A Multi-Preview Point Path Tracking Control Method and System Based on Neural Network

Abstract

The present invention discloses a multi-preview-point path tracking control method and system based on a neural network. The method includes: establishing a nonlinear model predictive controller based on a time-varying local model in a vehicle body coordinate system; controlling a controlled object to track a preset forward reference path through the nonlinear model predictive controller, collecting forward reference path information, and generating training samples; constructing a path tracking controller by using a preset neural network, and training the constructed path tracking controller by using the training samples; and realizing path tracking control of the controlled object by using the trained path tracking controller. The solution of the present invention solves the problems in the prior art that the forward reference path information cannot be effectively utilized and the accuracy is poor when the curvature of the reference path changes greatly.

Description

Technical Field

[0001] The present invention relates to the technical field of motion control of driverless devices, and particularly to a multi-preview point path tracking control method and system based on a neural network. Background Art

[0002] In the existing path tracking control technology, non-linear model predictive control has the advantages of being able to explicitly handle system constraints, effectively utilize the information of the forward reference path, and weaken the influence of disturbances such as positioning errors (Bai Guoxing, Meng Yu, Liu Li, etc. Research status of path tracking control for driverless vehicles [J]. Journal of Engineering Science, 2021, 43(4): 475-485; Bai Guoxing, Luo Weidong, Liu Li, etc. Research status and progress of path tracking control for mine articulated vehicles [J]. Journal of Engineering Science, 2021, 43(2): 193-204; Bai G, Meng Y, Liu L, et al. Review and comparison of path tracking based on model predictive control [J]. Electronics, 2019, 8(10): 1077). However, its real-time performance is still worse than that of control methods such as linear model predictive control even after optimization (Bai Guoxing, Liu Li, Meng Yu, etc. Real-time path tracking of mobile robots based on non-linear model predictive control [J]. Transactions of the Chinese Society for Agricultural Machinery, 2020, 51(9): 47-52).

[0003] To address the problem of poor real-time performance, there is currently a path tracking control method that improves real-time performance by using a neural network to learn a non-linear model predictive controller (CN111624992B). However, since the forward reference path information is not considered in the training samples, this path tracking control method cannot effectively utilize the forward reference path information and has poor accuracy when the curvature change amplitude of the reference path is large. Summary of the Invention

[0004] The present invention provides a multi-preview point path tracking control method and system based on a neural network to solve the problem that the existing path tracking control method using a neural network to learn a non-linear model predictive controller cannot effectively utilize the forward reference path information and has poor accuracy when the curvature change amplitude of the reference path is large.

[0005] To solve the above technical problems, the present invention provides the following technical solutions:

[0006] On the one hand, the present invention provides a multi-preview point path tracking control method based on a neural network, and the multi-preview point path tracking control method based on a neural network includes:

[0007] Establish a nonlinear model predictive controller based on a time-varying local model in the vehicle body coordinate system;

[0008] Through the nonlinear model predictive controller based on the time-varying local model, control the controlled object to track a preset forward reference path, and collect the forward reference path information to generate training samples; wherein, the reference path information includes: the horizontal and vertical coordinates of the reference path, and the heading angle;

[0009] Construct a path tracking controller using a preset neural network, and train the constructed path tracking controller using the training samples; wherein, the input of the path tracking controller includes the initial yaw rate and the reference path information, and the output of the path tracking controller includes the target yaw rate;

[0010] Use the trained path tracking controller to achieve the path tracking control of the controlled object.

[0011] Further, establishing a nonlinear model predictive controller based on a time-varying local model in the vehicle body coordinate system includes:

[0012] Transfer the predictive controller and a section of the forward reference path of the controlled object into the vehicle body coordinate system at the beginning of each control cycle, and establish a nonlinear model predictive controller based on the transformed coordinates and a time-varying local model.

[0013] Further, the constructing a path tracking controller using a preset neural network and training the constructed path tracking controller using the training samples includes:

[0014] Construct a path tracking controller using a preset neural network according to the structure of the training samples;

[0015] Train the constructed path tracking controller using the training samples.

[0016] Further, the input of the path tracking controller includes: the initial yaw rate, the horizontal and vertical coordinates and the heading angle of the first reference point in the reference path, the horizontal and vertical coordinates and the heading angle of an intermediate reference point in the reference path, and the horizontal and vertical coordinates and the heading angle of the last reference point in the reference path.

[0017] On the other hand, the present invention also provides a multi-preview-point path tracking control system based on a neural network, and the multi-preview-point path tracking control system based on a neural network includes:

[0018] A nonlinear model predictive controller construction module, configured to establish a nonlinear model predictive controller based on a time-varying local model in the vehicle body coordinate system;

[0019] A training sample generation module, which is used to predict, through the non-linear model predictive controller construction module, the non-linear model predictive controller based on the time-varying local model constructed by the non-linear model predictive controller construction module, control the controlled object to track a preset forward reference path, and collect the forward reference path information to generate training samples; wherein, the reference path information includes: the horizontal and vertical coordinates of the reference path, and the heading angle.

[0020] A path tracking controller construction and training module, which is used to construct a path tracking controller by using a preset neural network, and train the constructed path tracking controller by using the training samples generated by the training sample generation module; wherein, the inputs of the path tracking controller include the initial yaw rate and the reference path information, and the output of the path tracking controller includes the target yaw rate.

[0021] A path tracking control module, which is used to use the path tracking controller trained by the path tracking controller construction and training module to achieve the path tracking control of the controlled object.

[0022] Further, the non-linear model predictive controller construction module is specifically used for:

[0023] Transfer the predictive controller and a section of the forward reference path of the controlled object into the vehicle body coordinate system at the beginning of each control cycle, and establish a non-linear model predictive controller based on the time-varying local model based on the transformed coordinates.

[0024] Further, the path tracking controller construction and training module is specifically used for:

[0025] Construct a path tracking controller by using a preset neural network according to the structure of the training samples;

[0026] Train the constructed path tracking controller by using the training samples.

[0027] Further, the inputs of the path tracking controller include: the initial yaw rate, the horizontal and vertical coordinates and the heading angle of the first reference point in the reference path, the horizontal and vertical coordinates and the heading angle of an intermediate reference point in the reference path, and the horizontal and vertical coordinates and the heading angle of the last reference point in the reference path.

[0028] On the other hand, the present invention also provides an electronic device, which includes a processor and a memory; wherein, at least one instruction is stored in the memory, and the instruction is loaded and executed by the processor to implement the above method.

[0029] On another hand, the present invention also provides a computer-readable storage medium, in which at least one instruction is stored, and the instruction is loaded and executed by the processor to implement the above method.

[0030] The beneficial effects brought by the technical solution provided by the present invention at least include:

[0031] The path tracking control method provided by the present invention deeply analyzes the performance characteristics of a path tracking controller based on nonlinear model predictive control, decouples the lateral and longitudinal coordinates and the heading angle using the local coordinate system of the vehicle body, designs a nonlinear model predictive controller in this coordinate system, and uses this controller to generate training samples. Finally, a neural network-based multi-look-ahead point path tracking controller is established on this basis, solving the problem that the existing path tracking control method of learning a nonlinear model predictive controller through a neural network cannot effectively utilize the information of the forward reference path and has poor accuracy when the curvature of the reference path changes greatly. BRIEF DESCRIPTION OF THE DRAWINGS

[0032] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following will briefly introduce the drawings required for the description of the embodiments. Obviously, the following drawings are only some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can be obtained based on these drawings.

[0033] Figure 1 It is a schematic flowchart of the execution of the neural network-based multi-look-ahead point path tracking control method provided by the embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0034] To make the objectives, technical solutions, and advantages of the present invention clearer, the following will further describe the embodiments of the present invention in detail with reference to the drawings.

[0035] The First Embodiment

[0036] This embodiment provides a neural network-based multi-look-ahead point path tracking control method applicable to the path tracking control of mining articulated vehicles. This neural network-based multi-look-ahead point path tracking control method can be implemented by an electronic device. The execution process of this method is as Figure 1 shown and includes the following steps:

[0037] S1. Establish a nonlinear model predictive controller based on a time-varying local model in the vehicle body coordinate system;

[0038] S2. Through the nonlinear model predictive controller based on the time-varying local model, control the controlled object to track a preset forward reference path, and collect the forward reference path information to generate training samples; wherein, the reference path information includes: the lateral and longitudinal coordinates of the reference path, and the heading angle;

[0039] S3. Construct a path tracking controller using a preset neural network, and train the constructed path tracking controller using the training samples. Among them, the inputs of the path tracking controller include the initial yaw rate and the reference path information, and the output of the path tracking controller includes the target yaw rate.

[0040] S4. Use the trained path tracking controller to achieve path tracking control of the controlled object.

[0041] Among them, the nonlinear model predictive controller based on the time-varying local model is the fitting target of the neural network-based path tracking controller. It should be particularly noted that the input variables of the nonlinear model predictive controller are usually the coordinates and heading information of a section of the reference path ahead and the model input variables of the previous control period, and the output variable is the model input variable of the current control period. In the global coordinate system, the coordinate information of the reference path changes greatly and has weak regularity, making it difficult to be fitted by a neural network.

[0042] Considering that transferring the prediction model and a section of the reference path ahead of the controlled object into the vehicle body coordinate system will not affect the performance of the nonlinear model predictive controller, and since the reference path usually has a limited curvature, the change range of a section of the reference path ahead of the controlled object in the vehicle body coordinate system is small and has strong regularity. Therefore, first, the prediction model and a section of the reference path ahead of the controlled object need to be transferred into the vehicle body coordinate system at the beginning of each control period to establish a nonlinear model predictive controller based on the local time-varying model. Among them, the coordinate transformation process is a simple mathematical process and will not be elaborated here. The design method of the nonlinear model predictive controller is also publicly available knowledge (Bai Guoxing, Liu Li, Meng Yu, etc. Real-time path tracking of mobile robots based on nonlinear model predictive control [J]. Transactions of the Chinese Society for Agricultural Machinery, 2020, 51(9): 47-52), which will not be elaborated here either.

[0043] Furthermore, to generate training samples, first, the input information vector and output information vector of the neural network need to be determined according to the nonlinear model predictive controller. The prediction model of the nonlinear model predictive controller is based on the kinematic model, and it is assumed that the controlled object travels at a constant speed. Therefore, it can be known that the model state variables are \(x = [x\ y\ \theta]\) T , where \(x\) and \(y\) are the horizontal and vertical coordinates, and \(\theta\) is the heading angle (yaw angle). The initial value of each control period is \(x_0 = [0\ 0\ 0]\) T , the model input variable is \(u = [\omega]\) T , where \(\omega\) is the yaw rate. The mapping relationship between the model state variables and the input variables can be expressed as

[0044] Therefore, the input information vector of the neural network is:

[0045] ξ = [ξ1 ξ2 … ξ n T

[0046] = [ω0 x ref1 y ref1 θ ref1 x ref2 y ref2 θ ref2 … x refm y refm θ refm T

[0047] where the subscript ref represents the reference path, the number represents the i-th quantity starting from time t, and in particular, 0 represents the quantity at time t.

[0048] Furthermore, to improve real-time performance, the number of reference path points can be reduced. In this regard, this embodiment takes the first point, an intermediate point, and the last point as the input information of the neural network:

[0049] ξ = [ξ1 ξ2 ξ n T

[0050] = [ω0 x ref1 y ref1 θ ref1 x refq y refq θ refq x refp y refp θ refp T

[0051] where p is the prediction horizon, and q is the ceiling value of p / 2.

[0052] The output information vector of the neural network is ω1.

[0053] Let the nonlinear model predictive controller based on the local time-varying model control the controlled object to track a reference path with a higher complexity, and collect the above information vectors, then training samples can be generated. Then, a neural network model is constructed according to the structure of these information vectors, and the constructed neural network is trained using these information vectors. After the training is completed, a multi-preview-point path tracking controller based on the neural network can be established, thereby realizing the multi-preview-point path tracking control based on the neural network. Among them, the specific technical solution for constructing the neural network model and training the neural network controller can refer to the patent application with the publication number CN111624992B.

[0054] ​​​​In summary, through in-depth analysis of the performance characteristics of the path tracking controller based on nonlinear model predictive control, the above technical solution of this embodiment decouples the lateral and longitudinal coordinates and the heading angle using the local vehicle coordinate system, designs a nonlinear model predictive controller in this coordinate system, and uses this controller to generate training samples. Finally, based on this, a neural network-based multi-preview point path tracking controller is established, which solves the problem that the existing path tracking control method of learning the nonlinear model predictive controller through the neural network cannot effectively utilize the information of the forward reference path and has poor accuracy when the curvature of the reference path changes greatly.

[0055] Second Embodiment

[0056] This embodiment provides a neural network-based multi-preview point path tracking control system, including:

[0057] A nonlinear model predictive controller construction module for establishing a nonlinear model predictive controller based on a time-varying local model in the vehicle coordinate system;

[0058] A training sample generation module for controlling the controlled object to track a preset forward reference path through the nonlinear model predictive controller based on the time-varying local model constructed by the nonlinear model predictive controller construction module, and collecting forward reference path information to generate training samples; wherein, the reference path information includes: the lateral and longitudinal coordinates of the reference path, and the heading angle;

[0059] A path tracking controller construction and training module for constructing a path tracking controller using a preset neural network and training the constructed path tracking controller using the training samples generated by the training sample generation module; wherein, the input of the path tracking controller includes the initial yaw rate and the reference path information, and the output of the path tracking controller includes the target yaw rate;

[0060] A path tracking control module for using the path tracking controller trained by the path tracking controller construction and training module to implement the path tracking control of the controlled object.

[0061] The neural network-based multi-preview point path tracking control system of this embodiment corresponds to the neural network-based multi-preview point path tracking control method of the above first embodiment; wherein, the functions implemented by each functional module in the neural network-based multi-preview point path tracking control system correspond one-to-one to each process step in the above neural network-based multi-preview point path tracking control method; therefore, it will not be elaborated here.

[0062] Third Embodiment

[0063] This embodiment provides an electronic device, which includes a processor and a memory; wherein, at least one instruction is stored in the memory, and the instruction is loaded and executed by the processor to implement the method of the first embodiment.

[0064] The electronic device may have relatively large differences due to different configurations or performances, and may include one or more processors (central processing units, CPUs) and one or more memories. Among them, at least one instruction is stored in the memory, and the instruction is loaded and executed by the processor to implement the above method.

[0065] Fourth Embodiment

[0066] This embodiment provides a computer-readable storage medium, in which at least one instruction is stored, and the instruction is loaded and executed by the processor to implement the method of the first embodiment above. Among them, the computer-readable storage medium may be ROM, random access memory, CD-ROM, magnetic tape, floppy disk, optical data storage device, etc. The instruction stored therein can be loaded and executed by the processor in the terminal to implement the above method.

[0067] In addition, it should be noted that the present invention can be provided as a method, a device, or a computer program product. Therefore, the embodiments of the present invention can take the form of an all-hardware embodiment, an all-software embodiment, or an embodiment combining software and hardware aspects. Moreover, the embodiments of the present invention can take the form of a computer program product implemented on one or more computer-usable storage media containing computer-usable program code.

[0068] The embodiments of the present invention are described with reference to the flowcharts and / or block diagrams of the methods, terminal devices (systems), and computer program products according to the embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, and the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to the processor of a general-purpose computer, an embedded processor, or other programmable data processing terminal devices to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing terminal devices generate a device for implementing the functions specified in Figure 1 one process or multiple processes and / or blocks Figure 1 one block or multiple blocks.

[0069] These computer program instructions can also be stored in a computer-readable memory that can direct a computer or other programmable data processing terminal device to work in a specific manner, so that the instructions stored in the computer-readable memory generate a manufactured product including an instruction device, and the instruction device implements the functions in Figure 1 one process or multiple processes and / or blocksFigure 1 The functions specified in one or more boxes. These computer program instructions can also be loaded onto a computer or other programmable data processing terminal device, so that a series of operation steps are executed on the computer or other programmable terminal device to generate a computer-implemented process. Thus, the instructions executed on the computer or other programmable terminal device provide steps for implementing the functions specified in one or more processes and / or boxes. Figure 1 One or more processes and / or boxes Figure 1 The steps of the functions specified in one or more boxes.

[0070] It should also be noted that in this article, the term "comprising", "including" or any other variant thereof is intended to cover non-exclusive inclusion, so that a process, method, article or terminal device comprising a series of elements not only includes those elements, but also includes other elements not expressly listed, or further includes elements inherent to such process, method, article or terminal device. Without further limitation, an element defined by the statement "comprising one..." does not exclude the presence of additional identical elements in the process, method, article or terminal device comprising the said element.

[0071] Finally, it should be noted that the above is the preferred embodiment of the present invention. It should be pointed out that although the preferred embodiments of the present invention have been described, for those skilled in the art of this technology, once the basic creative concept of the present invention is known, without departing from the principle described in the present invention, several improvements and refinements can still be made, and these improvements and refinements should also be regarded as the protection scope of the present invention. Therefore, the appended claims are intended to be construed as including the preferred embodiments and all changes and modifications falling within the scope of the embodiments of the present invention.

Claims

1. A multi-preview-point path tracking control method based on a neural network, characterized in that, Comprising: Establishing a nonlinear model predictive controller based on a time-varying local model in the vehicle body coordinate system; Controlling a controlled object to track a preset forward reference path through the nonlinear model predictive controller based on the time-varying local model, and collecting forward reference path information to generate training samples; wherein, the reference path information includes: the horizontal and vertical coordinates of the reference path, and the heading angle; Constructing a path tracking controller using a preset neural network, and training the constructed path tracking controller using the training samples; wherein, the input of the path tracking controller includes the initial yaw rate and the reference path information, and the output of the path tracking controller includes the target yaw rate; Implementing path tracking control of the controlled object using the trained path tracking controller; Establishing a nonlinear model predictive controller based on a time-varying local model in the vehicle body coordinate system, including: Transferring the predictive controller and a section of the forward reference path of the controlled object into the vehicle body coordinate system at the beginning of each control cycle, and establishing a nonlinear model predictive controller based on the time-varying local model based on the transformed coordinates; Controlling a controlled object to track a preset forward reference path through the nonlinear model predictive controller based on the time-varying local model, and collecting forward reference path information to generate training samples, including: enabling the nonlinear model predictive controller based on the local time-varying model to control the controlled object to track a reference path with a complexity meeting preset requirements, and collecting an input information vector to generate training samples; wherein, the input information vector is composed of the initial yaw rate, the horizontal and vertical coordinates and the heading angle of the first reference point in the reference path, the horizontal and vertical coordinates and the heading angle of an intermediate reference point in the reference path, and the horizontal and vertical coordinates and the heading angle of the last reference point in the reference path.

2. The multi-preview-point path tracking control method based on a neural network according to claim 1, wherein The constructing a path tracking controller using a preset neural network and training the constructed path tracking controller using the training samples includes: Constructing a path tracking controller using a preset neural network according to the structure of the training samples; Training the constructed path tracking controller using the training samples.

3. A multi-preview-point path tracking control system based on a neural network, characterized in that, Comprising: A nonlinear model predictive controller construction module, configured to establish a nonlinear model predictive controller based on a time-varying local model in the vehicle body coordinate system; A training sample generation module, configured to control a controlled object to track a preset forward reference path through the nonlinear model predictive controller established by the nonlinear model predictive controller construction module, and collect forward reference path information to generate training samples; wherein, the reference path information includes: the horizontal and vertical coordinates of the reference path, and the heading angle; A path tracking controller construction and training module, configured to construct a path tracking controller using a preset neural network, and train the constructed path tracking controller using the training samples generated by the training sample generation module; wherein, the input of the path tracking controller includes the initial yaw rate and the reference path information, and the output of the path tracking controller includes the target yaw rate; A path tracking control module, which is used to implement path tracking control of a controlled object by using the path tracking controller trained by the path tracking controller construction and training module; The non-linear model predictive controller construction module is specifically used for: Transferring the predictive controller and a section of the reference path in front of the controlled object into the vehicle body coordinate system at the beginning of each control cycle, and establishing a non-linear model predictive controller based on a time-varying local model based on the transformed coordinates; Through the non-linear model predictive controller based on the time-varying local model, controlling the controlled object to track a preset forward reference path, and collecting forward reference path information to generate training samples, including: making the non-linear model predictive controller based on the local time-varying model control the controlled object to track a reference path with a complexity meeting the preset requirements, and collecting the input information vector to generate training samples; wherein, the input information vector is composed of the initial yaw rate, the horizontal and vertical coordinates and the heading angle of the first reference point in the reference path, the horizontal and vertical coordinates and the heading angle of an intermediate reference point in the reference path, and the horizontal and vertical coordinates and the heading angle of the last reference point in the reference path.

4. The multi-preview-point path tracking control system based on a neural network according to claim 3, characterized in that, The path tracking controller construction and training module is specifically used for: Constructing a path tracking controller by using a preset neural network according to the structure of the training samples; Training the constructed path tracking controller by using the training samples.

Images