CN110788859B - A Global Adaptive Adjustment System of Controller Parameters - Google Patents

Abstract

The invention discloses a controller parameter universe self-adaptive adjusting system, which comprises the following steps of firstly, determining a clustering center of three driving joint load inertias of a parallel mechanism in a dynamic platform reference point working space universe and the membership degree of a sample inertia about each center by means of a fuzzy clustering algorithm; then, off-line setting parameters of feedback and feedforward controllers at corresponding characteristic points of each clustering center, and storing the parameters and membership data in a control system database; and finally, calculating and updating the controller parameters of the three driving joints in real time by using the data in the database and the current coordinate information of the reference point of the movable platform of the parallel mechanism by adopting a gravity center method.

Description

A Global Adaptive Adjustment System of Controller Parameters

technical field

The invention relates to a global self-adaptive adjustment method for controller parameters of a five-degree-of-freedom hybrid robot, which relates to the field of robotics and automation, and can effectively improve the motion control accuracy of the end-effector of the five-degree-of-freedom hybrid robot.

Background technique

The robot control system generally adopts the composite control strategy of "feedback + feedforward". The feedback controller determines the system bandwidth and ensures the stability; the feedforward controller further improves the following accuracy of the system. However, for a hybrid robot system with a parallel mechanism, the load inertia converted to the drive joint of the parallel mechanism varies with the robot configuration. The controller with fixed gain is difficult to meet the requirements of high-speed and high-precision motion of such robots in the workspace, which restricts its application in occasions with high requirements for motion accuracy (such as machining).

For such strongly nonlinear time-varying systems, advanced control algorithms such as dynamic control algorithms, neural networks and genetic algorithms have been proposed. However, due to the complexity of the algorithms, the large amount of computation and the lack of hardware support, it is difficult to put them into practical applications. Therefore, there is an urgent need for an adaptive adjustment method of controller parameters suitable for such robot systems with simple and easy-to-implement algorithms, so that the parameters can match the changes of the load inertia in the entire workspace, and realize online adjustment in different robot configurations.

SUMMARY OF THE INVENTION

Aiming at a five-degree-of-freedom hybrid robot with a rotating bracket disclosed in patent CN104985596A, the present invention proposes a global self-adaptive adjustment system for controller parameters, which can effectively deal with the load inertia that changes with the configuration when the parallel mechanism drives the joints to move. The influence of a factor on the quality of control. The core of the present invention is to first determine the cluster center of the load inertia of the three drive joints of the parallel mechanism in the whole domain of the reference point work space of the moving platform and the membership degree of the sample inertia about each center by means of a fuzzy clustering algorithm, and set each cluster center offline. Corresponding to the controller parameters at the feature points, and then using the center of gravity method to estimate the controller parameters of the three drive joints online when the reference point of the moving platform is located at any point in the workspace. The invention has the advantages that only the controller parameters at a limited number of characteristic points need to be adjusted offline, so that the parameters can be matched with the load inertia change to realize the global self-adaptive adjustment, and the algorithm is simple, occupies less hardware resources, and is easy to implement.

A controller parameter global adaptive adjustment system, the system includes a rough interpolation module, a position inverse solution module and a control algorithm module, the system also includes a controller parameter estimation module, the controller parameter estimation module performs the following steps to adjust the parameters Adjustment to match load inertia changes, including:

The membership degree obtained by cluster analysis and the controller parameters obtained by offline tuning are written into the corresponding registers as global variables for calling, and the controller parameters tuned in the work space are used as their initial values;

The rough interpolation module performs rough interpolation on the NC code according to the motion law, and calculates the end pose after a rough interpolation cycle is completed;

The position inverse solution module calculates the corresponding moving platform reference point coordinates and drive joint commands, and writes the moving platform reference point coordinates as global variables into the temporary stack register;

并作为全局变量写入相应的寄存器；The coarse interpolation module calls the controller parameter estimation module in an interrupt mode according to the integer multiple of the coarse interpolation period, and reads the moving platform reference point coordinate, membership degree and n characteristic point controller parameters K _i,k from the corresponding register (k=1,2...,n), use the center of gravity method to estimate the controller parameters of the i-th drive joint corresponding to the reference point coordinates

And write to the corresponding register as a global variable;

利用更新后的

计算控制器的输出指令，并在下一次估算前保持

不变。Call the control algorithm module and read from the corresponding register

Use the updated

Calculate the output command of the controller and hold it until the next evaluation

constant.

是动平台的参考点位于工作空间中任意点时的驱动关节的控制器参数；该关节在所有特征点处控制器参数的加权估计，权重由与当前参考点距离最为接近的样本点确定，定义该样本点序号为m(j＝m)；控制器参数估计算法为the controller parameters

is the controller parameter of the drive joint when the reference point of the moving platform is located at any point in the workspace; the weighted estimation of the controller parameters of the joint at all feature points, the weight is determined by the sample point with the closest distance to the current reference point, the definition The sequence number of the sample point is m (j=m); the controller parameter estimation algorithm is

The cluster analysis step includes:

其中，根据机器人数学模型，由样本点计算各驱动关节负载惯量，定义第i个关节的第j个样本惯量为I_i,j；(1) Grid divide the working space of the reference point of the parallel mechanism moving platform, and define the grid nodes as sample points

Wherein, according to the robot mathematical model, the load inertia of each drive joint is calculated from the sample points, and the jth sample inertia of the ith joint is defined as I _i,j ;

(2) According to the fuzzy clustering analysis algorithm, given the number of cluster categories is n (n<<N), according to the formula to calculate in the lth iteration, the jth sample inertia is about the kth (k=1,2, ...,n) the membership degrees of the cluster centers μ _i,j,k,l and the new cluster center I _i,k,l+1 ;

In the formula, α represents a fuzzy index greater than 1, usually α=2;

(3) Continuously update μ _i,j,k,l and I _i,k,l+1 according to the above formula until it satisfies ||I _i,k,l+1 -I _{i,k, l} ||≤ε. Here, ε represents the cluster center iteration accuracy.

beneficial effect

In order to ensure the stability of the system operation, the task priority of calling the controller parameter estimation module is lower than that of calling the servo algorithm module, and the servo task interruption method is used to realize the calling of the module. The calling cycle of the controller parameter estimation module can be determined according to the task requirements and hardware computing capability, and can generally be set to an integer (n _T ) multiple of the coarse interpolation cycle. Accordingly, the controller parameters are estimated and updated every n _T coarse interpolation cycles and remain unchanged until the next estimation.

The global self-adaptive adjustment method for the controller parameters of the five-degree-of-freedom hybrid robot proposed by the present invention uses the cluster analysis algorithm to determine the feature points and the degree of membership of any point in the workspace with respect to the feature points, and based on the online estimation reference point is located at any point controller parameters at the time. The advantage is that it only needs to tune the controller parameters at a limited number of characteristic points offline, so that the parameters can be matched with the load inertia change to achieve global adaptive adjustment; the algorithm is simple and occupies less hardware resources; independent from the servo algorithm, it can ensure the servo Real-time adjustment of parameters under the premise of stable control.

Description of drawings

Figure 1. Block diagram of the global adaptive control strategy of the hybrid robot

Figure 2 Schematic diagram of workspace grid division and feature points

Fig. 3 Calculation flow chart of adaptive adjustment of controller parameters

Detailed ways

In order to make the technical solutions of the present invention clearer, the present invention will be further described in detail below with reference to the accompanying drawings. It should be understood that the specific examples described herein are only used to explain the present invention, but are not limited to this example. Taking a five-degree-of-freedom hybrid robot with a rotating bracket disclosed in patent CN104985596A as an example, the specific implementation of the present invention is described as follows.

A controller parameter global adaptive adjustment system of the present invention (as shown in FIG. 1 ) includes the following steps:

1. Fuzzy cluster analysis

The grid is divided into the working space of the reference point (referred to as the reference point) of the parallel mechanism motion platform, and the grid nodes are defined as sample points. According to the mathematical model of the robot, the load inertia of each drive joint is calculated from the sample points, which is defined as the sample inertia. The sample inertia is analyzed by means of fuzzy clustering algorithm, and the cluster center and the membership degree of each sample inertia to the cluster center are calculated.

2. Offline tuning of controller parameters

The sample points corresponding to the inertia of the cluster center are defined as feature points, and the controller parameters of the three drive joints of the parallel mechanism are set and recorded offline when the reference points reach these feature points.

3. Define global variables

In order to improve the update speed of the controller parameters, it is necessary to read the controller parameters of the feature points, the membership degree of the sample inertia about the cluster center (the sample points about the feature points) and the current coordinates of the reference point of the moving platform in real time. To do this, define the above parameters as global variables to increase the speed of reading and writing to them.

4. Online estimation and adjustment of controller parameters

In the process of executing the motion program, the current coordinate of the reference point is calculated by the position inverse solution module, and is written into the temporary stack register as a global variable, and the current reference point coordinate and the controller parameters at the feature point are read by the controller parameter estimation module. Determine the sample point closest to the current reference point and read its membership degree about each feature point, use the center of gravity method to estimate the drive joint controller parameters corresponding to the coordinates of the current reference point, and use it to update the controller parameters in the control algorithm module .

Step 1: Fuzzy Cluster Analysis

The grid is divided into the working space of the reference point (referred to as the reference point) of the parallel mechanism motion platform, and the grid nodes are defined as sample points. According to the mathematical model of the robot, the load inertia of each drive joint is calculated from the sample points, which is defined as the sample inertia. The sample inertia is analyzed by means of fuzzy clustering algorithm, and the cluster center and the membership degree of each sample inertia to the cluster center are calculated.

根据机器人数学模型，由样本点计算各驱动关节负载惯量，定义第i个关节的第j个样本惯量为I_i,j。(1) As shown in Figure 2, the grid is divided into the working space of the reference point (hereinafter referred to as the reference point) of the parallel mechanism motion platform, and the grid node is defined as the sample point

According to the robot mathematical model, the load inertia of each drive joint is calculated from the sample points, and the j-th sample inertia of the i-th joint is defined as I _i,j .

(2) According to the fuzzy clustering analysis algorithm, given the number of cluster categories is n (n<<N), according to the formula to calculate in the lth iteration, the jth sample inertia is about the kth (k=1,2, ...,n) membership degrees of cluster centers μ _i,j,k,l and new cluster centers I _i,k,l+1 :

In the formula, α represents a blur index greater than 1, and α=2 is usually taken.

(3) Continuously update μ _i,j,k,l and I _i,k,l+1 according to the above formula until it satisfies ||I _i,k,l+1 -I _{i,k, l} ||≤ε. Here, ε represents the cluster center iteration accuracy.

Step 2: Offline tuning of controller parameters

离线整定并记录参考点达到这些特征点时并联机构三个驱动关节的控制器参数Define the sample points corresponding to the cluster center inertia I _{i, k, l} as feature points

Off-line tuning and recording of the controller parameters of the three drive joints of the parallel mechanism when the reference point reaches these characteristic points

Ki _,k = {K _i,k,1 Ki _,k,2 Ki _,k,3 Ki _,k,4 Ki _,k,5 }

={K _i,k,P K _i,k,I K _i,k,D K _i,k,vff K _i,k,aff }

In the formula, K _i,k represents the control parameter sequence of the i-th (i=1, 2, 3) drive joint at the k-th (k=1 _, 2, ..., n) feature point. K _i,k,P ,K _i,k,I ,K _i,k,D ,K _i,k,vff ,K _i,k,aff represent the proportional, integral, differential, speed front Gains of feedforward and acceleration feedforward controllers.

According to the symmetry of the mechanism and the workspace, the feature points of the second and third drive joints are symmetrical about the midplane, and the parameters of the other joint can be obtained by adjusting the feature point parameters of one of the two joints.

Step 3: Define global variables

In order to improve the update speed of the controller parameters, it is necessary to read the controller parameters of the feature points, the membership degree of the sample inertia about the cluster center (the sample points about the feature points) and the current coordinates of the reference point of the moving platform in real time. To do this, define the above parameters as global variables to increase the speed of reading and writing to them.

Step 4: Online estimation and adjustment of controller parameters

In the process of executing the motion program, the present invention uses the position inverse solution module to calculate and obtain the current coordinates of the reference point, and writes it into the temporary stack register as a global variable, and the controller parameter estimation module reads the coordinates of the current reference point and the controller at the feature point. parameters, determine the sample point closest to the current reference point and read its membership degree with respect to each feature point, use the center of gravity method to estimate the parameters of the drive joint controller corresponding to the coordinates of the current reference point, and use it to update the control algorithm in the control algorithm module. device parameters.

The described centroid method assumes that the controller parameters when the reference point is located at any point in the workspace is a weighted estimate of the controller parameters of all feature points, and the weight (ie, the degree of membership) is determined by the sample point with the closest distance to the current reference point, which is defined as The sample point serial number is m (j=m). Accordingly, the controller parameter estimation algorithm can be expressed as

并作为全局变量写入相应的寄存器。调用控制算法模块，从相应的寄存器中读取

利用更新后的

计算控制器的输出指令，并在下一次估算前保持

不变。The calculation flow of the global adaptive adjustment of the controller parameters is shown in Figure 3. It can be seen from the figure that in the implementation process, the membership degree obtained by cluster analysis and the controller parameters obtained by offline tuning need to be written into the corresponding registers as global variables for calling, and the controller parameters tuned in the workspace will be set. as its initial value. During the execution of the NC program, the NC code is firstly subjected to rough interpolation according to the motion law to calculate the end pose after a rough interpolation cycle is completed. joint instruction, and write the coordinate of the moving platform reference point into the temporary stack register as a global variable. According to the integer multiple of the rough interpolation period, the controller parameter estimation module is called in an interrupt mode, and the reference point coordinates of the moving platform, the degree of membership and the controller parameters of the feature point are read from the corresponding registers, and the center of gravity method is used to estimate the coordinates corresponding to the reference point. The controller parameters of the ith driven joint of

And write to the corresponding register as a global variable. Call the control algorithm module and read from the corresponding register

Use the updated

Calculate the output command of the controller and hold it until the next evaluation

constant.

Claims (3)

1. a controller parameter global adaptive adjustment system, the system comprises a rough interpolation module, a position inverse solution module and a control algorithm module, it is characterized in that, this system also comprises a controller parameter estimation module, described controller parameter estimation The module performs the following steps to adjust the parameter matching load inertia change, including: Write the membership degree obtained by cluster analysis and the characteristic point controller parameters obtained by offline tuning as global variables into the corresponding registers for calling, and use the controller parameters tuned in the workspace as its initial values; The rough interpolation module performs rough interpolation on the NC code according to the motion law, and calculates the end pose after a rough interpolation cycle is completed; The position inverse solution module calculates the corresponding moving platform reference point coordinates and drive joint commands, and writes the moving platform reference point coordinates as global variables into the temporary stack register; The coarse interpolation module calls the controller parameter estimation module in an interrupt mode according to the integer multiple of the coarse interpolation period, reads the coordinates of the reference point of the moving platform, the degree of membership and the controller parameters of the feature point from the corresponding registers, and uses the center of gravity method to estimate and The controller parameters of the i-th drive joint corresponding to the coordinates of the reference point

And write to the corresponding register as a global variable; Call the control algorithm module and read from the corresponding register

Use the updated

Calculate the output command of the controller and hold it until the next evaluation

constant. 2 . The global adaptive adjustment system for controller parameters according to claim 1 , wherein the controller parameters

is the controller parameter of the drive joint when the reference point of the moving platform is located at any point in the workspace; the weighted estimation of the controller parameters of the joint at all feature points, the weight is determined by the sample point with the closest distance to the current reference point, and is defined by The sample point serial number is m (j=m); the controller parameter estimation algorithm is

3. a kind of controller parameter global self-adaptive adjustment system according to claim 1, is characterized in that, described global variable is written into corresponding register to prepare for invocation step comprises: (1) Grid divide the working space of the reference point of the parallel mechanism moving platform, and define the grid nodes as sample points

Wherein, according to the robot mathematical model, the load inertia of each drive joint is calculated from the sample points, and the jth sample inertia of the ith joint is defined as I _i,j ; (2) According to the fuzzy clustering analysis algorithm, given the number of cluster categories is n (n<<N), according to the formula to calculate in the lth iteration, the jth sample inertia is about the kth (k=1,2, ...,n) the membership degrees of the cluster centers μ _i,j,k,l and the new cluster center I _i,k,l+1 ;

In the formula, α represents a fuzzy index greater than 1; (3) Continuously update μ _i,j,k,l and I _i,k,l+1 according to the above formula until it satisfies ||I _i,k,l+1 -I _{i,k, l} ||≤ε; here, ε represents the cluster center iteration accuracy.

Images