A vehicle cornering control method based on lateral acceleration prediction

Abstract

A vehicle cornering control method based on lateral acceleration prediction solves the problem of cornering stability failure of autonomous driving vehicles under extreme working conditions. The present invention relates to the field of autonomous driving. The present invention includes a reference path module, a lateral acceleration prediction module, a desired braking torque design module, a model prediction control module and a vehicle module. Among them, the reference path module inputs the desired path into the model prediction control module; the lateral acceleration prediction module predicts the lateral acceleration based on the vehicle state information; the desired braking torque design module generates the desired braking torque based on the lateral acceleration; the model prediction control module combines the desired path, vehicle state information and desired braking torque, optimizes and solves the control information and sends it to the vehicle module to realize path tracking control, and at the same time feeds back the vehicle state information to the model prediction control module and the lateral acceleration prediction module to improve the path tracking performance of the vehicle under extreme working conditions.

Description

Technical field:

[0001] The present invention belongs to the field of autonomous driving, and in particular to a vehicle cornering control method based on lateral acceleration prediction. Background technology:

[0002] With the rapid development of autonomous driving technology, vehicle safety issues under extreme conditions have become increasingly prominent. Take high-speed cornering as an example. The sharp change in road curvature causes a rapid increase in lateral acceleration, which in turn affects the distribution of friction between the tire and the road. When the lateral force of the tire exceeds the adhesion provided by the road surface, it can easily cause the tire to slip and even cause the entire vehicle to become unstable. Therefore, how to regulate the vehicle's braking torque and steering angle during cornering to ensure driving stability has become one of the core issues in current autonomous driving control systems. In response to the problem of vehicle cornering control, researchers have proposed a variety of control strategies to optimize vehicle stability and handling performance. Each method has its advantages in specific application scenarios. However, under extreme conditions, existing methods generally have problems such as response lag and insufficient robustness, making it difficult to achieve comprehensive and effective control.

[0003] Among the many control methods, G-Vectoring Control (GVC) is regarded as a representative cornering control strategy. This method determines whether the vehicle is in the cornering stage by identifying the changes in lateral acceleration during vehicle operation in real time, and achieves a "braking" effect based on the active adjustment of the engine torque, so that the center of mass of the vehicle is moderately offset to increase the load on the front wheels, thereby enhancing the vehicle's steering response and cornering stability. On this basis, researchers have tried to introduce advanced control strategies to expand and optimize GVC. For example, patent CN116788242A combines Model Predictive Control (MPC) with GVC, and adopts a four-wheel equal lateral tire lateral force margin distribution strategy for torque control, thereby enhancing the stability of MPC in the coordinated control of path tracking and braking torque. However, patent CN116788242A fails to fully utilize historical lateral acceleration data and its changing trends. Furthermore, the four-wheel equal tire lateral force margin distribution strategy adjusts the braking torque by distributing the force from the center of mass to the front and rear wheelbases. This results in the front tire lateral force margin being greater than the rear tire lateral force margin. This leads to insufficient rear tire lateral force margin during cornering, which can cause vehicle instability. To further improve vehicle lateral stability, patent CN115257704A derives a lateral acceleration expression by constructing a two-degree-of-freedom vehicle model. It then introduces a linear tire model to construct state-space equations and applies this to an MPC framework to predict lateral acceleration. While patent CN115257704A can predict the vehicle's lateral dynamic response during driving to a certain extent, providing feedforward information for the control system. However, when the tires enter the nonlinear region, the linear tire model struggles to accurately describe the nonlinear characteristics of the tire forces, resulting in significant deviations between the predicted lateral acceleration and the actual value, thus affecting the vehicle's lateral stability control.

[0004] In summary, if historical lateral acceleration data can be fully integrated and long-term predictions of lateral acceleration can be achieved based on this, the ability to predict when a vehicle is about to become unstable can be improved. Furthermore, combining these predictions with coordinated steering and braking control can effectively improve the stability and safety of autonomous vehicles during cornering. Summary of the invention:

[0005] To address the shortcomings of existing technologies, this paper proposes a vehicle cornering control method based on lateral acceleration prediction. This method integrates path tracking, lateral acceleration prediction, and optimal braking torque distribution. It also employs an MPC algorithm to perform multi-objective collaborative optimization of front wheel angle and braking torque. This method achieves coordinated control of lateral stability and longitudinal braking force regulation, enhancing the vehicle's cornering control performance.

[0006] To achieve the above object, the present invention adopts the following technical solutions:

[0007] The present invention provides a vehicle cornering control method based on lateral acceleration prediction, which includes a reference path module, a lateral acceleration prediction module, a desired braking torque design module, a model prediction control module and a vehicle module.

[0008] The method comprises the following steps:

[0009] Step 1. Reference Path Module:

[0010] The reference path module provides the ideal driving trajectory of the vehicle in the future. This path provides the expected path P for the model predictive control algorithm. t ={x ref ,y ref}, guiding the vehicle to reasonably adjust its driving state in a dynamic environment to ensure that the vehicle drives safely and smoothly along the expected trajectory, where x ref with y ref are the horizontal and vertical coordinates of the expected path respectively.

[0011] Step 2, Vehicle Module:

[0012] The vehicle module executes the control information Q output by the model predictive control module t , realize path tracking control, and transmit vehicle status information C t Feedback to the model prediction control module and the lateral acceleration prediction module, where the vehicle state information C t Includes: C t ={v x ,v y ,F x.ij ,F y.ij ,F z.ij ,a x ,γ,ψ,y,x,κ}, where y is the lateral position of the vehicle, x is the longitudinal position of the vehicle, κ is the road curvature, v x ,v y are the longitudinal speed and lateral speed of the vehicle, F x.ij ,F y.ij ,F z.ij They are the longitudinal tire force, lateral tire force, and vertical load of the wheel, respectively, where i = f, r represents the front wheel and the rear wheel, respectively, j = ρ, σ represents the left wheel and the right wheel, respectively, and a x is the longitudinal acceleration of the vehicle, γ is the yaw rate of the vehicle, and ψ is the heading angle of the vehicle.

[0013] Step 3: Lateral acceleration prediction module:

[0014] The lateral acceleration prediction module is used to predict future lateral acceleration based on vehicle status information during cornering, improving the forward-looking nature of the control system. This module includes: a data preprocessing module, a model training module, and a prediction module. The specific steps are as follows:

[0015] Step 3.1, data preprocessing module:

[0016] Based on vehicle status information, samples are screened using a similarity measurement method to construct a dataset. The dataset is then normalized. On this basis, sliding window technology is applied to extract time-series input-output pairs to capture time-dependent features. The extracted features and labels are organized into tensor data for use in model training.

[0017] Step 3.2, model training module:

[0018] The LSTM network receives tensor data, generates a predicted value of lateral acceleration, and calculates the loss function based on the actual value to update the network parameters. The LSTM network model is generated and passed to the prediction module to perform the lateral acceleration prediction task.

[0019] Step 3.3, prediction module:

[0020] Receive the LSTM network model generated by the model training module, screen and retain the network model with the smallest loss value as the optimal model, combine it with the tensor data generated by the data preprocessing module to perform the lateral acceleration prediction task, and output the final lateral acceleration a when entering the corner y .

[0021] Step 4: Expected braking torque design module:

[0022] This module receives the real-time prediction results provided by the lateral acceleration prediction module, combines them with the vehicle status information, generates the expected braking torque that meets the path tracking and stability requirements, and provides it to the subsequent model predictive controller to realize the distribution of wheel braking torque.

[0023] Step 5: Model predictive control module:

[0024] The model predictive control module is used to track the reference path and distribute the desired braking torque among the wheels to ensure wheel stability. This module includes: vehicle model construction, local linearization of tire lateral force, tire lateral force margin distribution module construction, prediction model construction, and inverse tire model construction. The specific steps are as follows:

[0025] Step 5.1: Vehicle model construction:

[0026] In order to describe the yaw rate and lateral motion response of the vehicle under the action of braking force, the following vehicle dynamics model is established, as follows:

[0027]

[0028] in, is the derivative of the yaw rate, is the derivative of the lateral velocity, γ is the yaw rate, v y is the lateral velocity, l f is the distance from the center of mass to the front axle, l r is the distance from the center of mass to the rear axle, w is the wheelbase of the vehicle, m is the total mass of the vehicle, I z is the moment of inertia of the vehicle, F y.ij is the tire lateral force, T b,ij is the tire braking force, where i = f, r represents the front wheel and the rear wheel respectively, and j = ρ, σ represents the left wheel and the right wheel respectively.

[0029] The tracking error model is established as follows:

[0030]

[0031] in, is the derivative of the heading deviation, is the lateral deviation derivative, γ is the yaw angular velocity, Δψ is the heading angular rate error, e is the lateral error, v x is the vehicle longitudinal velocity, and κ is the road curvature.

[0032] Step 5.2: Local linearization of tire lateral force:

[0033] The function of the rear wheel lateral force with respect to the rear wheel slip angle is locally linearized in the prediction domain by using a table lookup method, as follows:

[0034]

[0035] in, is the tire slip angle at time k, Side slip angle The local state stiffness, F y,ij is the tire lateral force, α ij is the tire slip angle, Side slip angle The local state lateral force, j = ρ, σ, which represent the left and right wheels respectively.

[0036] Step 5.3: Construction of tire lateral force margin distribution module:

[0037] Constrain the vehicle tire forces as follows:

[0038]

[0039] Among them, Fy,ij is the tire lateral force, F x,ij is the tire longitudinal force, F z,ij is the vertical load, and μ is the road adhesion coefficient.

[0040] On this basis, in order to ensure sufficient tire force margin for all four wheels, the following tire force margin distribution strategy is adopted, as follows:

[0041]

[0042] in, is the tire lateral force margin, F z,ij is the vertical load, and μ is the road adhesion coefficient.

[0043] Step 5.4: Prediction model construction:

[0044] Substituting formula (3) in step 5.2 into formulas (1) and (2) in step 5.1, the integrated MPC controller model is obtained as follows:

[0045]

[0046] in, is the derivative of the yaw rate, is the derivative of the lateral velocity, γ is the yaw rate, v y is the lateral velocity, l f is the distance from the center of mass to the front axle, l r is the distance from the center of mass to the rear axle, w is the wheelbase of the vehicle, m is the total mass of the vehicle, I z is the moment of inertia of the vehicle, F y.ij is the tire lateral force, T b,ij is the tire braking force, κ is the road curvature, i = f, r represent the front and rear wheels respectively, j = ρ, σ represent the left and right wheels respectively.

[0047] Formula (6) is organized into a standard state space equation as follows:

[0048]

[0049] Where ξ=[γv y Δψe d ] T is the state variable,

[0050] Z=[F y,fρ F y,fσ T b,fρ T b,fσ T b,rρ T b,rσ ] T is the control variable, ζ is the control output, d is the disturbance input, Aξ 、B z 、B d and C ξ is the coefficient matrix.

[0051] Where,

[0052] Discretize formula (7) to obtain the incremental discrete system model, which is as follows:

[0053]

[0054] Among them, Δξ is the vehicle state increment, ΔZ is the control input increment, and Δd is the disturbance input increment.

[0055] In order to introduce the strategy of maximizing tire lateral force margin utilization into the control system for optimization, it is necessary to write it into the discrete system model formula (8) and use the Jacobian matrix linearization formula (5), as follows:

[0056]

[0057] in, It means that the formula (5) For the state variable ξ=[γv y Δψe d ] T Find each term of It means that the formula (5) For the control variable Z=[F y,fρ F y,fσ T b,fρ T b,fσ T b,rρ T b,rσ ] T Take the partial derivative of each term.

[0058] By combining formula (8) and formula (9), a controller system model based on tire lateral force margin is constructed as follows:

[0059]

[0060] In the formula O 3×6 Represents a matrix with a value of 0 and a dimension of 3 rows and 6 columns.

[0061] The predicted output can be expressed as follows:

[0062] Γ(h+1|h)=S ξ ·Δξ(h)+S Z ·ΔZ(h)+Sd Δd(h)+E·Γ(h) (11)

[0063] Among them, S Z =S BZ +S DZ , S BZ 、S DZ Represents the control input increment matrix, whose dimension is a matrix with P rows and M columns.

[0064]

[0065] Among them, P is the prediction time domain and M is the control time domain.

[0066] The predicted output sequence, control input sequence, and reference output sequence are defined at h as follows:

[0067]

[0068]

[0069]

[0070] Based on the discretized prediction model, the cost function of the controller is designed as follows:

[0071]

[0072] Where Q is the optimization objective function, Θ Γ is the weight of the predicted output, Θ z is the weight of the predicted input, Γ(h+1) is the predicted value, R(h+1) is the actual value, ΔZ(h) is the control increment, Aeq=[001111], beq is the torque generated by the desired braking torque design module, z min 、z max is the constraint on the control input, Δz min , Δz max is the constraint on the rate of change of the control input.

[0073] Finally, a constrained optimization problem is constructed to solve the optimal control quantity at the current moment, as follows:

[0074]

[0075] Convert the optimization problem into a quadratic programming problem and solve it:

[0076]

[0077] Among them, H,q,C, The definition is as follows:

[0078] E P (h+1)=R(h+1)-S ξ Δx(h)-Eζ(h), L z =diag([E nz ,…E nz ] M×M ),

[0079]

[0080] where Z(h-1) is the matrix of measurements of the control inputs at time h-1.

[0081] Step 5.5: Inverse tire model construction:

[0082] The inverse tire model is used to convert the MPC-optimized front wheel lateral force into the front wheel steering angle δ as follows:

[0083]

[0084] Among them, δ fj is the turning angle of the front wheel, β is the sideslip angle of the center of mass, l f is the distance from the front axle to the center of mass, γ is the yaw rate, v x is the longitudinal velocity of the vehicle, F y,fj is the lateral force on the front wheel, is used to transfer the front wheel lateral force F y,fj Converted into tire slip angle α fj , where j = ρ and σ represent the left and right wheels respectively.

[0085] Finally, the front wheel angle δ fj and braking torque T b,ij As the control information Q t The data is sent to the vehicle module to control the vehicle turning.

[0086] The present invention provides a vehicle cornering control method based on lateral acceleration prediction. By integrating a lateral acceleration prediction mechanism with a model predictive control strategy, this method leverages historical lateral acceleration data and, based on this, enables long-term prediction of future lateral acceleration. This allows the control system to proactively identify impending vehicle instability and, in conjunction with this prediction, proactively adjust the steering and braking control systems. This method not only enhances the control system's foresight but also improves the vehicle's stability and robustness under extreme operating conditions, demonstrating promising application prospects in the field of autonomous driving. Description of the drawings:

[0087] Figure 1It is a schematic diagram of a vehicle cornering control method based on lateral acceleration prediction.

[0088] Figure 2 It is a schematic diagram of the lateral acceleration prediction module. Specific implementation method:

[0089] The present invention will be described in detail below with reference to the accompanying drawings.

[0090] like Figure 1 As shown, the present invention is a vehicle cornering control method based on lateral acceleration prediction. With lateral acceleration prediction as the core, it combines path tracking and braking torque distribution to improve the stability of the vehicle during cornering. It includes five modules: a reference path module, a lateral acceleration prediction module, a desired braking torque design module, a model prediction control module and a vehicle module. The reference path module provides the desired path as the path tracking target; the lateral acceleration prediction module predicts the lateral acceleration of the corner based on the vehicle state information, and perceives the risk of instability in advance; the desired braking torque design module generates the braking torque according to the predicted lateral acceleration of the corner; the model prediction control module combines the desired path, vehicle state information and desired braking torque, outputs the optimized control information and sends it to the vehicle module to realize path tracking control, and transmits the vehicle state information to the model prediction control and lateral acceleration prediction modules to realize closed-loop control.

[0091] Step 1. Reference Path Module:

[0092] The reference path module uses a quintic polynomial fitting method to generate a smooth and controllable desired path P based on the position information of the starting and ending points and their velocity and acceleration boundary conditions. t ={x ref ,y ref}, ensuring that the trajectory is physically feasible and continuous. The fifth-order polynomial expression is as follows:

[0093] y(x)=a0+a1x+a2x 2 +a3x 3 +a4x 4 +a5x 5 (1)

[0094] Where x represents the longitudinal coordinate and y represents the corresponding lateral coordinate. By giving the boundary conditions of the starting and ending points, including position, velocity, and acceleration, the six unknown coefficients a0 to a5 can be solved to generate a desired trajectory that satisfies the constraints.

[0095] Where a0~a5 can be expressed as:

[0096]

[0097] Among them, a0 is the acceleration of the starting point, a1 is the acceleration of the end point, v0 is the speed of the starting point, v1 is the acceleration of the end point, h = q1-q0 is the lateral position change between the starting point and the end point, and T = t1-t0 is the time change between the starting point and the end point.

[0098] Step 2, Vehicle Module:

[0099] In order to verify the effectiveness and engineering feasibility of the proposed control strategy, the present invention introduces a vehicle dynamics model for system modeling and control verification. In the specific implementation method, the CarSim 2020 vehicle model is used as the simulation platform, and the vehicle dynamics system is constructed in combination with typical working conditions. Through the joint simulation of CarSim and Simulink, the effect evaluation of the path tracking and tire lateral force margin distribution strategy is realized, which serves as the verification basis before the actual vehicle experiment. The following is a specific description of the method of the present invention using a certain model of CarSim automobile simulation software as the platform. Its main parameters are shown in Table 1:

[0100] parameter unit Parameter value Car mass m kg 1414 <![CDATA[Distance l from the vehicle's center of mass to the front axle f > m 1.015 <![CDATA[Distance l from the vehicle's center of mass to the rear axle r > m 1.895 <![CDATA[Yaw moment of inertia I about the vertical axis passing through the center of mass of the vehicle z > <![CDATA[kgm 2 ]]> 1536.7 Car wheelbase w m 1.675 Wheel effective radius r m 0.31

[0101] The vehicle module executes the control information Q output by the model predictive control module t , realize path tracking control, and transmit vehicle status information C t Feedback to the model prediction control module and the lateral acceleration prediction module, where the vehicle state information C t Includes: C t ={v x ,v y ,F x.ij ,F y.ij ,F z.ij ,a x ,γ,ψ,y,x,κ}, where y is the lateral position of the vehicle, x is the longitudinal position of the vehicle, κ is the curvature of the path the vehicle has passed, and v x ,v y are the longitudinal speed and lateral speed of the vehicle, F x.ij ,F y.ij ,F z.ij They are the longitudinal tire force, lateral tire force, and vertical load of the wheel, respectively, where i = f, r represents the front wheel and the rear wheel, j = ρ, σ represents the left wheel and the right wheel, respectively, and a x is the longitudinal acceleration of the vehicle, γ is the yaw rate of the vehicle, and ψ is the heading angle of the vehicle.

[0102] Step 3: Lateral acceleration prediction module:

[0103] like Figure 2 As shown, the lateral acceleration prediction module is based on a deep neural network (such as LSTM network) and combines the vehicle state information C t, to predict lateral acceleration, specifically including: data preprocessing module, model training module, and prediction module.

[0104] Step 3.1, data preprocessing module:

[0105] Step 3.1.1: First, the system receives the vehicle status information C t , we filter samples by similarity measurement method and construct a data set. Among them, we choose Euclidean distance as the matching degree measurement method between vehicle status information and matching information in the information database. The details are as follows:

[0106]

[0107] Among them, the non-negative real number d(x,y) represents the Euclidean distance between two vectors, x i Represents n-dimensional vehicle status information, y i Indicates matching information in the information database.

[0108] Step 3.1.2: Use the Min-Max Normalization method to normalize the dataset. The purpose is to scale the data to the range of [0, 1] to avoid the influence of different features on the model training due to the difference in value range. The details are as follows:

[0109]

[0110] Step 3.1.3: After normalizing the dataset, use the sliding window technique to extract continuous input-output pairs to capture temporal features, as follows:

[0111] Take a time series data X with a length of N = [x1, x2, ..., x N ], set a fixed window size w, and define the sliding step s.

[0112] For each window, the input data consists of the historical values ​​within the window, and the output data is the data at the end of the window. In the time series, the input data in the tth window can be expressed as:

[0113] X input (t) = [x t ,x t+1 ,…x t+w-1 ] (5)

[0114] The output data corresponding to this window is:

[0115] X output (t) = x t+w (6)

[0116] Step 3.1.4: Adjust the dimensions of the extracted input and output pairs according to the requirements of the LSTM network structure and convert them into three-dimensional tensor data with dimensions (N, C, T), where N represents the batch size and C = 8 represents the feature dimension (corresponding to the lateral displacement y, lateral velocity v y , yaw rate γ, front wheel angle δ, road curvature κ, heading angle ψ, longitudinal acceleration a x and its lateral acceleration a y ), T=4 represents the length of the input sequence (i.e., the model predicts the deceleration at the next moment based on data from four consecutive historical moments).

[0117] Step 3.1.5. Divide the dataset into training set, validation set and test set. Divide the dataset according to the set ratio (80% training set, 10% validation set, 10% test set) and prepare for subsequent training and evaluation.

[0118] Step 3.2, model training module:

[0119] Step 3.2.1. Initialize the LSTM network model to recognize the input tensor data, where the LSTM network contains a 2-layer LSTM structure, with 128 hidden units in each layer.

[0120] Step 3.2.2, the LSTM network model calculates the prediction results through forward propagation, updates the state at each time step through the gating mechanism, and uses past historical information to predict the lateral acceleration.

[0121] Among them, the key gating mechanism of LSTM takes the input of time step t including the hidden state h of the previous time step t and the input x at the current time step t , these inputs are processed by different gating mechanisms. The key gating mechanisms of LSTM are as follows:

[0122] Forget Gate f t : Control the retention ratio of historical information. The specific formula is as follows:

[0123] f t =σ(W f ·[h t-1 ,x t ]+b f ) (7)

[0124] Among them, f t is the output of the forget gate, W f is the weight matrix of the forget gate, x t is the input of the current time step, h t-1 is the hidden state of the previous moment, b f is the bias term; σ is the Sigmoid activation function.

[0125] Input gate i t : Controls how the current input affects the memory unit. The specific formula is as follows:

[0126] i t =σ(W i ·[h t-1 ,x t ]+b i ) (8)

[0127] Among them, i t is the output of the input gate, W t is the weight matrix of the input gate, b t is the bias term.

[0128] Candidate memory units Generate candidate new memory content. The specific formula is as follows:

[0129]

[0130] Among them, W c is the weight matrix of the candidate memory unit, b c is the bias term, and tanh is the hyperbolic tangent activation function.

[0131] Memory unit update C t :Combining the output of the forget gate and the input gate, update the memory unit. The formula is as follows:

[0132]

[0133] Among them, f t Controls the history information retained, i t Control the addition of new information.

[0134] Output gate o t : Controls the output of the hidden state. The specific formula is as follows:

[0135] o t =σ(W o ·[h t-1 ,x t ]+b o ) (11)

[0136] o t is the output of the output gate, W o is the weight matrix of the output gate, b o is the bias term.

[0137] Step 3.2.3. Based on the difference between the predicted value generated by the model and the true value (label), the loss function is calculated using the mean square error (MSE). The specific formula is as follows:

[0138]

[0139] Where n is the total number of samples, y i is the true value of the i-th sample, is the predicted value of the i-th sample.

[0140] Step 3.2.4: Through the back-propagation algorithm, the loss value is propagated backward from the output layer to the input layer, and the gradient is calculated for each parameter (i.e., the weight and bias in the LSTM unit) based on the loss value.

[0141] Step 3.2.5. Using the calculated gradient, use the AdamW optimizer (with an initial learning rate of 0.01 and a weight decay coefficient of 0.001) to adjust the weights and biases in the LSTM network. The goal is to reduce the value of the loss function by adjusting the weights, thereby improving the accuracy of the model prediction.

[0142] Step 3.2.6. During the training process, the weights are continuously adjusted iteratively until the model converges, that is, the loss reaches the minimum value. After the training is completed, all the weights and biases obtained from the training are integrated to form the final LSTM network model. The network model is sent to the prediction module for subsequent prediction tasks.

[0143] Step 3.3, prediction module:

[0144] Step 3.3.1: Receive the trained LSTM network model from the model training module. The model includes the weights, biases, and corresponding loss function values ​​learned during the training process.

[0145] Step 3.3.2: Compare the validation set loss function values ​​in each round of training and select the model with the lowest loss. This means that the model is able to optimally predict the model on the validation set, thus avoiding overfitting. An "early stopping" strategy is usually adopted. That is, if the validation set loss does not significantly decrease over multiple consecutive training rounds, training is stopped and the model with the lowest loss is retained.

[0146] Step 3.3.3: After completing the training and verification of the model, the LSTM network model makes predictions based on the tensor data extracted by the data preprocessing module, and finally outputs the lateral acceleration a when entering the curve. y .

[0147] Step 4: Expected braking torque design module:

[0148] For the design of braking torque, this paper adopts the G-Vectoring Control (GVC) method. This method intervenes in the vehicle's lateral dynamic response by adjusting the longitudinal braking torque, optimizing the vehicle's posture control in corners. GVC can effectively suppress drastic changes in lateral acceleration, thereby improving the vehicle's stability and responsiveness in corners.

[0149] Step 4.1, GVC control module:

[0150] In the GVC control module design, the lateral acceleration change rate is first used as the core adjustment basis. After first-order filtering and smoothing, it is mapped to the longitudinal braking acceleration input, thereby achieving active adjustment of the vehicle posture. The details are as follows:

[0151]

[0152] Among them, G xc Longitudinal braking acceleration, G y is the lateral acceleration, is the rate of change of lateral acceleration, C xy is the control gain, T s is the first-order filter time constant.

[0153] Step 4.2, braking torque mapping:

[0154] To properly convert the longitudinal braking acceleration generated by the GVC module into wheel braking torque, this paper further designed a tire dynamics mapping model. This model comprehensively considers the wheel's moment of inertia, angular velocity change rate, and longitudinal acceleration to ensure the feasibility and dynamic response characteristics of the braking torque output. The details are as follows:

[0155]

[0156] Where J is the wheel moment of inertia, is the derivative of the wheel angular velocity, R is the wheel radius, m t is the mass of a single wheel, is the longitudinal acceleration of the wheel, T b is the output braking torque.

[0157] By combining the above-mentioned GVC control law with the braking torque mapping model, longitudinal braking control intervention of the lateral dynamic trend is achieved, providing the expected braking torque for model predictive control.

[0158] Step 5: Model predictive control module:

[0159] The model predictive control module is used to track the reference path and distribute the desired braking torque among the wheels to ensure wheel stability. This module includes: vehicle model construction, local linearization of tire lateral force, tire lateral force margin distribution module construction, prediction model construction, and inverse tire model construction. The specific steps are as follows:

[0160] Step 5.1: Vehicle model construction:

[0161] In order to describe the yaw rate and lateral motion response of the vehicle under the action of braking force, the following vehicle dynamics model is established, as follows:

[0162]

[0163] in, is the derivative of the yaw rate, is the derivative of the lateral velocity, γ is the yaw rate, v y is the lateral velocity, l f is the distance from the center of mass to the front axle, l r is the distance from the center of mass to the rear axle, w is the wheelbase of the vehicle, m is the total mass of the vehicle, I z is the vehicle's moment of inertia, F y.ij is the tire lateral force, T b,ij is the tire braking force, where i = f, r represents the front wheel and the rear wheel respectively, and j = ρ, σ represents the left wheel and the right wheel respectively.

[0164] The tracking error model is established as follows:

[0165]

[0166] in, is the derivative of the heading deviation, is the lateral deviation derivative, γ is the yaw angular velocity, Δψ is the heading angular rate error, e is the lateral error, v x is the vehicle longitudinal velocity, and κ is the road curvature.

[0167] Step 5.2: Local linearization of tire lateral force:

[0168] The rear wheel lateral force is locally linearized to the rear wheel slip angle in the prediction time domain by using the table lookup method, as follows:

[0169]

[0170] in, is the tire slip angle at time k, Side slip angle The local state stiffness, F y,ij is the tire lateral force, α ijTire slip angle, Side slip angle The local state lateral force.

[0171] Step 5.3: Construction of tire lateral force margin distribution module:

[0172] Constrain the vehicle tire forces as follows:

[0173]

[0174] Among them, F y,ij is the tire lateral force, F x,ij is the tire longitudinal force, F z,ij is the vertical load.

[0175] On this basis, in order to ensure sufficient tire lateral force margin on all four wheels, the following tire lateral force margin distribution strategy is adopted, as follows:

[0176]

[0177] in, is the tire lateral force margin, F z,ij is the vertical load, and μ is the road adhesion coefficient.

[0178] Step 5.4: Prediction model construction:

[0179] Substituting formula (17) in step 5.2 into formulas (15) and (16) in step 5.1, the integrated MPC controller model is obtained as follows:

[0180]

[0181] in, is the derivative of the yaw rate, is the derivative of the lateral velocity, γ is the yaw rate, v y is the lateral velocity, l f is the distance from the center of mass to the front axle, l r is the distance from the center of mass to the rear axle, w is the wheelbase of the vehicle, m is the total mass of the vehicle, I z is the moment of inertia of the vehicle, F y.ij is the tire lateral force, T b,ij is the tire braking force, κ is the road curvature, i = f, r represent the front and rear wheels respectively, j = ρ, σ represent the left and right wheels respectively.

[0182] Formula (20) is organized into a standard state-space equation as follows:

[0183]

[0184] Where ξ=[γv y Δψe d ] T is the state variable, Z=[F y,fρ F y,fσ T b,fρ T b,fσ T b,rρ T b,rσ ] T is the control variable, ζ is the control output, d is the disturbance input, A ξ 、B z 、B d and C ξ is the coefficient matrix.

[0185] Where,

[0186] For formula (21), by T s The system is discretized in seconds (according to the vehicle dynamic response requirements) to obtain the incremental discrete system model, as follows:

[0187]

[0188] Among them, Δξ is the vehicle state increment, ΔZ is the control input increment, and Δd is the disturbance input increment.

[0189] In order to introduce the strategy of maximizing the utilization rate of the tire lateral force margin into the control system for optimization, it is necessary to write it into the system model (22) and use the Jacobian matrix linearization formula (19), as follows:

[0190]

[0191] in, include Indicates the yaw rate γ based on Find the partial derivative, Indicates the lateral velocity v y right Find the partial derivative, Indicates the heading deviation Δψ Find the partial derivative, Indicates the lateral deviation e Find the partial derivative, include Indicates the lateral force F based on the left front wheel y,fρ right Find the partial derivative, Indicates the lateral force F based on the right front wheel y,fσ right Find the partial derivative, Indicates the longitudinal force F based on the left front wheel x,fρ right Find the partial derivative, Indicates the longitudinal force F based on the right front wheel x,fr right Find the partial derivative, Indicates the longitudinal force F based on the left rear wheel x,rρ right Find the partial derivative, Indicates the longitudinal force F based on the right rear wheel x,rσ right Find the partial derivative.

[0192] The controller system model based on the tire lateral force margin is constructed by combining formula (23) and formula (22), as follows:

[0193]

[0194] In the formula O 3×6 Represents a matrix with a value of 0 and a dimension of 3 rows and 6 columns.

[0195] The predicted output can be expressed as follows:

[0196] Γ(h+1|h)=S ξ ·Δξ(h)+S Z ·ΔZ(h)+S d Δd(h)+E·Γ(h) (25)

[0197] Among them, S Z =S BZ +S DZ , S BZ 、S DZ Represents the control input increment matrix, whose dimension is a matrix with P rows and M columns.

[0198]

[0199] Among them, P is the prediction time domain and M is the control time domain.

[0200] The predicted output sequence, control input sequence, and reference output sequence are defined at h as follows:

[0201]

[0202]

[0203]

[0204] Based on the discretized prediction model, a constrained optimization problem is constructed to solve the optimal control quantity at the current moment, as follows:

[0205]

[0206] Where Q is the optimization objective function, Θ Γ is the weight of the predicted output, Θ z is the weight of the prediction input, Γ(h+1) is the predicted value, R(h+1) is the actual value, ΔZ(h) is the control increment, Aeq=[0 0 1 1 1 1], beq is the torque generated by the desired braking torque design module, z min 、z max is the constraint on the control input, Δz min , Δz max is the constraint on the rate of change of the control input.

[0207] The MPC optimization problem can be formulated as follows:

[0208]

[0209] The optimization problem of formula (30) is transformed into a quadratic programming problem and solved:

[0210]

[0211] Among them, H,q,C, The definition is as follows:

[0212] E P (h+1)=R(h+1)-S ξ Δx(h)-Eζ(h), L z =diag([E nz ,…E nz ] M×M ),

[0213]

[0214] where Z(h-1) is the matrix of measurements of the control inputs at time h-1.

[0215] Step 5.5: Inverse tire model construction:

[0216] The inverse tire model is used to convert the MPC-optimized front wheel lateral force into the front wheel steering angle δ as follows:

[0217]

[0218] Among them, δ fj is the turning angle of the front wheel, β is the sideslip angle of the center of mass, l f is the distance from the front axle to the center of mass, γ is the vehicle's angular velocity, v xis the longitudinal velocity of the vehicle, F y,fj is the lateral force on the front wheel, For numerical calculations using the lookup table method, when a specific F is given y,fj The corresponding sideslip angle α can be quickly calculated fj , where j = ρ and σ represent the left and right wheels respectively.

[0219] Finally, the front wheel angle δ fj and braking torque T b,ij As the control information Q t The data is sent to the vehicle module to control the vehicle turning.

[0220] In summary, the present invention proposes a vehicle cornering control method based on lateral acceleration prediction. By collecting the real-time status of the vehicle and predicting future lateral acceleration, the braking torque can be adjusted in advance. Compared with the traditional GVC model, this method utilizes historical lateral acceleration data and, based on this, realizes long-term prediction of future lateral acceleration. It can sense potential lateral instability trends under cornering conditions to improve path tracking accuracy and stability. At the same time, the improved tire lateral force margin allocation strategy is combined with model predictive control to achieve optimal distribution of braking torque, enhance the stability and robustness of the vehicle under extreme conditions, and has good application prospects in the field of autonomous driving.

Claims

1. A vehicle cornering control method based on lateral acceleration prediction, characterized by: The method includes the following modules: a reference path module, a lateral acceleration prediction module, a desired braking torque design module, a model predictive control module, and a vehicle module; wherein the reference path module is used to provide the desired path to the model predictive control module; the lateral acceleration prediction module predicts the lateral acceleration of the vehicle during the cornering process based on vehicle state information; the desired braking torque design module generates the desired braking torque based on the predicted lateral acceleration; the model predictive control module combines the desired path, vehicle state information, and desired braking torque to optimize and solve control information and send it to the vehicle module to implement path tracking control, and feeds back the vehicle state information to the model predictive control module and the lateral acceleration prediction module; The method comprises the following steps: Step 1.1: The reference path module provides the vehicle's expected path P to the model predictive control module. t ={x ref ,y ref }, where x ref with y ref are the horizontal and vertical coordinates of the expected path respectively; Step 1.2: The vehicle module executes the control information Q output by the model predictive control module t , realize path tracking control, and transmit vehicle status information C t Feedback to the model prediction control module and the lateral acceleration prediction module; the vehicle state information C t Includes: C t ={v x ,v y ,F x.ij ,F y.ij ,F z.ij ,a x ,γ,ψ,y,x,κ}, where y is the lateral position of the vehicle, x is the longitudinal position of the vehicle, κ is the road curvature, and v x ,v y are the longitudinal speed and lateral speed of the vehicle, F x.ij ,F y.ij ,F z.ij They are the longitudinal tire force, lateral tire force and vertical load of the wheel, respectively, where i = f, r represents the front wheel and the rear wheel respectively, j = ρ, σ represents the left wheel and the right wheel respectively, and a x is the longitudinal acceleration of the vehicle, γ is the yaw rate of the vehicle, and ψ is the heading angle of the vehicle; Step 1.3, the lateral acceleration prediction module, based on the vehicle status information C fed back by the vehicle module t , predict the lateral acceleration a of the vehicle when entering a curve y ; Step 1.4: Design module for expected braking torque based on the predicted lateral acceleration a y , calculate the expected braking torque T z ; Step 1.5: Model predictive control module receives the expected path P t , expected braking torque T z and vehicle status information C t , optimize and solve the control information Q t And sent to the vehicle module to realize path tracking control, where the control information Q t Includes the braking torque of the four wheels and the front wheel angle.

2. The vehicle cornering control method based on lateral acceleration prediction according to claim 1, characterized in that: The lateral acceleration prediction module is used to predict the lateral acceleration a of the vehicle during the cornering process in real time. y , specifically including: data preprocessing module, model training module and prediction module. The specific steps are as follows: Step 2.1, data preprocessing module: A dataset is constructed based on the vehicle status information output by the vehicle module and normalized. Based on this, a sliding window technique is used to extract input and output pairs to capture the temporal features in the dataset. The extracted temporal features and labels are then converted into tensor data. Step 2.2, model training module: The long short-term memory (LSTM) network receives the tensor data output by the data preprocessing module to generate a lateral acceleration prediction value. It then combines this with the actual lateral acceleration value to calculate the loss function, update the network parameters, and generate the LSTM network model. Step 2.3, prediction module: Receive the LSTM network model generated by the model training module, screen and retain the network model with the smallest loss value as the optimal model, combine it with the tensor data generated by the data preprocessing module to predict the lateral acceleration, and output the lateral acceleration a during the cornering process y .

3. The vehicle cornering control method based on lateral acceleration prediction according to claim 1, characterized in that: The model predictive control module distributes the desired braking torque and performs path tracking control, including vehicle model construction, local linearization of tire lateral force, tire lateral force margin distribution module construction, prediction model construction, and inverse tire model construction. The specific steps are as follows: Step 3.1: Vehicle model construction: In order to describe the yaw rate and lateral motion response of the vehicle under the action of braking force, the following vehicle dynamics model is established, as follows: in, is the derivative of the yaw rate, is the derivative of the lateral velocity, γ is the yaw rate, v y is the lateral velocity, l f is the distance from the center of mass to the front axle, l r is the distance from the center of mass to the rear axle, w is the wheelbase of the vehicle, m is the total mass of the vehicle, I z is the vehicle's moment of inertia, F y.ij is the tire lateral force, T b,ij is the tire braking force, where i = f, r represents the front wheel and the rear wheel respectively, and j = ρ, σ represents the left wheel and the right wheel respectively; The tracking error model is established as follows: in, is the derivative of the heading deviation, is the lateral deviation derivative, γ is the yaw angular velocity, Δψ is the heading angular rate error, e is the lateral error, v x is the vehicle longitudinal velocity, κ is the road curvature; Step 3.2: Local linearization of tire lateral force: The rear wheel lateral force is locally linearized to the rear wheel slip angle in the prediction domain of the model predictive control by using a table lookup method, as follows: in, is the tire slip angle at time k, Side slip angle The local state stiffness, F y,ij is the tire lateral force, α ij Tire slip angle, Side slip angle The local state lateral force, j = ρ, σ, which represent the left and right wheels respectively; Step 3.3: Construction of tire lateral force margin distribution module: Constrain the vehicle tire forces as follows: Among them, F y,ij is the tire lateral force, F x,ij is the tire longitudinal force, F z,ij is the vertical load, where i = f, r represents the front wheel and the rear wheel respectively, j = ρ, σ represents the left wheel and the right wheel respectively; On this basis, in order to ensure that the wheels have sufficient tire lateral force margin, the following tire lateral force margin distribution strategy is adopted, as follows: in, is the lateral force margin of the left front wheel tire, F z,ij is the vertical load, μ is the road adhesion coefficient; Step 3.4: Prediction model construction: Substituting formula (3) in step 3.2 into formulas (1) and (2) in step 5.1, we can obtain the MPC controller model as follows: in, is the derivative of the yaw rate, is the derivative of the lateral velocity, γ is the yaw rate, v y is the lateral velocity, l f is the distance from the center of mass to the front axle, l r is the distance from the center of mass to the rear axle, w is the wheelbase of the vehicle, m is the total mass of the vehicle, I z is the vehicle's moment of inertia, F y.ij is the tire lateral force, T b,ij is the tire braking force, κ is the road curvature, i = f, r represents the front wheel and rear wheel respectively, j = ρ, σ represents the left wheel and right wheel respectively; Formula (6) is organized into a standard state space equation as follows: Where ξ=[γv y Δψe d ] T is the state variable, Z=[F y,fρ F y,fσ T b,fρ T b,fσ T b,rρ T b,rσ ] T is the control variable, ζ is the control output, d is the disturbance input, A ξ 、B z 、B d and C ξ is the coefficient matrix; Where, Discretize formula (7) to obtain the incremental discrete system model, which is as follows: Among them, Δξ is the vehicle state increment, ΔZ is the control input increment, and Δd is the disturbance input increment; In order to introduce the strategy of maximizing the utilization rate of the tire lateral force margin into the control system for optimization, the Jacobian matrix is ​​used to linearize formula (5), as follows: in, It means that the formula (5) For the state variable ξ=[γ v y Δψ e d ] T Find the partial derivative of each term It means that the formula (5) For the control variable Z=[F y,fρ F y,fσ T b,fρ T b,fσ T b,rρ T b,rσ ] T Take partial derivative of each term; By combining formula (8) and formula (9), a controller system model based on tire lateral force margin is constructed as follows: Where, O 3×6 Represents a matrix with a value of 0 and a dimension of 3 rows and 6 columns; The predicted output can be expressed as follows: C(h+1|h)=S ξ ·Dx(h)+S Z ·ΔZ(h)+S d Δd(h)+E·Γ(h) (11) Among them, S Z =S BZ +S DZ , S BZ 、S DZ Represents the control input increment matrix, whose dimension is a matrix with P rows and M columns; Where, Among them, P is the prediction time domain and M is the control time domain; The predicted output sequence, control input sequence, and reference output sequence are defined at h as follows: Based on the discretized prediction model, the cost function of the controller is designed as follows: Where Q is the optimization objective function, Θ Γ is the weight of the predicted output, Θ z is the weight of the prediction input, Γ(h+1) is the predicted value, R(h+1) is the actual value, ΔZ(h) is the control increment, Aeq=[0 0 1 1 1 1], beq is the torque generated by the desired braking torque design module, z min 、z max is the constraint on the control input, Δz min , Δz max The constraint on the rate of change of the control input; Finally, a constrained optimization problem is constructed to solve the optimal control quantity at the current moment, as follows: Furthermore, the optimization problem of formula (16) is transformed into a quadratic programming problem for solution, which is as follows: Among them, H,q,C, The definition is as follows: E P (h+1)=R(h+1)-S ξ Δx(h)-Eζ(h), where Z(h-1) is the matrix of measurements of the control input at time h-1; Step 3.5: Inverse tire model construction: The inverse tire model is used to convert the MPC-optimized front wheel lateral force into the front wheel steering angle δ as follows: Among them, δ fj is the turning angle of the front wheel, β is the sideslip angle of the center of mass, l f is the distance from the front axle to the center of mass, γ is the vehicle yaw rate, v x is the longitudinal velocity of the vehicle, F y,fj is the lateral force on the front wheel, is used to transfer the front wheel lateral force F y,fj Converted into tire slip angle α fj , where j = ρ and σ represent the left and right wheels respectively; Finally, the front wheel angle δ fj and braking torque T b,ij As the control information Q t The data is sent to the vehicle module to control the vehicle turning.

Images