CN-121978928-A - Intelligent automobile track tracking control method based on self-adaptive switching MPC

CN121978928ACN 121978928 ACN121978928 ACN 121978928ACN-121978928-A

Abstract

An intelligent automobile track tracking control method based on self-adaptive switching MPC (MPC) belongs to the technical field of intelligent automobile control and comprises the steps of building a vehicle mechanism dynamics model, building a vehicle dynamics model, designing an MPC controller based on three models, taking road curvature, road adhesion coefficient and longitudinal speed as fuzzy input, taking transverse error as variable domain scaling factors, and realizing self-adaptive switching of the three models through variable domain fuzzy control. According to the invention, modeling precision under complex working conditions is improved through the NARX neural network, tracking precision and calculation instantaneity are balanced through the self-adaptive switching mechanism, simulation verification shows that the minimum of a transverse error peak value can reach 0.102m, the control time is shortened to 4.56s, and the requirements of an automatic driving system on high precision and high instantaneity of track tracking are met.

Inventors

HE YILIN
ZHANG JIWEI
ZHAO XUAN
MA JIAN
YU MAN
YUAN XIAOLEI
WANG SHU

Assignees

长安大学

Dates

Publication Date: 20260505
Application Date: 20260116

Claims (10)

1. An intelligent automobile track tracking control method based on self-adaptive switching MPC is characterized by comprising the following steps: s1, establishing a vehicle mechanism dynamics model, wherein the vehicle mechanism dynamics model comprises a nonlinear model of three-degree-of-freedom vehicle dynamics and a linear model thereof; S2, establishing an NARX neural network with delay input, and training to obtain a vehicle dynamics model; S3, designing an MPC track tracking control algorithm, wherein the MPC track tracking control algorithm comprises an LMPC prediction controller, an NMPC prediction controller and an NARX-MPC prediction controller which correspond to the linear model, the nonlinear model and the vehicle dynamics model obtained in the S1 and the S2 respectively; S4, designing a self-adaptive switching MPC controller, taking road curvature rho, road adhesion coefficient mu and longitudinal vehicle speed vx as fuzzy inputs, taking transverse error e as variable domain scaling factor alpha, adopting variable domain fuzzy control, combining the scaling factor alpha to adjust an output domain, outputting a selection instruction of a linear model, a nonlinear model or a vehicle dynamics model, dynamically switching and activating the LMPC prediction controller, the NMPC prediction controller or the NARX-MPC prediction controller in the S3, and executing track tracking control at the current moment.
2. The intelligent vehicle track following control method based on adaptive switching MPC of claim 1, wherein the differential equation of the nonlinear model of three-degree-of-freedom vehicle dynamics in S1 is: Wherein m is the vehicle mass, v x and v y are the longitudinal speed and lateral speed of the mass center under the vehicle body coordinate system respectively, And The change rates of longitudinal and lateral speeds, namely longitudinal and lateral accelerations of the vehicle in the body coordinate system, are respectively expressed, X and Y are the position coordinates of the vehicle centroid in the global coordinate system, And Representing velocity components of the vehicle in global X and Y directions, respectively; For the yaw angle of the vehicle, The change rate of the yaw angle, that is, the yaw rate of the vehicle, Representing a change rate of yaw rate, that is, yaw acceleration of the vehicle; And The linear cornering stiffness of the front tire and the rear tire respectively, delta f is the cornering angle of the front wheel, l f and l r are the distances from the center of mass of the vehicle to the front axle and the rear axle respectively, C lf and C lr are the longitudinal cornering stiffness of the front tire and the rear tire of the vehicle respectively, s f and s r are the slip rates of the front tire and the rear tire respectively, and I z is the rotational inertia of the vehicle around the z axle.
3. The intelligent automobile track tracking control method based on adaptive switching MPC as claimed in claim 2, wherein after the nonlinear model of three-degree-of-freedom vehicle dynamics is linearized by S1, the forward Euler discretization is adopted, and a differential equation of the linear model is obtained as follows: The equation of state: the output equation: the state equation is that the state of the next moment (k+1) can be predicted according to the state of the current moment k and the control input, and the physical meaning is that the output quantity eta (k) required for tracking or feedback is selected from all the states χ (k) through a matrix C; Wherein, the , ; Where χ (k) represents the system state vector at discrete time step k, η (k) represents the system output vector at time step k, A k,t represents the state transition matrix at time step k, B k,t represents the control input matrix at time step k, C k,t represents the output matrix at time step k, T represents the sampling time, I n represents the n-dimensional identity matrix, and T represents the continuous time variable.
4. The intelligent automobile track following control method based on adaptive switching MPC as claimed in claim 1, wherein the NARX neural network in S2 comprises an input layer, a hidden layer and an output layer which are sequentially connected in series; the input features of the input layer comprise vehicle dynamics state and control information, and each input feature comprises current measured value and history information of previous time step, and the vehicle dynamics state comprises lateral speed Longitudinal speed And yaw rate The control information is a front wheel rotation angle delta; the hidden layer comprises 20 hidden units, and the activation function is tanh; the output characteristics of the output layer include lateral acceleration Longitudinal acceleration And yaw acceleration 。
5. The intelligent vehicle track following control method based on adaptive switching MPC as claimed in claim 4, wherein said NARX neural network adopts a Levenberg-Marquardt algorithm and trains with mean square error as a loss function to obtain a vehicle dynamics model.
6. The intelligent automobile track following control method based on adaptive switching MPC as claimed in claim 1, wherein the objective functions of LMPC, NMPC and NARX-MPC track following control algorithm in S3 are: Wherein, the 、 And The tracking error term, the course angle error and the term control quantity penalty term are respectively adopted; The method comprises the steps of Q Y 、Q φ and R are respectively weight coefficients of transverse errors, heading deviation and control quantity, N p 、N c is respectively a prediction time domain and a control time domain, J represents a cost function CostFunction, a controller calculates an optimal control sequence by minimizing J, the smaller value represents better tracking effect and smoother control, Y k+i represents an actual output vector at the moment k+i, a position coordinate Y of a vehicle in a two-dimensional plane, Y refk+i represents a reference output vector at the moment k+i, namely a target position at the corresponding moment on a desired path, phi represents an actual heading angle of the vehicle, namely an included angle between a vehicle body direction and an x axis of a global coordinate system, phi refk+i represents a reference heading angle of the vehicle at the moment k+i, namely a tangential direction of the desired path at the moment k+i, u k+i represents a control input vector at the moment k+i, front wheel steering angle, Q Y represents a weight matrix of the output errors, Q φ represents a weight matrix of the heading angle errors, R represents a weight matrix of control input, np represents a prediction time domain length, and Nc represents a control time domain length.
7. The intelligent vehicle track following control method based on adaptive switching MPC as claimed in claim 6, wherein the constraints of the MPC track following control algorithm include control quantity and output quantity constraints, and the following formula is: Wherein, the And (3) with The minimum and maximum control amounts, and y min and y max are the minimum and maximum system output amounts, respectively.
8. The intelligent automobile track tracking control method based on self-adaptive switching MPC as claimed in claim 1, wherein each parameter of the fuzzy input in S4 is divided into S, M AND L fuzzy sets by adopting a trapezoidal membership function, the fuzzy control of the variable domain specifically adopts a Mamdani fuzzy reasoning system with an AND rule, AND the reasoning rule is based on the mapping relation between the input parameter combination AND an optimal model; The expression of the scaling factor alpha is: Where e is the lateral error.
9. A computer readable storage medium, characterized in that the computer readable storage medium stores a computer program, which when executed by a processor, implements the intelligent vehicle trajectory tracking control method based on adaptive switching MPC according to any one of claims 1-8.
10. A computer program product comprising computer programs/instructions which, when executed by a processor, implement the intelligent vehicle trajectory tracking control method based on adaptively switching MPC according to any one of claims 1-9.

Description

Intelligent automobile track tracking control method based on self-adaptive switching MPC Technical Field The invention belongs to the technical field of intelligent automobile control, and particularly relates to an intelligent automobile track tracking control method based on self-adaptive switching MPC. Background With the development of automatic driving technology, track tracking is used as a core technology for realizing accurate path following of vehicles, and the performance of the track tracking is directly related to driving safety and riding comfort. Model Predictive Control (MPC) is widely used in the field of trajectory tracking because of its ability to explicitly handle multiple constraint and optimization problems. The control effect of an MPC is highly dependent on the accuracy of its internal predictive model. The traditional vehicle mechanism model (such as a linear two-degree-of-freedom model and a nonlinear three-degree-of-freedom model) has simple structure and high calculation efficiency, but under complex nonlinear working conditions such as large slip angle, low adhesion and the like, the model mismatch is serious, so that the control precision is reduced. The neural network model based on data driving can approach to complex vehicle dynamics characteristics and improve prediction accuracy, but the network structure is relatively complex, the online calculation time is long, and the high real-time requirement of an automatic driving system is difficult to meet. Therefore, how to ensure the track tracking precision and simultaneously consider the calculation efficiency of the control algorithm becomes a key difficult problem in the current intelligent automobile track tracking control field. In the prior art, the control requirement under the full working condition is difficult to meet by simply using a high-precision complex model or a low-precision simple model. Disclosure of Invention Aiming at the problems, the invention aims to provide an intelligent automobile track tracking control method based on self-adaptive switching MPC, aiming at intelligently selecting an optimal vehicle dynamics model for MPC prediction according to real-time driving conditions, so as to obtain optimal balance between control precision and calculation efficiency. In order to achieve the above purpose, the technical scheme adopted by the invention comprises the following steps: an intelligent automobile track tracking control method based on self-adaptive switching MPC comprises the following steps: S1, establishing a vehicle mechanism dynamics model, wherein the vehicle mechanism dynamics model comprises a nonlinear model of three-degree-of-freedom vehicle dynamics and a linear model thereof; S2, establishing an NARX neural network with delay input, and training to obtain a vehicle dynamics model; s3, designing an MPC track tracking control algorithm, wherein the MPC track tracking control algorithm comprises an LMPC prediction controller, an NMPC prediction controller and an NARX-MPC prediction controller which correspond to the linear model, the nonlinear model and the vehicle dynamics model obtained in the S1 and the S2 respectively; S4, designing a self-adaptive switching MPC controller, taking road curvature rho, road adhesion coefficient mu and longitudinal vehicle speed vx as fuzzy inputs, taking transverse error e as variable domain scaling factor alpha, adopting variable domain fuzzy control, combining scaling factor alpha to adjust an output domain, outputting a selection instruction of a linear model, a nonlinear model or a vehicle dynamics model, dynamically switching and activating the LMPC prediction controller, the NMPC prediction controller or the NARX-MPC prediction controller in S3, and executing track tracking control at the current moment. Preferably, the differential equation of the nonlinear model of three-degree-of-freedom vehicle dynamics in S1 is: Wherein m is the vehicle mass, v x and v y are the longitudinal speed and lateral speed of the mass center under the vehicle body coordinate system respectively, AndThe change rates of longitudinal and lateral speeds, namely longitudinal and lateral accelerations of the vehicle in the body coordinate system, are respectively expressed, X and Y are the position coordinates of the vehicle centroid in the global coordinate system,AndRepresenting velocity components of the vehicle in global X and Y directions, respectively; For the yaw angle of the vehicle, The change rate of the yaw angle, that is, the yaw rate of the vehicle,Representing a change rate of yaw rate, that is, yaw acceleration of the vehicle; And The linear cornering stiffness of the front tire and the rear tire respectively, delta f is the cornering angle of the front wheel, l f and l r are the distances from the center of mass of the vehicle to the front axle and the rear axle respectively, C lf and C lr are the longitudinal cornering stiffness of the front tire and the r