CN113325694B - Model prediction control parameter setting method based on machine learning - Google Patents
Model prediction control parameter setting method based on machine learning Download PDFInfo
- Publication number
- CN113325694B CN113325694B CN202110580904.6A CN202110580904A CN113325694B CN 113325694 B CN113325694 B CN 113325694B CN 202110580904 A CN202110580904 A CN 202110580904A CN 113325694 B CN113325694 B CN 113325694B
- Authority
- CN
- China
- Prior art keywords
- robot
- neural network
- control parameters
- control
- path
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
- 238000000034 method Methods 0.000 title claims abstract description 28
- 238000010801 machine learning Methods 0.000 title claims abstract description 26
- 238000013528 artificial neural network Methods 0.000 claims abstract description 48
- 238000012549 training Methods 0.000 claims abstract description 18
- 238000002474 experimental method Methods 0.000 claims abstract description 17
- 230000010355 oscillation Effects 0.000 claims description 19
- 238000005457 optimization Methods 0.000 claims description 9
- 230000006870 function Effects 0.000 claims description 5
- 230000000694 effects Effects 0.000 description 9
- 238000013461 design Methods 0.000 description 4
- 238000004836 empirical method Methods 0.000 description 4
- 238000013507 mapping Methods 0.000 description 4
- 238000012937 correction Methods 0.000 description 3
- 238000010586 diagram Methods 0.000 description 3
- 238000005516 engineering process Methods 0.000 description 3
- 230000008569 process Effects 0.000 description 3
- 230000003044 adaptive effect Effects 0.000 description 2
- 238000011161 development Methods 0.000 description 2
- 238000011160 research Methods 0.000 description 2
- 238000005096 rolling process Methods 0.000 description 2
- 241000219000 Populus Species 0.000 description 1
- 238000004458 analytical method Methods 0.000 description 1
- 238000013459 approach Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000004364 calculation method Methods 0.000 description 1
- 238000006555 catalytic reaction Methods 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 150000001875 compounds Chemical class 0.000 description 1
- 230000003750 conditioning effect Effects 0.000 description 1
- 238000011217 control strategy Methods 0.000 description 1
- 238000010891 electric arc Methods 0.000 description 1
- 238000011156 evaluation Methods 0.000 description 1
- 238000009776 industrial production Methods 0.000 description 1
- 238000011089 mechanical engineering Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000003068 static effect Effects 0.000 description 1
- 238000012546 transfer Methods 0.000 description 1
- 230000009466 transformation Effects 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G05—CONTROLLING; REGULATING
- G05B—CONTROL OR REGULATING SYSTEMS IN GENERAL; FUNCTIONAL ELEMENTS OF SUCH SYSTEMS; MONITORING OR TESTING ARRANGEMENTS FOR SUCH SYSTEMS OR ELEMENTS
- G05B11/00—Automatic controllers
- G05B11/01—Automatic controllers electric
- G05B11/36—Automatic controllers electric with provision for obtaining particular characteristics, e.g. proportional, integral, differential
- G05B11/42—Automatic controllers electric with provision for obtaining particular characteristics, e.g. proportional, integral, differential for obtaining a characteristic which is both proportional and time-dependent, e.g. P. I., P. I. D.
Landscapes
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Engineering & Computer Science (AREA)
- Automation & Control Theory (AREA)
- Feedback Control In General (AREA)
- Manipulator (AREA)
Abstract
The invention discloses a model prediction control parameter setting method based on machine learning, and belongs to the field of mechanical learning. The invention comprises the following steps: 1) Determining the structure of the BP neural network; 2) Selecting a series of control parameters as output data of the BP neural network training sample; 3) Inputting the control parameters into a model prediction controller of the robot, and acquiring corresponding performance indexes in a path tracking experiment as input data of a BP neural network training sample; 4) Training a BP neural network by using a training sample; 5) The set performance index is used as the input of a BP neural network and is output to obtain a control parameter, and the setting of the model control parameter is realized; 6) And inputting the set control parameters into a model prediction controller of the robot, collecting performance indexes generated by a path tracking experiment, comparing the performance indexes with preset performance indexes, and judging whether the set control parameters meet requirements or not. The method and the device are obviously improved in the rapidity and effectiveness of parameter setting.
Description
Technical Field
The invention belongs to the field of mechanical learning, and particularly relates to a model prediction control parameter setting method based on machine learning.
Background
The mobile robot can automatically collect surrounding environment information, plan a path, and control the robot to track a preset track to reach a destination through autonomous decision so as to complete a preset task. As a product of modern high and new technology development, related technologies relate to various disciplines, such as information and sensing engineering, mechanical engineering, computer control science and the like, and become important indexes for measuring the national scientific research and innovation level.
The model prediction control is a computer control optimization algorithm based on model prediction, and has three main characteristics of prediction model, rolling optimization and feedback correction. The model predictive controller has good adaptability and robustness, and plays a significant role in important fields of industrial production process control, robot control, aerospace and the like. Like other control strategies, the control effect of model predictive control is closely related to the parameters of the controller, but because there is no definite linear relationship between the design parameters and the control effect of predictive control, the design of the parameters of the controller is very difficult, and how to quickly and effectively set the parameters of the predictive controller becomes an increasingly important research topic.
In the early days, researchers usually adopt an empirical method and a trial and error method to set parameters, and although the two methods are convenient, the two methods also have the problems of high requirements on experience of the researchers, unsatisfactory parameter setting effect and the like. With the rapid development of control theory, the current intelligent technology is widely applied to the field of parameter setting. Populus chensinensis et al in the literature "Tuning parameters for model predictive controlled CSA-SQP" combines a clonal selection algorithm and a sequence quadratic programming algorithm based on a biological immune principle, and provides a new method for Tuning and predicting controller parameters. An improved mixed frog-leaping algorithm is provided by Wangzhesheng and the like in the document Design of a multi-variable PID controller of electrically scaled and smoothed out and scaling hm based on the sensing and decision principle of the frog, and is used in the field of PID controller parameter setting. In the document "A novel adaptive PID controller based on operator-critical learning", an intelligent optimization algorithm based on action-evaluation learning is used in the field of PID controller parameter setting. The document "Performance analysis and online fuzzy self-tuning of RTD-a controller's parameters" proposes a method for tuning RTD-a controller parameters based on advanced control theories such as fuzzy logic and adaptive learning. The mapping rule in the machine learning algorithm is obtained through training and learning, and clear mathematical expression is not needed, so that the method has great advantages in parameter setting. Nowadays, machine learning algorithms are increasingly widely used in the field of parameter tuning. In the literature, "PID type prediction self-correction controller based on BP network", wangqun Xian et al uses BP neural network in the self-correction field of PID controller. In the document "RBF neural network-based PID parameter tuning of an electric arc furnace electrode regulation system", lujun et al use the RBF network to tune parameters of a PID controller. The machine learning method is widely applied to the field of PID controller design and parameter setting, but has little related achievement in the field related to model predictive control.
Disclosure of Invention
The invention aims to provide a model prediction control parameter setting method based on machine learning, aiming at the problem of path tracking of a wheeled mobile robot.
In order to achieve the purpose, the invention adopts the following technical scheme to realize the purpose:
a model prediction control parameter setting method based on machine learning comprises the following steps:
1) Determining the structure of the BP neural network;
2) Selecting a series of control parameters as output data of the BP neural network training sample;
3) Inputting the control parameters into a model prediction controller of the robot, performing a path tracking experiment, and collecting corresponding performance indexes in the path tracking experiment as input data of a BP neural network training sample;
4) Training a BP neural network by using the input data and the output data of the training sample to obtain a trained BP neural network;
5) Setting a performance index as an input given value of a BP neural network, and acquiring output data of the neural network through the trained neural network to realize parameter setting of model predictive control;
6) Inputting the control parameters set by the BP neural network into a model prediction controller of the robot, performing a path tracking experiment, collecting performance indexes generated by the path tracking experiment, comparing the performance indexes with preset performance indexes, and judging whether the set control parameters meet the requirements or not;
and if the set control parameters do not meet the requirements, returning to the step 4) until the requirements are met.
Further, if the difference value between the generated performance index and the preset performance index in the step 6) is within a preset range, the set control parameter meets the requirement;
and if the difference value between the generated performance index and the preset performance index is not in the preset range, the set control parameter does not meet the requirement.
Further, the number of network layers of the BP neural network in step 1) is 3, the number of nodes in the input layer is 4, the number of nodes in the output layer is 4, the number of nodes in the hidden layer is 4, and the weighting coefficient of each layer is a random value.
Further, the control parameters in step 2) include a predicted step length P, a path deviation weight coefficient Q, a stability weight coefficient R, and a speed difference weight coefficient M.
Further, the performance index includes an adjustment time S t Steady state achievement distance D s Oscillation time O t And oscillation distance O d 。
Further, S t Represents the time required from the start of control to the achievement of steady state;
D s indicating the distance from the beginning of control to the achievement of steady-state robot movement along the path tangential direction;
O t representing the time required for the robot to reach a steady state from tracking an upper path;
O d which represents the distance the robot moves along the tangential direction of the path during the oscillation time.
Further, when the lateral deviation of the robot is smaller than a lateral deviation threshold value for the first time, the robot realizes path tracking;
when the yaw angle of the robot is smaller than a yaw angle threshold value and the difference value between the speed and the preset speed is smaller than a speed difference threshold value, the robot is controlled stably;
when the robot achieves path tracking and control is stable, the robot achieves a steady state.
Further, step 1) is preceded by:
establishing a kinematic model of the mobile robot;
designing a path tracking algorithm based on predictive control, and establishing an optimization objective function based on a kinematics model and a deviation kinematics model;
based on the requirements on the rapidity and the stability of path tracking control, performance indexes and control parameters are designed.
Compared with the prior art, the invention has the following beneficial effects:
the model predictive control parameter setting method based on machine learning solves the problem of difficult model predictive control parameter setting by utilizing the characteristic that a machine learning algorithm does not need a determined mathematical expression but establishes a mapping relation through training. The invention adopts the BP neural network in the machine learning algorithm to set the parameters, has simple structure and convenient application; the model prediction controller and the neural network are two independent modules, whether the neural network module is used for parameter setting can be determined according to specific use conditions when the model prediction controller and the neural network are used, and the model prediction controller and the neural network have the characteristic of flexible application; the invention provides a plurality of new performance indexes, which can more comprehensively reflect the path tracking control effect of the robot; compared with the experience method and trial and error method which are widely applied nowadays, the method provided by the invention has the advantages that the rapidity and effectiveness of parameter setting are obviously improved.
Drawings
FIG. 1 is a schematic plane coordinate system of a robot according to an embodiment;
FIG. 2 is a diagram of a neural network architecture of an embodiment;
FIG. 3 is a block diagram of a controller of an embodiment;
FIG. 4 is a flow chart of parameter tuning according to an embodiment;
FIG. 5 is a diagram of a robot software structure of the embodiment;
FIG. 6 is a graph of experimental movement trajectories for path tracking according to an embodiment;
FIG. 7 is a graph of experimental lateral deviation over time for path tracking according to an embodiment;
FIG. 8 is a graph of experimental yaw angle over time for path tracking according to an embodiment.
Detailed Description
In order to make the technical solutions of the present invention better understood, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
It should be noted that the terms "first," "second," and the like in the description and claims of the present invention and in the drawings described above are used for distinguishing between similar elements and not necessarily for describing a particular sequential or chronological order. It is to be understood that the data so used is interchangeable under appropriate circumstances such that the embodiments of the invention described herein are capable of operation in other sequences than those illustrated or described herein. Furthermore, the terms "comprises," "comprising," and "having," and any variations thereof, are intended to cover a non-exclusive inclusion, such that a process, method, system, article, or apparatus that comprises a list of steps or elements is not necessarily limited to those steps or elements expressly listed, but may include other steps or elements not expressly listed or inherent to such process, method, article, or apparatus.
The invention is described in further detail below with reference to the accompanying drawings:
the invention provides a model control parameter setting method based on machine learning, which is used for path tracking control of a wheeled mobile robot and comprises the following specific implementation steps:
As shown in FIG. 1, the pose of the robot
X and Y represent the coordinates of the robot,
representing the course angle of the robot and the pose of the tracked path point on the map coordinate system
Establishing a path tracking deviation model x of a robot e ：
The path tracking problem of the mobile robot can be described as that the path tracking deviation x of the robot is enabled to be given a reference path through designing a control scheme e And constantly approaches 0.
For the convenience of calculation, the robot control center is set to be located at the center of two differential driving wheels, v represents the velocity, and ω represents the angular velocity, so that the kinematic model of the robot is as follows:
after coordinate transformation, the deviation kinematics model of the robot is expressed as:
predictive controller output u = [ v, w] T . Because the robot adopts a differential driving mode, the output u of the prediction controller needs to be converted into u 2 ＝[v l ,v r ] T The final control can be realized in the form of (1), calculated as:
wherein v is l And v r The linear speeds of the left wheel and the right wheel of the robot are respectively, and d is the wheel track between the left driving wheel and the right driving wheel.
For the wheel type mobile robot system, the prediction control means that within a proper prediction step length of 1 ≤ k ≤ N, the following optimization problems are solved:
x t+k+1，t ＝f(x t+k，t ，u t+k，t )
x t+k，t ∈Γ
u t+k，t ∈ψ
wherein t is the current time, k is the prediction step number, and Γ and ψ are constraints.
Optimizing an objective function J N Specifically, the following are shown:
q is a path deviation weight coefficient, and mainly influences the tracking capability of the robot on the reference path;
r is a control stability weight coefficient, and mainly influences the stability maintaining capability of the robot;
m is a speed deviation weight coefficient, and mainly influences the capability of the linear speed of the robot reaching a set value.
If a solution exists in the above optimization problem, by solving the problem, an optimal control sequence can be obtained:
the model predictive control is a rolling optimization control, namely after obtaining an optimal control sequence, applying a first control quantity of the optimal control sequence to the robot:
and at the next moment, the robot system obtains a new state, takes the new state as an initial state, recalculates the optimization problem and applies control, and continuously cycles until the control process is completely finished.
The purpose of the parameter setting algorithm based on machine learning is to perform parameter setting on a model prediction controller, and the control index is required to be designed firstly for reflecting the effect of the parameter setting when the parameter setting is performed.
In the path tracking control of the wheeled mobile robot, a plurality of characteristic state quantities such as a control time t, a lateral deviation D, a yaw angle θ, a speed v, and the like are obtained, and a control target for achieving a steady state is set by integrating the above state quantities. In order to determine whether the robot has reached a steady state, the robot is setThree thresholds are determined: transverse deviation threshold D L Yaw angle threshold θ L And a speed difference threshold v L 。
When the transverse deviation D of the robot is smaller than a transverse deviation threshold value D for the first time L When (II | < D |) L ) And the robot realizes path tracking.
At this time, whether the robot deviates from the path again cannot be judged, and other characteristic quantities need to be introduced to ensure that the robot stably runs on the path.
When the yaw angle theta of the robot is smaller than the threshold value theta L and the speed v and the preset speed v are 2 Is less than a threshold value v L Time (| theta | < theta |) L )∩(‖v-v 2 ‖≤v L ) And the robot is controlled stably.
When the robot realizes path tracking and control is stable (II | < D |) L )∩(‖θ‖≤θ L )∩(‖v-v 2 ‖≤v L ) The robot reaches a steady state.
Selecting the adjustment time S t (setting time) as performance index 1. The conditioning time is the time it takes for the robot control to start until a steady state is achieved.
In the actual robot path tracking control, factors such as terrain limitation are often encountered, and not only the adjustment time S is adjusted t There is a demand that the distance required to achieve steady state is also high. To accommodate this situation, the present invention proposes a steady state achievement distance D s (distance before stability) and the adjustment time S t And the tracking effect can be more comprehensively reflected by matching. While the steady state is reached by a distance D s Will also be referred to as performance index 2.
Steady state achievement distance D s To control the distance the robot moves along the tangential direction of the path, D, from the start to the achievement of a steady state s Comprises the following steps:
selecting oscillation time O t (catalysis time) as a performance index 3, showing the machineThe robot follows the upper path to the time required to reach steady state.
Selecting the oscillation distance O d (oscilloting distance) as a performance index 4, which represents the distance that the robot moves along the tangential direction of the path during the oscillation time, O d Comprises the following steps:
of the above four indices, the time S is adjusted t And a steady state achievement distance D s Mainly reflecting the overall ability of the tracking path, the oscillation time O t Distance to oscillation O d More emphasis is placed on reflecting the stability of the system.
When the predictive control has constraints, the mapping relation between the proposed performance index and the control parameter is difficult to obtain, so the BP neural network is selected to describe the mapping relation.
According to an empirical method, selecting a neural network with the number of layers of 3 and the learning rate of 0.1, randomly giving an initial weight, and using a Sigmoid function as a transfer function, wherein the expression is as follows:
the parameter setting indexes are used as the data of the input layer, and as can be seen from the above, the performance indexes of the parameter setting include the adjusting time St, the steady state achievement distance Ds, the oscillation time Ot and the oscillation distance Od, and the expression of the input layer is as follows:
O (1) ＝(x 1 ,x 2 ,x 3 ,x 4 )＝(St,Ds,Ot,Od) (12)
selecting the number of nodes of the hidden layer as 4 according to an empirical formula, wherein the expression is as follows:
in the formula (I), the compound is shown in the specification,
the coefficients are weighted for the hidden layer.
The purpose of the neural network is to find optimal model predictive control parameters, so that the robot can stably and rapidly run along a given track, namely, the optimal control parameters are utilized to improve the dynamic and static performances of the model predictive controller. In the model predictive control of the robot, the parameters of the controller comprise a prediction step length P, a path deviation weight coefficient Q, a stability weight coefficient R and a speed difference weight coefficient M, and are used as output data of a neural network, and the expression of an output layer is as follows:
O (3) ＝(y 1 ，y 2 ，y 3 ，y 4 )＝(P，Q，R，M)
each change in the set of control parameters may result in a set of performance indicator data, and then a sample set may be obtained by continuously changing the control parameters. And the performance indexes and the control parameters corresponding to each group are respectively used as input data and output data of the neural network so as to train the whole network. The neural network structure is shown in fig. 2.
And 4, parameter setting.
And inputting a new performance index and outputting a control parameter by using the trained neural network.
The following describes a specific flow of parameter tuning, and a specific flow chart is shown in fig. 4:
first, the structure of the BP neural network is determined. The number of the neural network layers used in this embodiment is 3, the number of nodes of the input layer is 4, the number of nodes of the output layer is 4, the number of nodes of the hidden layer is 4, and the weighting coefficients of the layers are random values;
selecting several sets of model control parameters with good control effect as output data of the BP neural network training sample;
inputting the selected parameters into a model prediction controller of the robot, performing a path tracking experiment, and acquiring performance indexes of a path tracking effect as input data of a neural network training sample;
fourthly, inputting the data of the training sample into a BP neural network for training;
fifthly, setting a reasonable performance index as an input given value of the BP neural network, adjusting a weighting coefficient and realizing parameter setting of model predictive control;
sixthly, using the BP neural network setting parameters for a robot predictive control algorithm, and observing the condition that the robot tracks a given path;
collecting performance indexes of the robot path tracking effect, comparing the performance indexes with preset performance indexes, and judging whether the set parameters meet requirements or not;
if not, returning to the step 4 until the requirements are met.
Examples
The experimental platform adopted by the invention is a wheeled mobile robot with a singlechip as a core. And an STM32F103 single chip microcomputer is selected as the bottom layer controller to perform speed closed-loop control on the direct current motor. The upper computer uses the ROS system to perform complex operations. The software structure is shown in fig. 5.
The reference path is set to a straight line. Let reference speed v r 0.5m/s, a lateral deviation threshold D L Is 0.03m, yaw angle threshold theta L 0.02rad, threshold value of velocity difference v L Is 0.01m/s. The initial state speed of the robot is set to 0, the lateral deviation is set to 1m, and the yaw angle is set to 0.
The predetermined performance index is [5.0,2.5,0]I.e. the target that the path tracing wishes to achieve is to adjust the time S t At 5.0s, the steady state achieved distance D s 2.5m, oscillation time O t Is 0, oscillation distance O d Is 0.
Two groups of experimental data are selected to participate in comparison, one group uses parameters set by a machine learning algorithm, and the other group uses parameters set by an empirical method. The path tracking experiments were performed under the same conditions using two sets of parameter controlled robots, respectively, and the results of the experiments are shown in fig. 6 to 8. In fig. 6, the moving trajectory curve of the robot with the parameter set by the machine learning algorithm is smoother, the used space is smaller, and the robot can adapt to complex road conditions. In fig. 7, the robot with the machine learning algorithm setting parameters has a lateral deviation of 0.0294m at 4.9s, is smaller than 0.03m for the first time, and tracks the upper reference path. In fig. 8, the yaw angle of the robot with the machine learning algorithm setting parameters is 0.0196rad at 5.2s, and is less than 0.02rad for the first time. The adjusting time of the robot for setting parameters by the machine learning algorithm is 5.2s, and the oscillation time is 0.3s. And the adjusting time of the robot for setting the parameters by the empirical method is 5.9s, and the oscillation time is 0.2s.
The ratio of the characteristic value and the preset index obtained by the experiment of the two parameters is shown in table 1, and as can be seen from table 1, the experiment of the machine learning algorithm setting parameter is carried out at the adjusting time S t And a steady state achievement distance D s The two aspects are closer to the preset performance indexes and are greatly improved. At the oscillation time O t And oscillation distance O d In both aspects, the two experiments are very close to the preset index and have small difference. In a comprehensive view, the parameters of the model predictive controller of the wheeled mobile robot system can be effectively set through a machine learning algorithm, and the parameter setting capability is excellent.
Table 1 is a comparison table of the path tracking experiment of the present embodiment
The above-mentioned contents are only for illustrating the technical idea of the present invention, and the protection scope of the present invention is not limited thereby, and any modification made on the basis of the technical idea of the present invention falls within the protection scope of the claims of the present invention.
Claims (5)
1. A model predictive control parameter setting method based on machine learning is characterized by comprising the following steps:
1) Determining the structure of the BP neural network;
2) Selecting a series of control parameters as output data of the BP neural network training sample;
3) Inputting the control parameters into a model prediction controller of the robot, performing a path tracking experiment, and collecting corresponding performance indexes in the path tracking experiment as input data of a BP neural network training sample;
4) Training a BP neural network by using the input data and the output data of the training sample to obtain a trained BP neural network;
5) Setting a performance index as an input given value of the BP neural network, and acquiring output data of the neural network through the trained neural network to realize parameter setting of model predictive control;
the performance indicator includes an adjustment time S t Steady state achievement distance D s Oscillation time O t And oscillation distance O d ；
S t Represents the time required from the start of control to the achievement of steady state;
D s controlling the distance from the beginning to the end of the steady-state robot moving along the tangential direction of the path;
O t representing the time required for the robot to reach a steady state from tracking an upper path;
O d the distance of the robot moving along the tangential direction of the path in the oscillation time is represented;
when the transverse deviation of the robot is smaller than a transverse deviation threshold value for the first time, the robot realizes path tracking;
when the yaw angle of the robot is smaller than the threshold value and the difference value between the speed and the preset speed is smaller than the threshold value, the robot is controlled stably;
when the robot realizes path tracking and control is stable, the robot achieves a stable state;
6) Inputting the control parameters set by the BP neural network into a model prediction controller of the robot, performing a path tracking experiment, collecting performance indexes generated by the path tracking experiment, comparing the performance indexes with preset performance indexes, and judging whether the set control parameters meet the requirements or not;
and if the set control parameters do not meet the requirements, returning to the step 4) until the requirements are met.
2. The method for tuning model predictive control parameters based on machine learning according to claim 1, wherein the tuned control parameters meet requirements if the difference between the generated performance index and the preset performance index in step 6) is within a preset range;
and if the difference value between the generated performance index and the preset performance index is not in the preset range, the set control parameter does not meet the requirement.
3. The method for tuning model predictive control parameters based on machine learning according to claim 1, wherein the number of network layers of the BP neural network of step 1) is 3, the number of nodes of the input layer is 4, the number of nodes of the output layer is 4, the number of nodes of the hidden layer is 4, and the weighting coefficients of each layer are random values.
4. The method for tuning model predictive control parameters based on machine learning of claim 1, wherein the control parameters of step 2) include a prediction step length P, a path deviation weight coefficient Q, a stability weight coefficient R, and a velocity difference weight coefficient M.
5. The tuning method of model predictive control parameters based on machine learning according to claim 1, characterized in that step 1) is preceded by:
establishing a kinematic model of the mobile robot;
designing a path tracking algorithm based on predictive control, and establishing an optimization objective function based on a kinematics model and a deviation kinematics model;
based on the requirements on the rapidity and the stability of path tracking control, performance indexes and control parameters are designed.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110580904.6A CN113325694B (en) | 2021-05-26 | 2021-05-26 | Model prediction control parameter setting method based on machine learning |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110580904.6A CN113325694B (en) | 2021-05-26 | 2021-05-26 | Model prediction control parameter setting method based on machine learning |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN113325694A CN113325694A (en) | 2021-08-31 |
| CN113325694B true CN113325694B (en) | 2022-12-09 |
Family
ID=77421343
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110580904.6A Active CN113325694B (en) | 2021-05-26 | 2021-05-26 | Model prediction control parameter setting method based on machine learning |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113325694B (en) |
Families Citing this family (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115108261A (en) * | 2022-08-29 | 2022-09-27 | 成都盛锴科技有限公司 | Bogie dragging system |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104597753A (en) * | 2014-12-19 | 2015-05-06 | 徐工集团工程机械股份有限公司道路机械分公司 | Method and device for intelligently controlling asphalt and crushed stone spreading of synchronous chip sealer |
| CN111711400A (en) * | 2020-04-24 | 2020-09-25 | 成都理工大学工程技术学院 | PID speed regulation method and system based on BP fuzzy neural network |
Family Cites Families (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US6445963B1 (en) * | 1999-10-04 | 2002-09-03 | Fisher Rosemount Systems, Inc. | Integrated advanced control blocks in process control systems |
| CN107615186A (en) * | 2015-08-13 | 2018-01-19 | 华为技术有限公司 | The method and apparatus of Model Predictive Control |
| CN105700538B (en) * | 2016-01-28 | 2019-04-12 | 武汉光庭信息技术股份有限公司 | Track follower method based on neural network and pid algorithm |
| CN107561942B (en) * | 2017-09-12 | 2021-02-09 | 重庆邮电大学 | Intelligent vehicle trajectory tracking model prediction control method based on model compensation |
| CN109884900B (en) * | 2019-04-03 | 2022-04-12 | 东南大学 | Design method of harvester path tracking controller based on adaptive model predictive control |
| CN110221611B (en) * | 2019-06-11 | 2020-09-04 | 北京三快在线科技有限公司 | Trajectory tracking control method and device and unmanned vehicle |
| CN111624992B (en) * | 2020-04-28 | 2021-07-09 | 北京科技大学 | Path tracking control method of transfer robot based on neural network |
| CN112622903B (en) * | 2020-10-29 | 2022-03-08 | 东北大学秦皇岛分校 | Longitudinal and transverse control method for autonomous vehicle in vehicle following driving environment |
| CN112327625B (en) * | 2020-11-12 | 2022-02-22 | 西安建筑科技大学 | Uncertain system model predictive control parameter setting method based on machine learning |
-
2021
- 2021-05-26 CN CN202110580904.6A patent/CN113325694B/en active Active
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104597753A (en) * | 2014-12-19 | 2015-05-06 | 徐工集团工程机械股份有限公司道路机械分公司 | Method and device for intelligently controlling asphalt and crushed stone spreading of synchronous chip sealer |
| CN111711400A (en) * | 2020-04-24 | 2020-09-25 | 成都理工大学工程技术学院 | PID speed regulation method and system based on BP fuzzy neural network |
Also Published As
| Publication number | Publication date |
|---|---|
| CN113325694A (en) | 2021-08-31 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108008720B (en) | Fuzzy sliding mode trajectory tracking control and method for wheeled mobile robot | |
| CN111413966B (en) | Progressive model prediction unmanned planning tracking cooperative control method | |
| CN109491389B (en) | Robot trajectory tracking method with speed constraint | |
| CN109884900B (en) | Design method of harvester path tracking controller based on adaptive model predictive control | |
| Wang et al. | Spot welding robot path planning using intelligent algorithm | |
| CN110450156B (en) | Optimal design method of self-adaptive fuzzy controller of multi-degree-of-freedom mechanical arm system | |
| Wen et al. | Elman fuzzy adaptive control for obstacle avoidance of mobile robots using hybrid force/position incorporation | |
| CN111338346A (en) | Automatic driving control method and device, vehicle and storage medium | |
| CN111273551A (en) | Double-steering wheel automatic guided vehicle trajectory tracking control method and system | |
| Boukens et al. | Design of an intelligent optimal neural network-based tracking controller for nonholonomic mobile robot systems | |
| CN109799814A (en) | Crawler belt vehicle contrail tracker design method based on kinematics model | |
| CN109669463A (en) | A kind of section trace tracking method that considering AGV speed and the front-wheel deviation angle and can be changed | |
| CN109991934B (en) | Time-optimal online S-type acceleration and deceleration planning method | |
| CN113435025B (en) | Robot high-performance track automatic generation method combined with multistage optimization model | |
| CN111538328A (en) | Priority hierarchical prediction control method for obstacle avoidance trajectory planning and tracking control of autonomous driving vehicle | |
| CN115157238A (en) | Multi-degree-of-freedom robot dynamics modeling and trajectory tracking method | |
| CN113325694B (en) | Model prediction control parameter setting method based on machine learning | |
| CN116185014A (en) | Intelligent vehicle global optimal track planning method and system based on dynamic planning | |
| CN115236973A (en) | AGV trajectory tracking control method based on PSO Lyapunov function | |
| Anderson et al. | Modeling a professional driver in ultra-high performance maneuvers with a hybrid cost MPC | |
| Chen et al. | An adaptive path tracking controller based on reinforcement learning with urban driving application | |
| CN109249393B (en) | Multi-parameter robot real-time behavior correction method based on empirical control | |
| CN115047882A (en) | Omnidirectional mobile robot track tracking method based on fuzzy sliding mode control | |
| CN115079569A (en) | AGV nonlinear active disturbance rejection control method based on longicorn stigma search algorithm | |
| Chen et al. | Prescribed‐time stabilisation control of differential driven automated guided vehicle |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |