A vehicle path tracking control method and system based on a preset state space NMPC

Abstract

The application provides a vehicle path tracking control method and system based on a preset state space NMPC, relates to the motion control technical field of unmanned vehicles, and comprises the following steps: a preset state space NMPC framework is established, a real-time path tracking controller layer is set, a plurality of real-time path tracking controllers are contained, path information and vehicle pose information of the vehicle are input into the real-time path tracking controller, and feasible control input is solved; a prediction model layer predicts future pose states of the vehicle according to each feasible control input, and a plurality of groups of pose states form a state space; and an optimization objective function layer finds a group of input closest to the reference path in the state space and transmits the input to the controlled vehicle, so that the real-time performance of path tracking control is improved under the premise of ensuring the accuracy of path tracking control. The application is derived from traditional NMPC control, but is quite different from the state space; and the connection part of the prediction model and the optimization objective function and the state space is also adaptively improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the motion control technical field of unmanned vehicles, and particularly relates to a vehicle path tracking control method and system based on a preset state space NMPC. BACKGROUND

[0002] In the field of path tracking control of unmanned vehicles, there are currently various path tracking control methods, such as pure tracking control, Stanley control, sliding mode control, model predictive control (MPC), and the like. Compared with other control methods, MPC has advantages such as being able to explicitly handle constraints and directly consider the dynamics characteristics. At the same time, there are various forms of MPC control methods, among which nonlinear model predictive control (NMPC) not only has the advantages of general MPC, but also has the advantage of being able to effectively utilize the reference path information in front of the vehicle, thereby being able to achieve high-precision path tracking in the case of a large reference path curvature and a large curvature change amplitude.

[0003] However, NMPC has poor real-time performance, and current real-time optimization schemes still have some shortcomings. For example, optimizing computer code (KAYACAN E, KAYACAN E, RAMON H, et al. Learning in centralized nonlinear model predictive control: Application to an autonomous tractor-trailer system[J]. IEEE Transactions on Control Systems Technology, 2014, 23(1):197-205), replacing the solution algorithm (SHEN C, BUCKHAM B, SHI Y. Modified C / GMRES algorithm for fast nonlinear model predictive tracking control of AUVs[J]. IEEE Transactions on Control Systems Technology, 2016, 25(5):1896-1904; GUO N, ZHANG X, ZOU Y, et al. A computationally efficient path-following control strategy of autonomous electric vehicles with yaw motion stabilization[J]. IEEE Transactions on Transportation Electrification, 2020, 6(2): 728-739; Wang Hongwei, Liu Chenyu, Li Lei, et al. Research on trajectory tracking control of unmanned vehicles based on efficient NMPC algorithm [J]. Automotive Engineering, 2022, 44(10): 1494-1502, 1618), adjusting the prediction time domain parameters (Bai Guoxing, Liu Li, Meng Yu, et al. 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; BAI G, MENG Y, GU Q, et al. Some rules for setting the horizon parameters of NMPC-based vehicle path tracking [C] / / 2020 Chinese Automation Congress (CAC). IEEE, 2020: 5167-5170) and other schemes still need online optimization, and the real-time performance improvement effect is limited.The chip hardware solution (GUO H, LIU F, XU F, et al. Nonlinear model predictive lateral stability control of active chassis for intelligent vehicles and its FPGA implementation[J].IEEE Transactions on Systems, Man, and Cybernetics: Systems, 2017, 49(1): 2-13) is too expensive and has poor applicability. Offline optimization solutions (TAVERNINI D, METZLER M, GRUBER P, et al. Explicit nonlinear model predictive control for electric vehicle traction control[J]. IEEE Transactions on Control Systems Technology, 2018, 27(4): 1438-1451) and solutions using neural networks to learn NMPC (BAI G, MENG Y, LIU L, et al. Path tracking for car-like robots based on neural networks with NMPC as learning samples[J]. Electronics, 2022, 11(24): 4232;CN111624992B;CN115576317A) cannot guarantee the accuracy of path tracking control, i.e., they cannot simultaneously guarantee accuracy and real-time performance.

[0004] In summary, NMPC-based vehicle path tracking control methods have many advantages, but their poor real-time performance remains a problem that is difficult to solve effectively. Summary of the Invention

[0005] This invention provides a vehicle path tracking control method and system based on preset state space NMPC, which solves the problem of poor real-time performance of vehicle path tracking control based on NMPC in the prior art.

[0006] To achieve the aforementioned objectives, the present invention provides the following technical solution: a vehicle path tracking control method based on preset state space NMPC, characterized in that the steps include:

[0007] S1. Obtain the vehicle's path information, vehicle pose information, and reference path;

[0008] S2. Preset state-space NMPC framework; the framework is divided into three layers, including: real-time path tracking control layer, prediction model layer and optimization objective function layer;

[0009] S3. Set up several real-time path tracking controllers in the real-time path tracking layer, input the vehicle's path information and vehicle pose information to the real-time path tracking controllers, calculate the control input, and form an input space;

[0010] S4. Substitute the control inputs in the input space into the prediction model layer for iteration to predict the future pose of the vehicle and form a state space.

[0011] S5. Substitute the pose states in the state space into the optimization objective function layer, find the set of pose states in the state space that is closest to the reference path, and transmit the control input corresponding to the set of pose states that is closest to the reference path to the controlled vehicle, thus completing the vehicle path tracking control based on the preset state space NMPC.

[0012] Preferably, in step S3, the vehicle's path information and vehicle pose information are input to the real-time path tracking controller to calculate the control input, including:

[0013] Find the point closest to the vehicle in the reference path as the tracking target point;

[0014] To retain the advantage of the NMPC controller in responding to the reference path information ahead of the vehicle, a preview information is introduced into the real-time path tracking controller. A specific point on the reference path 0.5m to 5m ahead of the tracking target point is selected as the preview point, and the control input is calculated by real-time control law based on the preview point pose information and the vehicle pose information.

[0015] Preferably, the reference path arc length between the pre-aiming point and the tracking target point is the pre-aiming distance.

[0016] Preferably, in step S3, the real-time path tracking control module needs to have n real-time path tracking controllers with different parameters, where n is a positive integer greater than 1.

[0017] Preferably, the n real-time path tracking controllers with different parameters can be operated serially or in parallel; the operation results are stored as an array, serving as the input space for the NMPC prediction model.

[0018] δ Si =f(l i ), i = 1, 2, ..., n (1)

[0019] Where l represents the aiming distance, that is, the input space generated by the real-time path tracking control law takes the aiming distance as the independent variable.

[0020] Preferably, in step S4, the control inputs in the input space are substituted into the prediction model layer for iteration to predict the future pose state of the vehicle, forming a state space, including:

[0021] Based on the vehicle dynamics model, the preliminary model is shown in the following formula (2):

[0022]

[0023] Where x represents the state variable and u represents the input variable, and its element is usually δ;

[0024] Based on the vehicle dynamics model or kinematics model, the prediction model is discretized to obtain the prediction model shown in the following formula (3):

[0025] x(j+1)=x(j)+Tf(x(j),u),j=1,2,…,p (3)

[0026] Where T is the iteration period and p is the prediction time domain;

[0027] Each feasible input δ in the input space Si Substituting the values ​​into the prediction model and iterating, n sets of possible future vehicle pose states are generated.

[0028] The prediction model module generates n sets of possible future vehicle pose states, which are the state spaces in the preset state space NMPC framework.

[0029] Preferably, in step S5, the pose states in the state space are substituted into the optimization objective function layer to find the set of pose states in the state space that are closest to the reference path. The control input corresponding to the closest set of pose states is transmitted to the controlled vehicle to complete the vehicle path tracking control based on the preset state space NMPC, including:

[0030] Find the point closest to the vehicle in the reference path point list and use it as the first point for the reference value of the optimization objective function;

[0031] Along the vehicle's direction of travel, take p-1 points on the reference path as the remaining points for optimizing the objective function reference value. The arc length of the reference path between any two points is equal to the product of the vehicle speed and the control cycle.

[0032] Take the vehicle pose state information in the state space and subtract it from the pose state of the reference point sequence to form a penalty term;

[0033] The objective function is then expressed as shown in formula (4) below:

[0034]

[0035] Where, δ SiThe variable represents the independent variable, Q represents the weighting coefficient, and the subscript Ref represents the reference value.

[0036] Using the independent variables in the input space as the independent variables of the optimization objective function, the optimization objective function values ​​corresponding to different independent variables are calculated. The independent variable corresponding to the minimum value is taken as the final control input and passed to the controlled vehicle to complete the vehicle path tracking control based on the preset state space NMPC.

[0037] A vehicle path tracking control system based on preset state space NMPC is provided. The system is used in the aforementioned vehicle path tracking control method based on preset state space NMPC. The system includes:

[0038] The information acquisition unit is used to acquire the vehicle's path information, vehicle pose information, and information from the reference path;

[0039] The framework preset unit is used to preset the state space NMPC framework; the framework is divided into three layers, including: real-time path tracking control layer, prediction model layer and optimization objective function layer;

[0040] A real-time path tracking unit is used to set up several real-time path tracking controllers in the real-time path tracking layer, input the vehicle's path information and vehicle pose information to the real-time path tracking controllers, calculate the control input, and form an input space.

[0041] The prediction unit is used to substitute the control input in the input space into the prediction model layer for iteration to predict the future pose state of the vehicle and form a state space.

[0042] The tracking control unit is used to substitute the pose states in the state space into the optimization objective function layer, find the set of pose states in the state space that are closest to the reference path, and transmit the control input corresponding to the set of pose states that are closest to the reference path to the controlled vehicle, thereby completing the vehicle path tracking control based on the preset state space NMPC.

[0043] Preferably, the real-time path tracking unit is further used to find the point closest to the vehicle in the reference path as the tracking target point;

[0044] To retain the advantage of the NMPC controller in responding to the reference path information ahead of the vehicle, a preview information is introduced into the real-time path tracking controller. A specific point on the reference path 0.5m to 5m ahead of the tracking target point is selected as the preview point, and the control input is calculated by real-time control law based on the preview point pose information and the vehicle pose information.

[0045] Preferably, the reference path arc length between the pre-aiming point and the tracking target point is the pre-aiming distance.

[0046] On the one hand, an electronic device is provided, the electronic device including a processor and a memory, the memory storing at least one instruction, the at least one instruction being loaded and executed by the processor to implement the above-described vehicle path tracking control method based on preset state space NMPC.

[0047] On the one hand, a computer-readable storage medium is provided, wherein at least one instruction is stored in the storage medium, and the at least one instruction is loaded and executed by a processor to implement the above-described vehicle path tracking control method based on preset state space NMPC.

[0048] The above technical solution has at least the following advantages compared with the existing technology:

[0049] The above scheme adopts a predictive control approach. First, a pre-defined state-space NMPC framework is established. Then, a real-time path tracking control module, a predictive model module, and an optimization objective function module are integrated. This improves the real-time performance of path tracking control while ensuring its accuracy. This patent is derived from traditional NMPC control, but it differs significantly in its state space (also known as the search space). Furthermore, the connection between the predictive model and the optimization objective function and the state space has been significantly improved. Attached Figure Description

[0050] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0051] Figure 1 This is a schematic flowchart of a vehicle path tracking control method based on a preset state space NMPC provided in an embodiment of the present invention;

[0052] Figure 2 This is a detailed flowchart of the vehicle path tracking control method based on preset state space NMPC provided in the embodiments of the present invention;

[0053] Figure 3 This is a block diagram of a vehicle path tracking control system based on a preset state space NMPC provided in an embodiment of the present invention;

[0054] Figure 4 This is a schematic diagram of the structure of an electronic device provided in an embodiment of the present invention. Detailed Implementation

[0055] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. All other embodiments obtained by those skilled in the art based on the described embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0056] This invention addresses the problem of poor real-time performance in existing NMPC-based vehicle path tracking control by providing a vehicle path tracking control method and system based on preset state space NMPC.

[0057] like Figure 1 As shown, this embodiment of the invention provides a vehicle path tracking control method based on a preset state space NMPC, which can be implemented by an electronic device. Figure 1 The flowchart shown is for a vehicle path tracking control method based on a preset state space NMPC. The processing flow of this method may include the following steps:

[0058] S101. Obtain the vehicle's path information, vehicle pose information, and reference path;

[0059] S102, Preset state-space NMPC framework; the framework is divided into three layers, including: real-time path tracking control layer, prediction model layer and optimization objective function layer.

[0060] In one feasible implementation, the pre-defined state-space NMPC framework is the core invention of this patent. This framework divides the controller into three layers: the first layer is a real-time path tracking control module, the second layer is a prediction model module, and the third layer is an optimization objective function module. For example... Figure 2 The flowchart shown is a detailed process for a vehicle path tracking control method based on a preset state space NMPC. The real-time path tracking control module requires n real-time path tracking controllers with different parameters, where n is typically a positive integer around 10. These n controllers can calculate n different control inputs based on the reference path information and vehicle pose information, forming an input space. The predictive model module iterates through the n different control inputs from the input space to predict the vehicle's n possible pose states in the future, forming a state space. In other words, the real-time path tracking control module and the predictive model module combine to complete the preset state space function. The optimization objective function aims to find the set of n possible pose states that is closest to the reference path. Finally, the control inputs corresponding to this set of pose states are used as the actual control inputs to the controlled vehicle, thus achieving NMPC control within the preset state space.

[0061] S103. Set up several real-time path tracking controllers in the real-time path tracking layer, input the vehicle's path information and vehicle pose information to the real-time path tracking controllers, calculate the control input, and form an input space.

[0062] In one feasible implementation, the real-time path tracking control module is an essential module of this patent, and the method of connecting the real-time path tracking control module to the preset state space NMPC framework is the invention content of this patent.

[0063] In the predefined state-space NMPC framework, the inputs to the real-time path tracking control module are the reference path information input from the path planning system and the vehicle pose information input from the positioning system. The reference path information can typically be given in the form of a list of points.

[0064] In one feasible implementation, in step S103, the vehicle's path information and vehicle pose information are input to the real-time path tracking controller to calculate the control input, including:

[0065] Find the point closest to the vehicle in the reference path as the tracking target point;

[0066] The NMPC controller can respond to reference path information ahead of the vehicle. To retain this advantage, the real-time path tracking controller needs to respond to aiming information. A specific point on the reference path 0.5m to 5m ahead of the target point is selected as the aiming point, and the control input is calculated through a real-time control law based on the aiming point pose information and the vehicle pose information.

[0067] In one feasible implementation, the reference path arc length between the aiming point and the tracking target point is the aiming distance.

[0068] In one feasible implementation, after selecting the target point, considering that the NMPC controller needs to respond to the reference path information ahead of the vehicle, the real-time path tracking control law must be able to take into account the reference path information ahead of the vehicle, i.e., the pre-aiming information. Therefore, a specific point on the reference path a certain distance ahead of the target point can be selected as the pre-aiming point. Based on the pose information of the pre-aiming point and the vehicle pose information, the feasible control input is calculated through the real-time control law. The arc length of the reference path between the pre-aiming point and the target point is the pre-aiming distance.

[0069] This method requires that the real-time control law must be a real-time control law that does not require online optimization and solution. Therefore, pure tracking control, Stanley control, LQR (Linear Quadratic Regulator) and other methods can be selected as the design basis for the real-time path tracking control module.

[0070] In one feasible implementation, in step S3, the real-time path tracking control module needs to have n real-time path tracking controllers with different parameters, where n is a positive integer greater than 1. Preferably, it can be set to 10.

[0071] In one feasible implementation, n real-time path tracking controllers with different parameters can be operated serially or in parallel; the operation results are stored as an array, serving as the input space for the NMPC prediction model.

[0072] δ Si =f(l i ), i = 1, 2, ..., n (1)

[0073] Here, l represents the aiming distance, meaning the input space generated by the real-time path tracking control law uses the aiming distance as the independent variable. The independent variable can also be the adjustment parameters of the control law. The independent variable can be a two-dimensional table composed of the aiming distance and the adjustment parameters.

[0074] S104. Substitute the control inputs in the input space into the prediction model layer for iteration to predict the future pose of the vehicle and form a state space.

[0075] In one feasible implementation, within the preset state space NMPC framework, the input to the prediction model module is the input space generated by the real-time path tracking control module, and each feasible input in the input space is substituted into the prediction model.

[0076] For the prediction model, in order to consider the influence of dynamic state variables such as lateral velocity on path tracking control, the prediction model in the prediction model module is generally based on the vehicle dynamics model. If the influence of dynamic state variables such as lateral velocity is not considered, the vehicle kinematics model can also be used as the basis.

[0077] In one feasible implementation, in step S104, the control inputs in the input space are substituted into the prediction model layer for iteration to predict the future pose state of the vehicle, forming a state space, including:

[0078] Based on the vehicle dynamics model, the preliminary model is shown in the following formula (2):

[0079]

[0080] Where x represents the state variable and u represents the input variable, and its element is usually δ;

[0081] Based on the vehicle dynamics model or kinematics model, the prediction model is discretized to obtain the prediction model shown in the following formula (3):

[0082] x(j+1)=x(j)+Tf(x(j),u),j=1,2,…,p (3)

[0083] Where T is the iteration period and p is the prediction time domain;

[0084] Each feasible input δ in the input space Si Substituting the values ​​into the prediction model and iterating, n sets of possible future vehicle pose states are generated.

[0085] The prediction model module generates n sets of possible future vehicle pose states, which are the state spaces in the preset state space NMPC framework.

[0086] In one feasible implementation, the discretization process can employ either the Newton-Euler method or the Runge-Kutta method.

[0087] S105. Substitute the pose states in the state space into the optimization objective function layer, find the set of pose states in the state space that is closest to the reference path, and transmit the control input corresponding to the set of pose states that is closest to the reference path to the controlled vehicle, thereby completing the vehicle path tracking control based on the preset state space NMPC.

[0088] In one feasible implementation, the objective function optimization module is an essential module of this patent, and the method of connecting the objective function optimization module to the preset state space NMPC framework is the invention content of this patent.

[0089] In the pre-defined state-space NMPC framework, the input to the objective function module is the state space generated by the prediction model module.

[0090] In one feasible implementation, in step S105, the pose states in the state space are substituted into the optimization objective function layer to find the set of pose states in the state space that are closest to the reference path. The control input corresponding to the closest set of pose states is transmitted to the controlled vehicle to complete the vehicle path tracking control based on the preset state space NMPC, including:

[0091] Find the point closest to the vehicle in the reference path point list and use it as the first point for the reference value of the optimization objective function;

[0092] Along the vehicle's direction of travel, take p-1 points on the reference path as the remaining points for optimizing the objective function reference value. The arc length of the reference path between any two points is equal to the product of the vehicle speed and the control cycle.

[0093] Take the vehicle pose state information in the state space and subtract it from the pose state of the reference point sequence to form a penalty term;

[0094] The objective function is then expressed as shown in formula (4) below:

[0095]

[0096] Where, δ Si The variable represents the independent variable, Q represents the weighting coefficient, and the subscript Ref represents the reference value.

[0097] Using the independent variables in the input space as the independent variables of the optimization objective function, the optimization objective function values ​​corresponding to different independent variables are calculated. The independent variable corresponding to the minimum value is taken as the final control input and passed to the controlled vehicle to complete the vehicle path tracking control based on the preset state space NMPC.

[0098] In one feasible implementation, if only the accuracy of path tracking control is considered, the penalty terms are the horizontal coordinate deviation penalty term, the vertical coordinate deviation penalty term, and the heading angle deviation penalty term. In particular, if vehicle driving stability is considered, optimization target penalty terms such as lateral speed penalty term and center of gravity sideslip angle penalty term can usually be added.

[0099] In this embodiment of the invention, a predictive control approach is adopted. First, a pre-defined state-space NMPC framework is established. Then, a real-time path tracking control module, a predictive model module, and an optimization objective function module are integrated. This improves the real-time performance of path tracking control while ensuring its accuracy. This patent is derived from traditional NMPC control, but it differs significantly in its state space (also known as the search space). Furthermore, the connection between the predictive model and the optimization objective function and the state space has been significantly improved. These are the main inventive contents of this patent.

[0100] Figure 3 This is a schematic diagram of a vehicle path tracking control system based on a preset state space NMPC according to the present invention. The system 200 is used for the above-mentioned vehicle path tracking control based on the preset state space NMPC, and the system 200 includes:

[0101] The information acquisition unit 210 is used to acquire the vehicle's path information, vehicle pose information, and reference path;

[0102] The framework preset unit 220 is used to preset the state space NMPC framework; the framework is divided into three layers, including: a real-time path tracking control layer, a prediction model layer, and an optimization objective function layer;

[0103] The real-time path tracking unit 230 is used to set up a plurality of real-time path tracking controllers in the real-time path tracking layer, input the path information and vehicle pose information of the vehicle to the real-time path tracking controller, calculate the control input, and form an input space.

[0104] Prediction unit 240 is used to substitute the control input in the input space into the prediction model layer for iteration, predict the future pose state of the vehicle, and form a state space.

[0105] The tracking control unit 250 is used to substitute the pose states in the state space into the optimization objective function layer, find the set of pose states in the state space that are closest to the reference path, and transmit the control input corresponding to the closest set of pose states to the controlled vehicle to complete the vehicle path tracking control based on the preset state space NMPC.

[0106] Preferably, the real-time path tracking unit 230 is further used to find the point closest to the vehicle in the reference path as the tracking target point;

[0107] To retain the advantage of the NMPC controller in responding to the reference path information ahead of the vehicle, a preview information is introduced into the real-time path tracking controller. A specific point on the reference path 0.5m to 5m ahead of the tracking target point is selected as the preview point, and the control input is calculated by real-time control law based on the preview point pose information and the vehicle pose information.

[0108] Preferably, the reference path arc length between the pre-aiming point and the tracking target point is the pre-aiming distance.

[0109] Preferably, the real-time path tracking unit 230 is further used in a real-time path tracking control module that requires n different parameters for the real-time path tracking controller, where n is a positive integer greater than 1.

[0110] Preferably, the n real-time path tracking controllers with different parameters can be operated serially or in parallel; the operation results are stored as an array, serving as the input space for the NMPC prediction model.

[0111] δ Si =f(l i ), i = 1, 2, ..., n (1)

[0112] Where l represents the aiming distance, that is, the input space generated by the real-time path tracking control law takes the aiming distance as the independent variable.

[0113] Preferably, the prediction unit 240 is further used to base a preliminary model on a vehicle dynamics model, as shown in the following formula (2):

[0114]

[0115] Where x represents the state variable and u represents the input variable, and its element is usually δ;

[0116] Based on the vehicle dynamics model or kinematics model, the prediction model is discretized to obtain the prediction model shown in the following formula (3):

[0117] x(j+1)=x(j)+Tf(x(j),u),j=1,2,…,p (3)

[0118] Where T is the iteration period and p is the prediction time domain;

[0119] Each feasible input δ in the input space Si Substituting the values ​​into the prediction model and iterating, n sets of possible future vehicle pose states are generated.

[0120] The prediction model module generates n sets of possible future vehicle pose states, which are the state spaces in the preset state space NMPC framework.

[0121] Preferably, the tracking control unit 250 is further configured to find the point closest to the vehicle in the reference path point column as the first point for optimizing the objective function reference value;

[0122] Along the vehicle's direction of travel, take p-1 points on the reference path as the remaining points for optimizing the objective function reference value. The arc length of the reference path between any two points is equal to the product of the vehicle speed and the control cycle.

[0123] Take the vehicle pose state information in the state space and subtract it from the pose state of the reference point sequence to form a penalty term;

[0124] The objective function is then expressed as shown in formula (4) below:

[0125]

[0126] Where, δ Si The variable represents the independent variable, Q represents the weighting coefficient, and the subscript Ref represents the reference value.

[0127] Using the independent variables in the input space as the independent variables of the optimization objective function, the optimization objective function values ​​corresponding to different independent variables are calculated. The independent variable corresponding to the minimum value is taken as the final control input and passed to the controlled vehicle to complete the vehicle path tracking control based on the preset state space NMPC.

[0128] In this embodiment of the invention, a predictive control approach is adopted. First, a pre-defined state-space NMPC framework is established, and then a real-time path tracking control module, a predictive model module, and an optimization objective function module are integrated. This improves the real-time performance of path tracking control while ensuring its accuracy. This patent is derived from traditional NMPC control, but there are significant differences in the state space (also known as the search space). Furthermore, the connection between the predictive model and the optimization objective function and the state space has been significantly improved.

[0129] Figure 4 This is a schematic diagram of the structure of an electronic device 300 provided in an embodiment of the present invention. The electronic device 300 can vary considerably due to differences in configuration or performance. It may include one or more central processing units (CPUs) 301 and one or more memories 302. The memories 302 store at least one instruction, which is loaded and executed by the processors 301 to implement the steps of the following vehicle path tracking control method based on a preset state space NMPC:

[0130] S1. Obtain the vehicle's path information, vehicle pose information, and reference path;

[0131] S2. Preset state-space NMPC framework; the framework is divided into three layers, including: real-time path tracking control layer, prediction model layer and optimization objective function layer;

[0132] S3. Set up several real-time path tracking controllers in the real-time path tracking layer, input the vehicle's path information and vehicle pose information to the real-time path tracking controllers, calculate the control input, and form an input space;

[0133] S4. Substitute the control inputs in the input space into the prediction model layer for iteration to predict the future pose of the vehicle and form a state space.

[0134] S5. Substitute the pose states in the state space into the optimization objective function layer, find the set of pose states in the state space that is closest to the reference path, and transmit the control input corresponding to the set of pose states that is closest to the reference path to the controlled vehicle, thus completing the vehicle path tracking control based on the preset state space NMPC.

[0135] In an exemplary embodiment, a computer-readable storage medium is also provided, such as a memory including instructions that can be executed by a processor in a terminal to complete the vehicle path tracking control method based on a preset state space NMPC. For example, the computer-readable storage medium may be a ROM, random access memory (RAM), CD-ROM, magnetic tape, floppy disk, or optical data storage device.

[0136] Those skilled in the art will understand that all or part of the steps of the above embodiments can be implemented by hardware or by a program instructing related hardware. The program can be stored in a computer-readable storage medium, such as a read-only memory, a disk, or an optical disk.

Claims

1. A vehicle path tracking control method based on preset state space NMPC, characterized in that, The method steps include: S1. Obtain the vehicle's path information, vehicle pose information, and reference path; S2. Preset state-space NMPC framework; the framework is divided into three layers, including: real-time path tracking control layer, prediction model layer and optimization objective function layer; S3. Several real-time path tracking controllers are set in the real-time path tracking layer. These real-time path tracking controllers calculate a finite number of control inputs based on the reference path information and vehicle pose information to form an input space. The real-time path tracing control layer requires n real-time path tracing controllers with different parameters, where n is a positive integer greater than 1. Real-time path tracking controllers with n different parameters can operate serially or in parallel. Based on reference path information and vehicle pose information, the n controllers can calculate n different control inputs, and the calculation results are stored as an array, serving as the input space for the NMPC prediction model. （1） Where l represents the aiming distance, and the input space generated by the real-time path tracking control law takes the aiming distance as the independent variable; S4. Substitute the n different control inputs in the input space into the prediction model layer for iteration to predict the n possible pose states of the vehicle in the future, forming a state space. S5. Substitute the pose states in the n possible state spaces into the optimization objective function layer, find the set of pose states in the state space that is closest to the reference path, and transmit the control input corresponding to the closest set of pose states to the controlled vehicle to complete the vehicle path tracking control based on the preset state space NMPC.

2. The method according to claim 1, characterized in that, In step S3, the number n of the real-time path tracking controllers is 10.

3. The method according to claim 1, characterized in that, In step S3, the vehicle's path information and vehicle pose information are input to the real-time path tracking controller to calculate the control input, including: Find the point closest to the vehicle in the reference path as the tracking target point; To retain the advantage of the NMPC controller in responding to the reference path information ahead of the vehicle, a preview information is introduced into the real-time path tracking controller. A specific point on the reference path 0.5m to 5m ahead of the tracking target point is selected as the preview point, and the control input is calculated by real-time control law based on the preview point pose information and the vehicle pose information.

4. The method according to claim 3, characterized in that, The reference path arc length between the pre-aiming point and the tracking target point is the pre-aiming distance.

5. The method according to claim 4, characterized in that, In step S4, the control inputs in the input space are substituted into the prediction model layer for iteration to predict the future pose state of the vehicle, forming a state space, including: Based on the vehicle dynamics model, the preliminary model is shown in the following formula (2): （2） in, Represents state variables. Indicates the input quantity. The elements are usually ; Based on the vehicle dynamics model or kinematics model, the prediction model is discretized to obtain the prediction model shown in the following formula (3): （3） in, For the iteration period, For prediction in the time domain; Each feasible input in the input space Substituting the values ​​into the prediction model and iterating, n sets of possible future vehicle pose states are generated. The prediction model module generates n sets of possible future vehicle pose states, which are the state spaces in the preset state space NMPC framework.

6. The method according to claim 5, characterized in that, In step S5, the pose states in the state space are substituted into the optimization objective function layer to find the set of pose states in the state space that are closest to the reference path. The control input corresponding to the closest set of pose states is transmitted to the controlled vehicle to complete the vehicle path tracking control based on the preset state space NMPC, including: Find the point closest to the vehicle in the reference path point list and use it as the first point for the reference value of the optimization objective function; Along the vehicle's direction of travel, take p-1 points on the reference path as the remaining points for optimizing the objective function reference value. The arc length of the reference path between any two points is equal to the product of the vehicle speed and the control cycle. Take the vehicle pose state information in the state space and subtract it from the pose state of the reference point sequence to form a penalty term; The objective function is then expressed as shown in formula (4) below: （4） in, Indicates the independent variable. This indicates the weighting coefficient, and the subscript Ref indicates the reference value; Using the independent variables in the input space as the independent variables of the optimization objective function, the optimization objective function values ​​corresponding to different independent variables are calculated. The independent variable corresponding to the minimum value is taken as the final control input and passed to the controlled vehicle to complete the vehicle path tracking control based on the preset state space NMPC.

7. A vehicle path tracking control system based on a preset state space NMPC, characterized in that, The system is used in the vehicle path tracking control method based on preset state space NMPC as described in any one of claims 1 to 6, and the system includes: The information acquisition unit is used to acquire the vehicle's path information, vehicle pose information, and information from the reference path; The framework preset unit is used to preset the state space NMPC framework; the framework is divided into three layers, including: real-time path tracking control layer, prediction model layer and optimization objective function layer; A real-time path tracking unit is used to set up several real-time path tracking controllers in the real-time path tracking layer, input the vehicle's path information and vehicle pose information to the real-time path tracking controllers, calculate the control input, and form an input space. The prediction unit is used to substitute the control input in the input space into the prediction model layer for iteration to predict the future pose state of the vehicle and form a state space. The tracking control unit is used to substitute the pose states in the state space into the optimization objective function layer, find the set of pose states in the state space that are closest to the reference path, and transmit the control input corresponding to the set of pose states that are closest to the reference path to the controlled vehicle, thereby completing the vehicle path tracking control based on the preset state space NMPC.

8. The system according to claim 7, characterized in that, The real-time path tracking unit is further used to find the point closest to the vehicle in the reference path as the tracking target point; To retain the advantage of the NMPC controller in responding to the reference path information ahead of the vehicle, a preview information is introduced into the real-time path tracking controller. A specific point on the reference path 0.5m to 5m ahead of the tracking target point is selected as the preview point, and the control input is calculated by real-time control law based on the preview point pose information and the vehicle pose information.

9. The system according to claim 8, characterized in that, The reference path arc length between the pre-aiming point and the tracking target point is the pre-aiming distance.

Images