Method and device for trajectory tracking of autonomous vehicle based on data-driven model predictive control

Abstract

The application discloses an automatic driving vehicle trajectory tracking method and device based on a data-driven model predictive control, and the method comprises the following steps: discretizing a continuous vehicle lateral and longitudinal error model to obtain a discrete vehicle lateral and longitudinal error model and selecting control input and system output; collecting multiple data as prior input data based on the control input and constructing a Hankel matrix; applying the prior input data to the discrete vehicle lateral and longitudinal error model, collecting corresponding generated prior output data based on the system output, and constructing a Hankel matrix; based on the behavior system theory knowledge, the Hankel matrix constructed by the prior input and output data is used to represent the dynamic characteristics of the system; based on the data representation of the nominal system model corresponding to the discrete vehicle lateral and longitudinal error model, a cost function is constructed, and then an optimization problem of the data-driven model predictive control is constructed; the optimization problem is solved, and the obtained optimal control is applied to the system to complete the control. The application can drive the system to run along the expected trajectory while meeting the system constraints, and ensure the high precision and robustness of the trajectory tracking.

Description

Technical Field

[0001] The present invention relates to the field of autonomous driving technology, and in particular to a method, device, terminal device and computer-readable storage medium for tracking the trajectory of an autonomous driving vehicle based on data-driven model predictive control. Background Art

[0002] Autonomous vehicle systems consist of multiple modules, including perception, planning, decision-making, and control. The control module is one of the most critical. With the continuous development of autonomous driving technology, trajectory tracking has become a key technology within the control module. Existing trajectory tracking methods mainly include proportional-integral-derivative (PID) control, model predictive control (MPC), adaptive control, sliding mode control (SMC), and linear quadratic regulator (LQR). These different trajectory tracking methods each have their own advantages and disadvantages. Among them, model predictive control (MPC) is a commonly used advanced control method in trajectory tracking. MPC solves an optimization problem to predict the vehicle's trajectory over a period of time and calculates the optimal control input to minimize the trajectory error. Compared with other trajectory tracking methods, MPC's main advantages lie in its ability to implement rolling optimization and explicitly handle constraints. However, traditional MPC relies on an accurate vehicle dynamics model, which has a large number of parameters and requires experimental parameter identification. In addition, the complexity and uncertainty of real-world road environments often result in poor robustness in the face of unexpected events, external disturbances, or incomplete system models. Summary of the Invention

[0003] In order to address the deficiencies of the above-mentioned prior art, the present invention provides a method, apparatus, terminal device and computer-readable storage medium for autonomous driving vehicle trajectory tracking based on data-driven model predictive control. These methods avoid the precise modeling requirements of traditional model predictive control for the system, directly utilize limited offline data and behavioral system theory, and provide real-time feedback to calculate the optimal and safe control strategy. This method drives the system along the desired trajectory while satisfying system constraints, ensuring high accuracy and robustness of trajectory tracking.

[0004] The first object of the present invention is to provide a method for tracking the trajectory of an autonomous driving vehicle based on data-driven model predictive control.

[0005] The second object of the present invention is to provide an autonomous driving vehicle trajectory tracking device based on data-driven model predictive control.

[0006] The third object of the present invention is to provide a terminal device.

[0007] A fourth object of the present invention is to provide a computer-readable storage medium.

[0008] The first object of the present invention can be achieved by adopting the following technical solutions:

[0009] A method for tracking trajectory of an autonomous driving vehicle based on data-driven model predictive control, the method comprising:

[0010] Discretize the continuous vehicle lateral and longitudinal error models to obtain discrete vehicle lateral and longitudinal error models; select control inputs and system outputs based on the discrete vehicle lateral and longitudinal error models;

[0011] Based on the control input, multiple input data are collected and used as a set of prior input data; a corresponding Hankel matrix is ​​constructed according to the prior input data and it is confirmed that the collected prior input data is continuously excited;

[0012] Applying the priori input data to the discrete vehicle lateral and longitudinal error model, collecting the corresponding priori output data based on the system output; constructing the corresponding Hankel matrix based on the priori output data;

[0013] Based on the theoretical knowledge of behavioral systems, the dynamic characteristics of the system are expressed using the Hankel matrix constructed from the prior input data and the Hankel matrix constructed from the prior output data;

[0014] Based on the data representation of the nominal system model corresponding to the discrete vehicle lateral and longitudinal error model, a cost function is constructed; based on the cost function, an optimization problem of data-driven model predictive control is constructed;

[0015] The optimization problem is solved at every l time step, and the first l input data in the obtained optimal control sequence are applied to the autonomous driving vehicle system to complete the control of l time step; l is a positive integer greater than or equal to 1.

[0016] Furthermore, the discrete vehicle lateral and longitudinal error model is:

[0017]

[0018] in:

[0019] I is the identity matrix;

[0020]

[0021]

[0022] Where x(k+1) and x(k) represent the states of the system at time k+1 and k respectively, u(k) represents the control input at time k, and y(k) represents the system output at time k; are all coefficient matrices; m is the vehicle mass; I zis the moment of inertia of the vehicle around the z axis in the vehicle coordinate system, l f and l r Respectively front wheelbase and rear wheelbase; C af and C ar are the cornering stiffness of the front and rear wheels respectively; V x is the vehicle longitudinal acceleration; Indicates the angular velocity required for the vehicle body to turn the desired angle;

[0023] According to the discrete vehicle lateral and longitudinal error model, the front wheel angle δ and acceleration a of the autonomous vehicle are selected as control inputs to select input data; the lateral error y is selected. e , lateral error rate Heading error θ e , heading error rate Longitudinal error x e and longitudinal error rate Select the output data as system output.

[0024] Furthermore, the dynamic characteristics of the system are represented by the Hankel matrix constructed based on the prior input data and the Hankel matrix constructed based on the prior output data based on the behavioral system theory knowledge, including:

[0025] Ignoring the disturbance term in the discrete vehicle lateral and longitudinal error model

[0026] According to the theoretical knowledge of behavioral systems, any set of input and output data sequences of the system can be expressed as in, is the input vector consisting of the input data sequence, is the output vector consisting of the output data sequence, The input vector and the output vector The augmented vector composed of is the Hankel matrix H constructed from the prior input data L+n (u 0 ) and the Hankel matrix H constructed from the prior output data L+n (y 0 ), the vector g is an NL-l+1 dimensional real vector, N is the number of prior input data and N ≥ (m+1)(L+l+n-1), n ​​is the state dimension of the discrete vehicle lateral and longitudinal error model, m is the input data dimension, L is the prediction domain of data-driven model predictive control and L>n+l.

[0027] Furthermore, constructing a corresponding Hankel matrix based on the priori input data and confirming that the collected priori input data is continuously excited includes:

[0028] Form the prior input data into a prior input vector;

[0029] Construct the corresponding Hankel matrix based on the prior input vector;

[0030] If the Hankel matrix is ​​of full row rank, it is confirmed that the collected priori input data is continuously excited; otherwise, a new set of priori input data is collected and judged whether it is continuously excited, until the collected priori input data is continuously excited.

[0031] Furthermore, solving the optimization problem at every l time step and applying the first l control input data in the obtained optimal control sequence to the autonomous driving vehicle system to complete the control of l time step includes:

[0032] At the initial time k=0, the optimization problem is solved using the actual control input data and output data before k=0 as initial values ​​to obtain the optimal control input data sequence and the optimal control output data sequence at time 0;

[0033] The first l items in the optimal control input data sequence are selected as the control input data for the next l time steps and applied to the system; then, when k=l, the l actual control input data and output data before k=l are used as initial values ​​to solve the optimization problem, and the first l items in the obtained optimal control input data sequence are selected as the control input data for the next l time steps and applied to the system; the optimization problem is solved in this way, and the first l input data in the obtained optimal control input data sequence are applied to the system to complete the control of l time steps and apply to the system to ensure the stability of the system and improve the robustness.

[0034] Furthermore, the optimization problem is:

[0035]

[0036] Where, Represents the cost function J at time k L (u(k),y(k)) is minimized, and the optimal solution is found under the constraints. and g * (k); represents an L+l-dimensional vector consisting of the predictive control input data at time k, which is l time steps backward and L time steps forward. Represents the L+l dimensional vector composed of the predicted output data of l time steps backward and L time steps forward at time k, H L+l (u 0 ) represents the input data u controlled by the prior 0 The generated L+l order Hankel matrix, H L+l (y 0) represents the prior output data y 0 The generated L+l order Hankel matrix, u [k-l,k-1] represents the actual control input data of the system l time steps before time k; y [k-l,k-1] represents the actual control output data of the system l time steps before time k; represents the system input data to be predicted at time k; represents the system output data to be predicted at time k; represents the terminal constraint region; is a bounded set of tight constraints on the predictive control input, is a bounded set of tight constraints on the predicted output.

[0037] Furthermore, the cost function is:

[0038]

[0039] Where, J L (u(k), y(k)) represents the cost function, u(k) represents the actual control input data of the k system at the current moment, and y(k) represents the actual output data of the k system at the current moment. represents the predictive control input data at the current time k and j time steps forward, Represents the predicted output data at the current time k j time steps ahead, Represents the expected output data at the current time k j time steps forward; is and The augmented state of the composition, represents the vector consisting of the last l prediction control inputs within the current k prediction range, Represents the vector composed of the last l predicted outputs within the current k prediction range; R>0 is the control input weight matrix, Q>0 is the output weight matrix, P>0 is the terminal cost weight matrix, λ g is the weight factor, n is the upper bound of the noise; g(k) represents the NL-l+1 dimensional real vector composed of the linear combination coefficients of the column vector of the k Hankel matrix at the current moment.

[0040] The second object of the present invention can be achieved by adopting the following technical solutions:

[0041] A device for tracking trajectory of an autonomous driving vehicle based on data-driven model predictive control, the device comprising:

[0042] Selecting a module for discretizing the continuous vehicle lateral and longitudinal error model to obtain a discrete vehicle lateral and longitudinal error model; selecting a control input and a system output based on the discrete vehicle lateral and longitudinal error model;

[0043] An acquisition module is configured to acquire multiple input data based on the control input and use them as a set of prior input data; construct a corresponding Hankel matrix based on the prior input data and confirm that the acquired prior input data is continuously excited; apply the prior input data to a discrete vehicle lateral and longitudinal error model, acquire corresponding prior output data based on the system output; and construct a corresponding Hankel matrix based on the prior output data;

[0044] The representation module is used to represent the dynamic characteristics of the system using the Hankel matrix constructed from the prior input data and the Hankel matrix constructed from the prior output data based on the theoretical knowledge of the behavioral system;

[0045] A construction module is used to construct a cost function based on the data representation of the nominal system model corresponding to the discrete vehicle lateral and longitudinal error model; based on the cost function, an optimization problem of data-driven model predictive control is constructed;

[0046] A control module is configured to solve the optimization problem at every l time steps and apply the first l input data in the obtained optimal control sequence to the autonomous driving vehicle system to complete the control of l time steps; l is a positive integer greater than or equal to 1.

[0047] The third object of the present invention can be achieved by adopting the following technical solutions:

[0048] A terminal device includes a processor and a memory for storing a program executable by the processor. When the processor executes the program stored in the memory, it implements the above-mentioned autonomous driving vehicle trajectory tracking method based on data-driven model predictive control.

[0049] The fourth object of the present invention can be achieved by adopting the following technical solutions:

[0050] A computer-readable storage medium stores a program, which, when executed by a processor, implements the above-mentioned autonomous driving vehicle trajectory tracking method based on data-driven model predictive control.

[0051] The present invention has the following beneficial effects compared to the prior art:

[0052] The present invention directly utilizes limited data samples from offline measurements and theoretical knowledge of behavioral systems to achieve real-time optimal control of unknown systems without the need for complex system identification, thus simplifying the design process of the control system and improving adaptability. It can effectively cope with system uncertainties and external disturbances in dynamic traffic environments, ensuring high-precision trajectory tracking while improving the robustness of the control system. It calculates the optimal and safe control strategy through real-time feedback, quickly drives the vehicle along the desired trajectory while satisfying system constraints, and is suitable for autonomous driving needs in complex dynamic traffic scenarios. BRIEF DESCRIPTION OF THE DRAWINGS

[0053] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on the structures shown in these drawings without paying any creative work.

[0054] Figure 1 This is a flowchart of a method for tracking trajectory of an autonomous driving vehicle based on data-driven model predictive control according to Example 1 of the present invention;

[0055] Figure 2 This is a diagram showing the principle of data-driven model predictive control according to embodiment 1 of the present invention;

[0056] Figure 3 This is a structural block diagram of an autonomous driving vehicle trajectory tracking device based on data-driven model predictive control according to embodiment 2 of the present invention;

[0057] Figure 4 This is a structural block diagram of the terminal device of embodiment 3 of the present invention. DETAILED DESCRIPTION

[0058] To make the purpose, technical solutions, and advantages of the embodiments of the present invention more clear, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention. It should be understood that the specific embodiments described are only used to explain this application and are not used to limit this application.

[0059] Example 1:

[0060] like Figure 1 、 2As shown, this embodiment provides a method for tracking the trajectory of an autonomous driving vehicle based on data-driven model predictive control, comprising the following steps:

[0061] S101. Discretize the continuous vehicle lateral and longitudinal error models to obtain discrete vehicle lateral and longitudinal error models; and select control inputs and system outputs based on the discrete vehicle lateral and longitudinal error models.

[0062] Based on the vehicle dynamics relationship, the continuous vehicle lateral and longitudinal error model can be established as:

[0063]

[0064] Among them: x represents the state of the system, represents the derivative of x; u represents the control input; A and B represent the coefficient matrices; W is the disturbance matrix; and:

[0065]

[0066]

[0067] Where y e represents the lateral error, i.e. the distance between the vehicle's center of mass and the lane centerline; is the lateral error rate; θ e is the heading error; is the heading error rate; x e is the longitudinal error; is the longitudinal error rate; δ is the front wheel angle, a is the acceleration; m is the vehicle mass; I z is the moment of inertia of the vehicle around the z axis in the vehicle coordinate system, l f and l r Respectively front wheelbase and rear wheelbase; C af and C ar are the cornering stiffness of the front and rear wheels respectively; V x is the vehicle longitudinal acceleration; Indicates the angular velocity required for the vehicle body to turn the desired angle.

[0068] Since the subsequent steps require the use of discrete data, the continuous vehicle lateral and longitudinal error model is discretized using the bilateral discretization method. Assuming the sampling period is 1s, let The discrete vehicle lateral and longitudinal error model is obtained as follows:

[0069]

[0070] Where x(k) represents the state of the system at time k, u(k) represents the control input at time k, y(k) represents the system output at time k, and x(k+1) represents the state of the system at the next time k+1. The specific parameter values ​​are unknown, the disturbance term The matrix norm of has a constant upper bound Right now

[0071] Based on the discrete vehicle lateral and longitudinal error model (2), the front wheel steering angle δ and acceleration a of the autonomous vehicle are selected as control inputs to select the input data u in the subsequent steps; the lateral error y is selected e , lateral error rate Heading error θ e , heading error rate Longitudinal error x e and longitudinal error rate As system output so that the subsequent steps can select output data y.

[0072] S102 , collecting multiple input data based on the control input and using them as a set of priori input data; constructing a corresponding Hankel matrix according to the priori input data and confirming that the collected priori input data is continuously excited.

[0073] According to the control input, collect N input data to form a priori input data set where N ≥ (m + 1)(L + l + n - 1), where n is the state dimension of the discrete vehicle lateral and longitudinal error model (2), m is the control input dimension, L is the prediction horizon of the data-driven model predictive control, and l is the number of time steps before the current moment.

[0074] The collected priori input data set is formed into a priori input vector u 0 =[u(0) u(1) … u(N-1)]; construct the corresponding L+n-order Hankel matrix based on the prior input vector:

[0075]

[0076] And verify whether the Hankel matrix of the prior input vector is full row rank, that is, rank(H L+n (u 0 ))=m(L+n) holds. If so, the Hankel matrix is ​​full row rank, and the acquired prior input data set is continuously excited, that is, it is an ideal data set. If the Hankel matrix is ​​not full row rank, reacquire the prior input data until the acquired prior input data set is continuously excited.

[0077] S103 , applying the priori input data to a discrete vehicle lateral and longitudinal error model, collecting the corresponding priori output data based on the system output; and constructing a corresponding Hankel matrix based on the priori output data.

[0078] The collected continuous excitation prior input data set is open-loop applied to the discrete vehicle lateral and longitudinal error model (2), and the corresponding N prior output data sets are collected. The collected priori output data set is formed into a priori output vector y 0 =[y(0) y(1) … y(N-1)]; construct the corresponding L+n-order Hankel matrix based on the prior output vector:

[0079]

[0080] S104. Based on the theoretical knowledge of behavioral systems, the dynamic characteristics of the system are represented by the Hankel matrix constructed using the prior input data and the Hankel matrix constructed using the prior output data.

[0081] Ignore the disturbance term in the discrete vehicle lateral and longitudinal error model (2) The nominal system model corresponding to the discrete vehicle lateral and longitudinal error model (2) is defined as:

[0082]

[0083] Where, is the nominal system state, is the nominal system output, is the control input of the nominal system. Model parameters Unknown. According to the knowledge of behavioral system theory, any set of input and output sequences of the nominal system (3) It can be expressed as in is the input vector consisting of the input sequence, is the output vector consisting of the output sequence, The input vector and the output vector The augmented vector composed of is the Hankel matrix H L+n (u 0 ) and H L+n (y 0 ), and the vector g is an NL-l+1 dimensional real vector.

[0084] S105. Construct a cost function based on the data representation of the nominal system model corresponding to the discrete vehicle lateral and longitudinal error model; and construct an optimization problem for data-driven model predictive control based on the cost function.

[0085] The cost function is:

[0086]

[0087] Where, JL (u(k), y(k)) represents the cost function, u(k) represents the actual control input data of the k system at the current moment, and y(k) represents the actual output data of the k system at the current moment. represents the predictive control input data at the current time k and j time steps forward, Represents the predicted output data at the current time k j time steps ahead, Represents the expected output data at the current time k j time steps forward; is and The augmented state of the composition, represents the vector consisting of the last l prediction control inputs within the current k prediction range, Represents the vector composed of the last l predicted outputs within the current k prediction range; R>0 is the control input weight matrix, Q>0 is the output weight matrix, P>0 is the terminal cost weight matrix, λ g is the weight factor, is the upper bound of the noise; g(k) represents the NL-l+1 dimensional real vector composed of the linear combination coefficients of the column vector of the k Hankel matrix at the current moment.

[0088] Based on the cost function J L (u(k), y(k)), the data-driven model predictive control optimization problem is further constructed as:

[0089]

[0090] Where, Represents the cost function J at time k L (u(k),y(k)) is minimized, and the optimal solution is found under the constraints. Represents the L+l dimensional vector composed of the predictive control input data at time k, which is l time steps backward and L time steps forward. Similarly, Represents the L+l dimensional vector composed of the predicted output data of l time steps backward and L time steps forward at time k, H L+l (u 0 ) represents the input data u controlled by the prior 0 The generated L+l order Hankel matrix, H L+l (y 0 ) represents the prior output data y 0 The generated L+l order Hankel matrix, u [k-l,k-1] represents the actual control input data of the system l time steps before time k; y [k-l,k-1] represents the actual control output data of the system l time steps before time k; represents the system input data to be predicted at time k; represents the system output data to be predicted at time k; represents the terminal constraint region; is a bounded set of tight constraints on the predictive control input, is a bounded set of tight constraints on the predicted output.

[0091] S106. Solve the optimization problem at each moment, and apply the first l input data in the obtained optimal control sequence to the autonomous driving vehicle system to complete l time step control.

[0092] Initial time k = 0, with l actual control input data u before k = 0 [-l,-1] (0) and output data y [-l,-1] (0) is used as the initial value to solve the optimization problem in step 105 and obtain the optimal solution and g * (0).

[0093] Selecting optimal control input data The first item As the control input data for the next l time steps, it is applied to the autonomous driving vehicle system; then, when k=l, the actual control input data u before k=l is used. [-l,-1] (k) and output data y [-l,-1] (k) is used as the initial value, and the optimization problem in step 105 is solved when k=1 to obtain the optimal solution. g * (l). Select the optimal control input The first item As the control input data for the next l time steps, it is applied to the autonomous driving system, and so on. Towards terminal area Ξ ∈ When selecting As the control input data for the next moment, it ensures that the autonomous driving vehicle system remains stable and robust to external interference and environmental changes.

[0094] Those skilled in the art will appreciate that all or part of the steps in the method for implementing the above embodiments may be completed by instructing related hardware through a program, and the corresponding program may be stored in a computer-readable storage medium.

[0095] It should be noted that although the method operations of the above embodiments are described in a particular order in the accompanying drawings, this does not require or imply that the operations must be performed in this particular order, or that all of the illustrated operations must be performed to achieve the desired results. Rather, the depicted steps may be performed in a different order. Additionally or alternatively, certain steps may be omitted, multiple steps may be combined into a single step, and / or a single step may be broken down into multiple steps.

[0096] Example 2:

[0097] like Figure 3 As shown, this embodiment provides an autonomous driving vehicle trajectory tracking device based on data-driven model predictive control, which includes a selection module 301, an acquisition module 302, a representation module 303, a construction module 304 and a control module 305, wherein:

[0098] A selection module 301 is used to discretize the continuous vehicle lateral and longitudinal error models to obtain discrete vehicle lateral and longitudinal error models; based on the discrete vehicle lateral and longitudinal error models, a control input and a system output are selected;

[0099] The acquisition module 302 is configured to acquire a plurality of input data based on the control input and use them as a set of prior input data; construct a corresponding Hankel matrix based on the prior input data and confirm that the acquired prior input data is continuously excited; apply the prior input data to a discrete vehicle lateral and longitudinal error model, acquire corresponding prior output data based on the system output; and construct a corresponding Hankel matrix based on the prior output data.

[0100] A representation module 303 is used to represent the dynamic characteristics of the system using a Hankel matrix constructed from a priori input data and a Hankel matrix constructed from a priori output data based on theoretical knowledge of behavioral systems;

[0101] A construction module 304 is configured to construct a cost function based on the data representation of the nominal system model corresponding to the discrete vehicle lateral and longitudinal error model; and to construct an optimization problem for data-driven model predictive control based on the cost function.

[0102] The control module 305 is configured to solve the optimization problem every l time steps and apply the first l input data in the obtained optimal control sequence to the autonomous driving vehicle system to complete control for l time steps; l is a positive integer greater than or equal to 1.

[0103] The specific implementation of each module in this embodiment can be found in the above-mentioned embodiment 1, and will not be described one by one here; it should be noted that the device provided in this embodiment is only illustrated by the division of the above-mentioned functional modules. In actual applications, the above-mentioned functions can be allocated to different functional modules as needed, that is, the internal structure can be divided into different functional modules to complete all or part of the functions described above.

[0104] Example 3:

[0105] This embodiment provides a terminal device, which can be a computer, such as Figure 4 As shown, it comprises a processor 402, a memory, an input device 403, a display 404 and a network interface 405 connected via a system bus 401. The processor is used to provide computing and control capabilities. The memory includes a non-volatile storage medium 406 and an internal memory 407. The non-volatile storage medium 406 stores an operating system, a computer program and a database. The internal memory 407 provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. When the processor 402 executes the computer program stored in the memory, the autonomous driving vehicle trajectory tracking method based on data-driven model predictive control of the above-mentioned embodiment 1 is implemented.

[0106] Example 4:

[0107] This embodiment provides a computer-readable storage medium storing a computer program. When the computer program is executed by a processor, the autonomous driving vehicle trajectory tracking method based on data-driven model predictive control of the above-mentioned embodiment 1 is implemented.

[0108] It should be noted that the computer-readable storage medium of the present embodiment may be a computer-readable signal medium or a computer-readable storage medium or any combination thereof. The computer-readable storage medium may be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, device, or component, or any combination thereof. More specific examples of computer-readable storage media may include, but are not limited to, an electrical connection having one or more wires, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination thereof.

[0109] In summary, the present invention achieves real-time optimal control of unknown systems by directly utilizing limited offline data and behavioral systems theory. This addresses the reliance of traditional trajectory tracking methods on precise system modeling, overcomes the complex parameter identification and the difficulty in balancing control accuracy and robustness in traditional model predictive control, and effectively addresses the challenges of uncertainty and complexity in dynamic traffic environments. This invention achieves the ability to drive an unknown system along a desired trajectory while satisfying system constraints, ensuring high-precision and robust trajectory tracking, and provides an efficient and reliable control strategy for autonomous vehicles.

[0110] The above is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any technician familiar with the technical field can make equivalent replacements or changes based on the technical solution and inventive concept of the present invention within the scope disclosed by the present invention, which falls within the scope of protection of the present invention.

Claims

1. A method for tracking the trajectory of an autonomous vehicle based on data-driven model predictive control, characterized in that: The method comprises: Discretize the continuous vehicle lateral and longitudinal error models to obtain discrete vehicle lateral and longitudinal error models; select control inputs and system outputs based on the discrete vehicle lateral and longitudinal error models; Based on the control input, multiple input data are collected and used as a set of prior input data; a corresponding Hankel matrix is ​​constructed according to the prior input data and it is confirmed that the collected prior input data is continuously excited; Applying the priori input data to the discrete vehicle lateral and longitudinal error model, collecting the corresponding priori output data based on the system output; constructing the corresponding Hankel matrix based on the priori output data; Based on the theoretical knowledge of behavioral systems, the dynamic characteristics of the system are expressed using the Hankel matrix constructed from the prior input data and the Hankel matrix constructed from the prior output data; Based on the data representation of the nominal system model corresponding to the discrete vehicle lateral and longitudinal error model, a cost function is constructed; based on the cost function, an optimization problem of data-driven model predictive control is constructed; The optimization problem is solved at every l time step, and the first l input data in the obtained optimal control sequence are applied to the autonomous driving vehicle system to complete the control of l time step; l is a positive integer greater than or equal to 1.

2. The autonomous driving vehicle trajectory tracking method according to claim 1, characterized in that: The discrete vehicle lateral and longitudinal error model is: in: I is the identity matrix; Where x(k+1) and x(k) represent the states of the system at time k+1 and k respectively, u(k) represents the control input at time k, and y(k) represents the system output at time k; are all coefficient matrices; m is the vehicle mass; I z is the moment of inertia of the vehicle around the z axis in the vehicle coordinate system, l f and l r Respectively front wheelbase and rear wheelbase; C af and C ar are the cornering stiffness of the front and rear wheels respectively; V x is the vehicle longitudinal acceleration; Indicates the angular velocity required for the vehicle body to turn the desired angle; According to the discrete vehicle lateral and longitudinal error model, the front wheel angle δ and acceleration a of the autonomous vehicle are selected as control inputs to select input data; the lateral error y is selected. e , lateral error rate Heading error θ e , heading error rate Longitudinal error x e and longitudinal error rate Select the output data as system output.

3. The autonomous driving vehicle trajectory tracking method according to claim 2, characterized in that: The Hankel matrix constructed based on the behavioral system theory and the Hankel matrix constructed based on the prior input data and the Hankel matrix constructed based on the prior output data represent the dynamic characteristics of the system, including: Ignoring the disturbance term in the discrete vehicle lateral and longitudinal error model According to the theoretical knowledge of behavioral systems, any set of input and output data sequences of the system can be expressed as in, is the input vector consisting of the input data sequence, is the output vector consisting of the output data sequence, The input vector and the output vector The augmented vector composed of is the Hankel matrix H constructed from the prior input data L+n (u 0 ) and the Hankel matrix H constructed from the prior output data L+n (y 0 ), the vector g is an NL-l+1 dimensional real vector, N is the number of prior input data and N ≥ (m+1)(L+l+n-1), n ​​is the state dimension of the discrete vehicle lateral and longitudinal error model, m is the input data dimension, L is the prediction domain of data-driven model predictive control and L>n+l.

4. The method for tracking the trajectory of an autonomous driving vehicle according to any one of claims 1 to 3, characterized in that: The constructing of a corresponding Hankel matrix according to the priori input data and confirming that the collected priori input data is continuously excited includes: Form the prior input data into a prior input vector; Construct the corresponding Hankel matrix based on the prior input vector; If the Hankel matrix is ​​of full row rank, it is confirmed that the collected priori input data is continuously excited; otherwise, a new set of priori input data is collected and judged whether it is continuously excited, until the collected priori input data is continuously excited.

5. The autonomous driving vehicle trajectory tracking method according to claim 1, characterized in that: Solving the optimization problem at each time step and applying the first control input data in the obtained optimal control sequence to the autonomous driving vehicle system to complete control for the time step includes: At the initial time k=0, the optimization problem is solved using the actual control input data and output data before k=0 as initial values ​​to obtain the optimal control input data sequence and the optimal control output data sequence at time 0; The first l items in the optimal control input data sequence are selected as the control input data for the next l time steps and applied to the system; then, when k=l, the l actual control input data and output data before k=l are used as initial values ​​to solve the optimization problem, and the first l items in the obtained optimal control input data sequence are selected as the control input data for the next l time steps and applied to the system; the optimization problem is solved in this way, and the first l input data in the obtained optimal control input data sequence are applied to the system to complete the control of l time steps and apply to the system to ensure the stability of the system and improve the robustness.

6. The autonomous driving vehicle trajectory tracking method according to any one of claims 1 to 3 and 5, characterized in that: The cost function is: Where, J L (u(k), y(k)) represents the cost function, u(k) represents the actual control input data of the k system at the current moment, and y(k) represents the actual output data of the k system at the current moment. represents the predictive control input data at the current time k and j time steps forward, Represents the predicted output data at the current time k j time steps ahead, Represents the expected output data at the current time k j time steps forward; is and The augmented state of the composition, represents the vector consisting of the last l prediction control inputs within the current k prediction range, Represents the vector consisting of the last l prediction outputs within the k prediction range at the current moment; is the control input weight matrix, is the output weight matrix, is the weight matrix of the terminal cost, λ g is the weight factor, is the upper bound of the noise; g(k) represents the NL-l+1 dimensional real vector composed of the linear combination coefficients of the column vector of the k Hankel matrix at the current moment.

7. The autonomous driving vehicle trajectory tracking method according to claim 6, characterized in that: The optimization problem is: Where, Represents the cost function J at time k L (u(k),y(k)) is minimized, and the optimal solution is found under the constraints. and g * (k); represents the L+l dimensional vector consisting of the predictive control input data at time k, which is l time steps backward and L time steps forward. Represents the L+l dimensional vector composed of the predicted output data of l time steps backward and L time steps forward at time k, H L+l (u 0 ) represents the input data u controlled by the prior 0 The generated L+l order Hankel matrix, H L+l (y 0 ) represents the prior output data y 0 The generated L+l order Hankel matrix, u [k-l,k-1] represents the actual control input data of the system l time steps before time k; y [k-l,k-1] represents the actual control output data of the system l time steps before time k; represents the system input data to be predicted at time k; represents the system output data to be predicted at time k; represents the terminal constraint region; is a bounded set of tight constraints on the predictive control input, is a bounded set of tight constraints on the predicted output.

8. An autonomous driving vehicle trajectory tracking device based on data-driven model predictive control, characterized in that: The device comprises: Selecting a module for discretizing the continuous vehicle lateral and longitudinal error model to obtain a discrete vehicle lateral and longitudinal error model; selecting a control input and a system output based on the discrete vehicle lateral and longitudinal error model; An acquisition module is configured to acquire multiple input data based on the control input and use them as a set of prior input data; construct a corresponding Hankel matrix based on the prior input data and confirm that the acquired prior input data is continuously excited; apply the prior input data to a discrete vehicle lateral and longitudinal error model, acquire corresponding prior output data based on the system output; and construct a corresponding Hankel matrix based on the prior output data; The representation module is used to represent the dynamic characteristics of the system using the Hankel matrix constructed from the prior input data and the Hankel matrix constructed from the prior output data based on the theoretical knowledge of the behavioral system; A construction module is used to construct a cost function based on the data representation of the nominal system model corresponding to the discrete vehicle lateral and longitudinal error model; based on the cost function, an optimization problem of data-driven model predictive control is constructed; A control module is configured to solve the optimization problem at every l time steps and apply the first l input data in the obtained optimal control sequence to the autonomous driving vehicle system to complete the control of l time steps; l is a positive integer greater than or equal to 1.

9. A terminal device comprising a processor and a memory for storing a program executable by the processor, characterized in that: When the processor executes the program stored in the memory, it implements the autonomous driving vehicle trajectory tracking method described in any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the autonomous driving vehicle trajectory tracking method according to any one of claims 1 to 7 is implemented.

Images