: Construct a vehicle dynamics model based on the vehicle state and the control input of the electric vehicle handling stability controller. Use the Euler discretization method to discretize the vehicle dynamics model to obtain discrete nonlinear state equations.

[0009] Step S2: Based on the discrete nonlinear state equation, construct a control objective function, and obtain the optimal control input by solving the control objective function. The optimal control input is determined by the vehicle state and control weight parameters.

[0010] Step S3: Construct a calibration evaluation function, obtain the desired calibration evaluation index and the actual calibration evaluation index through the calibration evaluation function, and obtain the relaxation factor by subtracting the actual calibration evaluation index from the desired calibration evaluation index.

[0011] Step S4: Design the update rule for calibration parameters: Add the control weight parameter increment to the control weight parameter at the current time to obtain the control weight parameter at the next time.

[0012] Step S5: Model the relaxation factor and calibration parameter control increment as two independent random variables, both of which conform to a zero-mean Gaussian distribution. Calculate the relaxation factor and calibration parameter control increment using an unscented Kalman filter to obtain the calibration parameter control increment.

[0013] Step S6: Substitute the calibration parameter control increment into step S4 to obtain the control weight parameter at each time point, thus completing the automatic calibration of the control weight parameter.

[0014] Further, step S1 includes:

[0015] Let the vehicle state be X, and the control input of the electric vehicle handling stability controller be U. Then, the nonlinear vehicle dynamics model is constructed as follows:

[0016]

[0017] Let the current time be k, and the discrete time be T. s The nonlinear vehicle dynamics model is discretized using the Euler discretization method, resulting in the following discrete nonlinear state equations:

[0018] X(k+1)=f d (X(k),U(k)) (2)

[0019] Where X(k+1) is the vehicle state at time k+1, f d() represents the state equation function of the discrete nonlinear system, X(k) represents the vehicle state at time k, and U(k) represents the control input at time k.

[0020] Further, step S2 includes:

[0021] In vehicle handling stability control research, the construction of the control objective function typically considers tracking the desired state variables and suppressing the system's kinetic energy. By designing the control objective weight coefficients, the vehicle handling stability performance is optimized. Based on this, the control objective function is constructed as follows:

[0022]

[0023] Where N is the length of the future prediction time window, X(k+i) is the vehicle state at time k+i, and X ref For the desired state quantity, The weight parameters are used to track the desired state variables. U(k+i-1) is the weighting parameter for suppressing the system's kinetic energy, U(k+i-1) is the control input at time k, st is the constraint representation, and X is the weighting parameter for suppressing the system's kinetic energy. max ,X min These are the upper and lower boundary constraints for the state variables, U. max U min These are the upper and lower boundary constraints for the control variables.

[0024] By solving the above optimization problem, the optimal control input is obtained as follows:

[0025] U * (k)=f c (X(k),θ k (4)

[0026] Among them, U * (k) is the optimal control input, f c () represents the optimal control input function, X(k) represents the vehicle state at time k, and θ k The control weight parameters at time k are...

[0027] Further, step S3 includes:

[0028] Taking a historical time window length of n, the evaluation function is constructed based on historical time data as follows:

[0029] h k =h(X) k-n|k U k-n|k-1 (5)

[0030] In the above formula, X k-n|k U is the sequence of system state variables from time kn to time k within the historical time window.k-n|k-1 It is the sequence of system control variables from time kn to time k-1 within the historical time window.

[0031] Assuming the parameter θ is calibrated within a historical time window k If the initial state X(kn) remains unchanged, then the iterative evolution process of the state variable sequence and the control variable sequence can be expressed as follows:

[0032]

[0033] Analysis of the above formula shows that the calibration parameter θ k The only factor that determines the evaluation function h at the current time k is... k Therefore, the evaluation function can be simplified to:

[0034] h(θ k )=h0(X(kn),θ k (7)

[0035] The expected calibration evaluation index is: Then there is

[0036]

[0037] Among them, v k The relaxation factor represents the deviation between the expected and actual calibrated evaluation indicators. To determine the desired evaluation index, h(θ) k () is the international standard evaluation indicator.

[0038] Further, step S4 includes:

[0039] The update rules for the calibration parameters are as follows:

[0040] θ k+1 =θ k +Δθ k (9)

[0041] Where, θ k+1 Here, θ represents the control weight parameter at time k+1, Δθ is the calibration parameter control increment, and θ k The control weight parameters are at time k.

[0042] Further, step S5 includes:

[0043] The calibration parameter control increment Δθ and relaxation factor v are adjusted. k The model is based on two independent random variables, both of which follow a zero-mean Gaussian distribution, i.e.

[0044] Δθ k ~N(0,C θ ),v k~N(0,C v (10)

[0045] The parameter calibration process can then be constructed as a Kalman filter process, by finding the optimal calibration parameter θ. k This ensures that the deviation between the expected and actual calibration evaluation indicators is as small as possible.

[0046] This invention uses unscented Kalman filtering to automatically calibrate the control weight parameters. The process is as follows:

[0047] (1) Calculate 2n according to the theory of unscented transformation. θ +1 parameter sampling points, n θ Dimensions of parameters to be calibrated:

[0048]

[0049] In the above formula, λ is a scaling parameter used to reduce errors in prediction. Matrix A is calculated by applying the covariance matrix P... k|k Decomposition yields, i.e., P k|k =AA T .

[0050] (2) Set sampling weights:

[0051]

[0052] In the above formula, the subscripts m and c represent the mean weight and variance weight, respectively.

[0053] (3) Calculate 2n θ +1 parameter sampling points mean and sampling variance P k+1|k :

[0054]

[0055] (4) Based on each parameter sampling point Calculate its corresponding calibration evaluation function:

[0056]

[0057] (5) Calculate 2n θ +1 mean of the evaluation functions Covariance P hh :

[0058]

[0059] (6) Calculate the cross covariance P θh :

[0060]

[0061] (7) Calculate the Kalman gain K k :

[0062]

[0063] (8) Solve for the calibration parameter control increment:

[0064]

[0065] (9) Posterior covariance update:

[0066]

[0067] The automatic parameter calibration method based on unscented Kalman filtering, i.e., formula (18), yields the calibration parameter control increment Δθ. k .

[0068] In a second aspect, the present invention provides an in-vehicle controller, a processor, and a memory, the memory being used to store execution instructions of the processor, the execution instructions being used to perform the method described in any of the preceding claims.

[0069] Thirdly, the present invention provides a storage medium having a program stored thereon, wherein the program, when executed, implements the method described in any one of the above claims.

[0070] Compared with the prior art, the present invention has the following beneficial effects:

[0071] (1) Based on the control requirements of electric vehicle handling stability, this invention achieves automated calibration of control weight parameters by designing calibration evaluation functions, designing calibration parameter update criteria, and solving calibration parameter control increments based on unscented Kalman filtering. This method does not require a lot of manpower and parameter calibration experience, greatly reducing labor costs and improving parameter calibration efficiency.

[0072] (2) This invention first designs a calibration evaluation function to meet the handling stability requirements of electric vehicles, using a relaxation factor to characterize the deviation between the actual and desired calibration evaluation functions. Furthermore, the parameter update process is transformed into a process of solving for the calibration parameter control increment. By modeling the relaxation factor and the calibration parameter control increment as Gaussian distributions, the entire parameter calibration process is transformed into an unscented Kalman filter iterative process. Through unscented Kalman filter iteration, the calibration parameter control increment is obtained, thus completing the automated parameter calibration process. This invention eliminates the need for trial and error, achieving automated parameter calibration through data iteration, thus solving the problems of strong environmental dependence and low parameter space search efficiency associated with search-based automated parameter calibration methods. Attached Figure Description

[0073] Figure 1This is a framework diagram for the automated parameter calibration of the present invention for vehicle handling stability control;

[0074] Figure 2 This is a flowchart of the method steps of the present invention;

[0075] Figure 3 This is a schematic diagram of the two-degree-of-freedom vehicle model of the present invention;

[0076] Figure 4 This diagram illustrates the tracking errors of yaw rate and center of mass sideslip angle during the automated parameter calibration process of the automated parameter calibration method of the present invention. Figure 4 (a) shows the yaw rate tracking error curves under eight identical double-track change tests during the automated parameter calibration process. Figure 4 (b) shows the tracking error curves of the centroid side slip angle tracking under eight identical double-track shift tests during the automated parameter calibration process;

[0077] Figure 5 This is a schematic diagram of the calibration parameter curves obtained based on the automated parameter calibration method of the present invention, wherein, Figure 5 (a) is the calibration curve of the yaw rate tracking weight parameter. Figure 5 (b) is the calibration curve of the weight parameters for tracking the centroid sideslip angle;

[0078] Figure 6 The vehicle yaw rate tracking curve obtained by the vehicle handling stability controller in the simulation process of this invention using the automatic parameter calibration results described in this invention;

[0079] Figure 7 The vehicle yaw rate tracking curve obtained by the vehicle handling stability controller in the simulation process of this invention using the results of manual parameter calibration.

[0080] Figure 8 The vehicle center of gravity sideslip angle curve is obtained by applying two different parameter calibration results to the vehicle handling stability controller during the simulation process of this invention. Detailed Implementation

[0081] The present invention will now be described in detail with reference to the accompanying drawings and specific embodiments. These embodiments are based on the technical solution of the present invention and provide detailed implementation methods and specific operating procedures. However, the scope of protection of the present invention is not limited to the following embodiments.

[0082] Example

[0083] This embodiment provides an automated parameter calibration method for a handling stability controller in electric vehicles, such as... Figure 1The diagram shows the framework for automated parameter calibration for vehicle handling stability control. This framework comprises two parts. The first part is the design of the handling stability controller for the electric vehicle. This part first establishes a vehicle dynamics model, including a whole vehicle model and a tire model; then, it breaks down driving demand analysis into handling requirements, stability requirements, and low-energy consumption requirements, and designs the desired reference target accordingly; finally, it designs the handling stability controller based on the controlled object model and driving requirements, including objective function design, control weight selection, system constraint construction, and problem sequence solving. The control weights are obtained through an automated parameter calibration method. The second part is the automated parameter calibration method. This part first determines the control weights to be calibrated based on the handling stability controller; then, it designs a calibration evaluation function based on the handling stability driving requirements of the electric vehicle and constructs a parameter update criterion; finally, it obtains the control weight parameter update result based on an unscented Kalman filter process. Specifically, it constructs control weight calibration parameter sampling points based on unscented transformation, calculates the corresponding calibration evaluation function for each sampling point, and then calculates the mean, variance, and covariance of the sampling points and the evaluation function to obtain the control weight parameter update result.

[0084] In this embodiment, the electric vehicle selected is a distributed drive electric vehicle equipped with four-wheel hub motors. The driver controls the vehicle steering by turning the steering wheel, and the handling stability controller assists the driver in maneuvering the vehicle by sending four-wheel additional torque commands to the hub motors, thereby improving vehicle handling and driving stability.

[0085] This embodiment provides an automated parameter calibration method for a handling stability controller in electric vehicles, such as... Figure 2 As shown, the steps are as follows:

[0086] Step 1: Construct a vehicle dynamics model.

[0087] Specifically, (1) Construction of the vehicle dynamics model:

[0088] In this embodiment, a two-degree-of-freedom vehicle model is used to describe the vehicle's lateral and yaw motions, such as... Figure 3 As shown in the schematic diagram of this model, F yf ,F yr These represent the lateral forces of the front and rear axle tires, respectively; F xf ,F xr These represent the longitudinal forces of the front and rear axle tires, respectively; α f ,α r These are the front and rear axle tire slip angles, respectively; L f ,L r , respectively, are the distances from the vehicle's center of gravity to the front and rear axles; m is the vehicle's mass; I z It is the moment of inertia of the vehicle about the Z-axis; V xβ represents the longitudinal velocity; β represents the sideslip angle; γ represents the yaw rate; δ ... f Indicates the front wheel steering angle;

[0089] according to Figure 3 The two-degree-of-freedom model of the vehicle is established as follows:

[0090]

[0091] Where, β up γ is the upper limit of the centroid sideslip angle; up This represents the upper limit of the yaw rate; ΔM zup The purpose of introducing these quantities to add an upper limit value for the yaw moment is to normalize the state and control quantities, which are all considered to be known quantities in this invention.

[0092] (2) Tire Model Construction

[0093] The Fiala brush model is used to describe the lateral forces of the front and rear axles tires. This nonlinear model effectively improves the accuracy of tire force estimation compared to linear models. The tire slip angle is used as an internal variable in this model. When the tire slip angle α is very small, tanα≈α, and the tire model can then be approximated as:

[0094]

[0095] Where μ is the road surface adhesion coefficient; F z For vertical loads; C α The tire lateral stiffness is denoted as C. To distinguish between the front and rear wheels, the lateral stiffness of the front wheel is denoted as C. f The rear wheel lateral stiffness is denoted as C. r .

[0096] The tire slip angle is calculated using the following formula:

[0097]

[0098] The normalized vehicle yaw rate and sideslip angle are selected as the system state variables, i.e. The system control vector is chosen to be an additional yaw moment, i.e., u = ΔM z Considering that the tire model is a nonlinear model, the system state equation is also nonlinear, and can be expressed as:

[0099]

[0100] Let the current time be k, and the discrete time be T. s The nonlinear system equations are discretized using the Euler discretization method, resulting in the discrete nonlinear state equations:

[0101] x(k+1)=f d (x(k),u(k)) (24)

[0102] Where X(k+1) is the state variable at time k+1, f d () represents the state equation function of the discrete nonlinear system, x(k) represents the vehicle state at time k, and u(k) represents the control vector at time k.

[0103] Step 2: Construct the control objective function.

[0104] Specifically, the vehicle handling stability controller should improve the driver's handling and driving stability while minimizing energy consumption. Based on this, the control objective function is designed, considering the tracking of the desired yaw rate and sideslip angle, as well as the suppression of the system's kinetic energy:

[0105]

[0106] stu min ≤u(k+i-1)≤u max

[0107] In the above formula, N is the length of the future prediction time window; γ(k+i) is the vehicle yaw rate at time i, and γ ref The desired yaw rate is assumed to be known in this invention; β(k+i) is the sideslip angle of the vehicle's center of gravity at time i, and β... ref To determine the desired sideslip angle, based on vehicle stability requirements, the desired sideslip angle in this embodiment is set to β. ref =0; The weighted parameters are used to track the desired yaw rate; u(k+i-1) is the control input. Weighting parameters for tracking the desired centroid sideslip angle; Weighting parameters for suppressing actuation energy; u max ,u min These are the upper and lower boundary constraints for the control variables.

[0108] Since tracking the desired yaw rate and the desired sideslip angle are the primary objectives of electric vehicle handling stability control, this embodiment selects... and As an automatic calibration parameter, n θ =2, denoted as For the weighting parameters of suppressing actuation energy To improve the efficiency of automated parameter calibration, in this embodiment, it is treated as a fixed weight and does not participate in the automated parameter calibration process.

[0109] By solving the above optimization problem, the optimal additional yaw moment is obtained as follows:

[0110]

[0111] In the formula, To add the desired yaw moment, f c () represents the optimal control input function, X(k) represents the vehicle state at time k, and θ k The control weight parameters at time k are...

[0112] The optimal additional yaw moment is distributed to obtain the additional torque for all four wheels, which is then sent as control commands to the four-wheel hub motors of the electric vehicle to improve vehicle handling and driving stability. Using an average distribution method, the additional yaw moment is distributed to obtain the following additional torque for all four wheels:

[0113]

[0114] In the above formula, ΔT fl ΔT fr ΔT rl ΔT rr These represent the additional driving torque for each of the four wheels, where fl, fr, rl, and rr represent the left front wheel, right front wheel, left rear wheel, and right rear wheel, respectively; R e d is the tire rolling radius; d is the vehicle track width.

[0115] Step 3: Construct the calibration evaluation function.

[0116] Specifically, for the vehicle handling stability control problem, maintaining vehicle handling and driving stability is the primary goal of controller development and design. Therefore, the automated calibration results of the control weight parameters should first minimize the deviation between the actual yaw rate and the desired yaw rate to improve vehicle handling performance. Furthermore, the calibration results should minimize the vehicle's center-of-gravity sideslip angle to meet vehicle driving stability requirements. In this embodiment, the optimal additional yaw moment is obtained by solving an optimization problem; therefore, it is also necessary to ensure that the control weight parameters are non-negative to guarantee the non-convexity of the optimization problem. Therefore, the calibration evaluation function and desired calibration evaluation index are designed as follows:

[0117]

[0118] In the above formula, n is the length of the historical time window; (·) k-n|k This represents the sequence of state variables from time kn to time k within the historical time window; δ is the non-negative penalty factor for the weights, v k The relaxation factor represents the deviation between the expected calibration evaluation index and the actual calibration evaluation index.

[0119] The weighted non-negative penalty factor is expressed as follows:

[0120]

[0121] In the above formula, δ0 is a positive constant.

[0122] Step 4: Design the update rules for calibration parameters.

[0123] Specifically, the update criteria for the calibration parameters are designed as follows:

[0124] θ k+1 =θ k +Δθ k (29)

[0125] In the above formula, θ k The initial weight calibration parameters remain unchanged during the iterative calculation of state variables within the historical time window in step 3; Δθ is the calibration parameter control increment, which is obtained through the automated parameter calibration process based on unscented Kalman filtering in step 5; θ k+1 The updated weight calibration parameters will be fed into the handling stability controller for closed-loop control of vehicle handling stability.

[0126] Step 5: Calculate the relaxation factor and calibration parameter control increment using unscented Kalman filtering to obtain the calibration parameter control increment.

[0127] Specifically, the calibration parameter control increment Δθ and relaxation factor v are... k The model is based on two independent random variables, both of which follow a zero-mean Gaussian distribution, i.e.

[0128] Δθ k ~N(0,C θ ),ν k ~N(0,C v (30)

[0129] In the above formula, Δθ k The variance is taken as X θ =∈I2, where I2 is the second-order identity matrix, and ∈ is a balancing factor used to balance the convergence speed and robustness of parameter updates. When ∈ is large, the parameter update converges faster, but the robustness is poorer; C v The variance is taken as C v =I5, where I5 ​​is a fifth-order identity matrix.

[0130] Considering h(θ) k Since h(θ) is a nonlinear function, and for some control methods, such as optimization-based control, the optimization process is a black box process, it is impossible to obtain h(θ). k ) for θ kThe Jacobian differential matrix is ​​used. To avoid the above problems, this invention employs unscented Kalman filtering to achieve automated calibration of the control weight parameters. Unscented Kalman filtering uses unscented transformation to select several sampling points near the estimation point. By calculating the mean and variance of each sampling point, it achieves an infinite approximation of the system's nonlinearity. This method solves the information transmission problem of nonlinear systems without directly solving the Jacobian differential matrix.

[0131] The unscented Kalman filter process is constructed to automate the calibration of the control weight parameters. The process is as follows:

[0132] (1) Based on the theory of unscented transformation, calculate the sampling points of the 5 parameters. The calculation method is as follows:

[0133]

[0134] In the above formula, λ is a scaling parameter used to reduce prediction errors; here, λ = 1. Matrix A is calculated by applying the covariance matrix P... k|k Decomposition yields, i.e., P k|k =AA T .

[0135] (2) Set sampling weights:

[0136]

[0137] In the above formula, the subscripts m and c represent the mean weight and variance weight, respectively.

[0138] (3) Calculate the mean of the five parameter sampling points. and sampling variance P k+1|k :

[0139]

[0140] (4) Based on each parameter sampling point Calculate its corresponding calibration evaluation function:

[0141]

[0142] (5) Calculate the mean of the 5 evaluation functions. Covariance P hh :

[0143]

[0144] (6) Calculate the cross covariance:

[0145]

[0146] (7) Calculate the Kalman gain K k :

[0147]

[0148] (8) Solve for the calibration parameter control increment:

[0149]

[0150] (9) Posterior covariance update:

[0151]

[0152] The calibration parameter control increment θ is obtained through the above-described automated parameter calibration method based on unscented Kalman filtering. k .

[0153] Step 6: Substitute the calibration parameter control increment obtained from formula (37) into step 4 to obtain the updated control weight parameter θ. k+1 This enables the automatic calibration of the control weight parameters. The updated control weight parameters will then be fed into the handling stability controller for closed-loop control of vehicle handling stability.

[0154] To verify the effectiveness of the automated parameter calibration method for electric vehicle handling stability control described in this invention, a double lane change driving condition under low-friction road surface was selected for method verification. The comparison methods were the automated parameter calibration method designed in this invention and the manual calibration method. The specific process is as follows:

[0155] 1. Software Selection

[0156] The programming of the handling stability control algorithm and the automated parameter calibration algorithm proposed in this invention is implemented using the software Matlab / Simulink. The simulation model of the controlled object is implemented using the high-fidelity vehicle dynamics simulation software CarSim, with the software versions being Matlab R2021b and CarSim2019.1, respectively.

[0157] 2. Co-simulation settings

[0158] To achieve co-simulation between the two software programs, the input / output interface modules of CarSim were first configured, and the Simulink model path was added to CarSim to enable joint communication. Then, CarSim was compiled, and the corresponding S-Function module was generated in Simulink. Finally, the parameters of the S-Function were configured, and its input / output signal interfaces were brought out. The co-simulation step size was set to 0.001s. While the Simulink simulation model was running, the CarSim model was simultaneously performing calculations and solving. Data exchange between the two software programs continued throughout the simulation. If the model structure or parameter settings in CarSim were modified, recompilation was required, followed by regenerating the S-Function module to update the CarSim software configuration information.

[0159] 3. Simulation Condition Settings

[0160] To verify the effectiveness of the automated parameter calibration method for electric vehicle handling stability control described in this invention, a double lane change driving condition on a low-adhesion road surface was selected as the simulation verification condition. During the test, the road adhesion coefficient μ = 0.35, the vehicle speed was 65 km / h and remained constant during driving, and the vehicle steering wheel angle was implemented using the driver model built into Carsim. The vehicle model parameter used in the simulation test was the vehicle's moment of inertia I about the Z-axis. z =2059.2 kg·m 2 The total vehicle mass m = 1430 kg; the distance from the vehicle's center of gravity to the front axle L f = 1.05m, distance L from the rear axle r =1.61m; Front tire lateral stiffness C f = 43082 N / rad, rear tire lateral stiffness C r = 59950 N / rad, vehicle wheelbase d = 1.55 m, discrete time T s =0.01. For the control stability controller parameter configuration, the prediction time window length N = 5, and the upper and lower boundary constraints of the control quantity are U. max =5000Nm,U min = -5000Nm. For the parameter configuration of the automated parameter calibration method, the historical time window length n = 3, the weight penalty factor constant δ0 = 1, and the balance factor ∈ = 0.5.

[0161] 4. Comparison Method Settings

[0162] For both comparison methods, the initial weight value is set to 10, i.e. In the manual calibration method, the parameter values ​​used are the initial parameter values, which remain unchanged during the simulation. For the automated parameter calibration method, automated parameter calibration is performed based on the initial parameter values. The calibration process involves continuously implementing eight identical double-track test conditions until the control parameters converge. The obtained control parameter calibration results are then output to the handling stability controller, and the control effect of the controller after automated parameter calibration is obtained. By comparing the two control effects, the effectiveness of the method described in this invention is demonstrated. The simulation results of the two methods are shown in the attached figures.

[0163] Figure 4 To illustrate the yaw rate tracking error curve and the centroid sideslip angle tracking error curve during the automated parameter calibration process based on the automated parameter calibration method described in this invention, wherein... Figure 4 (a) The yaw rate tracking error curves under eight identical double lane change tests during the automated parameter calibration process. Analysis of the curves shows that the yaw rate tracking error decreases significantly with the increase in the number of test cycles, indicating that the vehicle's handling performance gradually improves as the yaw rate tracking weight parameters are updated. Furthermore, the tracking error curve exhibits periodic fluctuations because the expected yaw rate changes in real time according to the driver's intentions. During double lane change tests, the tire adhesion limit is easily reached at the entry and exit points, resulting in periodic abrupt changes in the tracking error. Figure 4 (b) The tracking error curves of the center of gravity sideslip angle under eight identical double lane change tests during the automated parameter calibration process. Analysis of the curves shows that the tracking error of the center of gravity sideslip angle decreases with the increase of the number of test cycles, indicating that the vehicle's driving stability gradually improves as the tracking weights of the center of gravity sideslip angle are updated. Furthermore, the reduction in tracking error is slow as the calibration process progresses. This is because the vehicle requires a certain lateral speed when turning, therefore the vehicle's center of gravity sideslip angle cannot be completely suppressed to zero.

[0164] Figure 5 The calibration parameter curve is obtained based on the automated parameter calibration method described in this invention, wherein... Figure 5 (a) is the calibration curve of the yaw rate tracking weight parameter. Figure 5 (b) shows the calibration curve of the weight parameters for tracking the center of gravity sideslip angle. Analysis of the curve shows that the two weight parameters gradually converge as the calibration process progresses. However, due to the continuous changes in the desired state and vehicle attitude during the double lane change test, the calibration curve of the control weight parameters exhibits some fluctuations. Based on the calibration curve of the control weight parameters, the desired yaw rate tracking weight parameter is ultimately selected as θ. γ =2800, the expected centroid sideslip angle tracking weight parameter is θ β =40. This is fed into the handling stability controller for vehicle handling and driving stability control.

[0165] Figure 6 The image shows the vehicle yaw rate tracking curve obtained by the vehicle handling stability controller during the simulation process using the automated parameter calibration results described in this invention. Analysis of the curve shows that the actual yaw rate of the vehicle can track the desired yaw rate with high accuracy.

[0166] Figure 7 The image shows the vehicle yaw rate tracking curve obtained by the vehicle handling stability controller during the simulation, based on the results of manual parameter calibration. Analysis of the curve reveals that the actual yaw rate fails to track the expected yaw rate, and the tracking error is significant. The large fluctuations in the actual yaw rate curve indicate that the vehicle is in an unstable state and has lost its handling capabilities.

[0167] contrast Figure 6 and Figure 7 It can be seen that the automated parameter calibration method described in this invention can effectively enhance the control capability of the handling stability controller on vehicle handling.

[0168] Figure 8 The figures show the vehicle sideslip angle curves obtained by the vehicle handling stability controller during simulation using two different parameter calibration results. Analysis of the curves shows that the handling stability controller obtained using the automated parameter calibration method described in this invention can effectively suppress the vehicle sideslip angle, allowing the vehicle to maintain stable driving on low-friction surfaces. In contrast, the results obtained using the manual calibration method cannot effectively suppress the vehicle sideslip angle; the large fluctuations in the sideslip angle curve indicate that the vehicle has lost its driving stability. Further improving the control effect of the handling stability controller would require more manpower and resources for manual calibration of the control weight parameters, which would increase the burden of algorithm development. In summary, the method proposed in this invention can effectively achieve all the benefits described in this invention.

[0169] This embodiment provides an in-vehicle controller, a processor, and a memory. The memory stores execution instructions of the processor, which execute the method provided in the above embodiment. The in-vehicle controller can be used in data transmission processes of various types of controllers within a vehicle. These controllers include, but are not limited to, vehicle controllers, gateway controllers, and instrument cluster controllers.

[0170] This embodiment also provides a storage medium on which a program is stored, which, when executed, implements the method provided in the above embodiment. The storage medium includes various media capable of storing program code, such as a USB flash drive, a portable hard drive, a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disk.

[0171] The preferred embodiments of the present invention have been described in detail above. It should be understood that those skilled in the art can make numerous modifications and variations based on the concept of the present invention without creative effort. Therefore, all technical solutions that can be obtained by those skilled in the art based on the concept of the present invention through logical analysis, reasoning, or limited experimentation on the basis of existing technology should be within the scope of protection defined by the claims.

Claims

1. An automated parameter calibration method for a handling stability controller for electric vehicles, characterized in that, Includes the following steps: Step S1: Construct a vehicle dynamics model based on the vehicle state and the control input of the electric vehicle handling stability controller. Use the Euler discretization method to discretize the vehicle dynamics model to obtain discrete nonlinear state equations. Step S2: Based on the discrete nonlinear state equation, construct a control objective function. By solving the control objective function, obtain the optimal control input, wherein the optimal control input is determined by the vehicle state and control weight parameters. The control objective function is expressed by the following formula: in, To predict the length of the time window for the future, for Vehicle status at any given time. For the desired state quantity, The weight parameters are used to track the desired state variables. These are the weighting parameters for suppressing the system's kinetic energy. For the control input at any given time, For constraint representation, These are the upper and lower constraint boundaries for the state variables, respectively. These are the upper and lower boundary constraints of the control quantity; Step S3: Construct a calibration evaluation function. Using this function, obtain the desired calibration evaluation index and the actual calibration evaluation index. Subtract the actual calibration evaluation index from the desired calibration evaluation index to obtain the relaxation factor. The calibration evaluation function is expressed by the following formula: in, Within a historical time window Time's up The sequence of system state variables at time t. Within a historical time window Time's up The sequence of system control variables at any given time; Step S4: Design the update rule for calibration parameters: Add the control weight parameter increment to the control weight parameter at the current time to obtain the control weight parameter at the next time. Step S5: Model the relaxation factor and calibration parameter control increment as two independent random variables, both of which conform to a zero-mean Gaussian distribution. Calculate the relaxation factor and calibration parameter control increment using an unscented Kalman filter to obtain the calibration parameter control increment. Step S6: Substitute the calibration parameter control increment into step S4 to obtain the control weight parameter at each time point, thus completing the automatic calibration of the control weight parameter.

2. The automated parameter calibration method for a handling stability controller for electric vehicles according to claim 1, characterized in that, The discrete nonlinear state equation is expressed by the following formula: in, for Vehicle status at any given time. For the state equation function of a discrete nonlinear system, for Vehicle status at any given time. for The amount of control input at any given time.

3. The automated parameter calibration method for a handling stability controller for electric vehicles according to claim 1, characterized in that, The optimal control input is expressed by the following formula: in, This is the optimal control input. This is the optimal control input function. for Vehicle status at any given time. for Control weight parameters at any time .

4. The automated parameter calibration method for a handling stability controller for electric vehicles according to claim 1, characterized in that, The relaxation factor is calculated using the following formula: in, As a relaxation factor, To define the evaluation indicators, Internationally defined evaluation indicators.

5. The automated parameter calibration method for a handling stability controller for electric vehicles according to claim 1, characterized in that, The update rule for the calibration parameters is calculated using the following formula: in, for Control weight parameters at any time To control the increment of calibration parameters, for Control weight parameters at any given time.

6. The automated parameter calibration method for a handling stability controller for electric vehicles according to claim 5, characterized in that, The calibration parameter control increment is calculated using the following formula: in, To control the increment of calibration parameters, For Kalman gain, To define the evaluation indicators, The mean of the evaluation function is used to calibrate it.

7. A vehicle-mounted controller, characterized in that, A processor and a memory, the memory being used to store execution instructions of the processor, the execution instructions being used to perform the method of any one of claims 1-6.

8. A storage medium having a program stored thereon, characterized in that, When the program is executed, it implements the method as described in any one of claims 1-6.

Images