Optimization method of robot joint module control system

Abstract

To provide a theoretical basis effective for optimizing a performance by effectively improving optimization efficiency and accuracy of a robot joint module control parameter.SOLUTION: An optimization method of a robot joint module control system includes the steps of: performing a response performance test of a robot joint module, designing a performance test using, as test variables, a mass m representing a load, an output speed v of the joint module, a proportionality coefficient P, an integral coefficient I, and a differential coefficient D, and acquiring a joint module response performance data set; performing response performance prediction of the robot joint module, constructing a prediction model, and training the prediction model by the response performance data set; and optimizing the joint module response performance, and acquiring a control parameter after optimization by the trained prediction model.SELECTED DRAWING: Figure 1

Description

Technical Field

[0001] The present invention relates to an optimization method for a control system, and particularly to an optimization method for a robot joint module control system, belonging to the technical field of robot joint control.

Background Art

[0002] As the utilization of artificial intelligence and robots increasingly integrates into human life and work, the robot joint module has become an essential core moving member for a joint robot to achieve motion. Therefore, the performance of the robot joint module directly affects the transmission accuracy of the robot body and performance indicators such as operating speed, vibration, and noise. The control method adopted by a single robot joint module is PID control, and the proportional coefficient P, integral coefficient I, and differential coefficient D are its main control parameters. The main response performance indicators of the robot joint module are mainly the overshoot amount and response time. The factors affecting the response performance of the robot joint module include load, output rotation speed, and control parameters (proportional coefficient P, integral coefficient I, and differential coefficient D). Among them, the load and output rotation speed are determined by the operating conditions of the robot joint module, and the control parameters are set artificially. Therefore, the quality of the control parameters directly determines the superiority or inferiority of the performance of the robot joint module. Summing up the above, by optimizing the control parameters of the robot joint module and improving the response performance of the joint module, the overall performance of the robot can be effectively improved.

[0003] Currently, the conventional debugging method often uses a test method to debug and calibrate the control parameters for the entire robot. However, in addition to the difficulty of accurately obtaining the optimal control parameters for a single joint module, a large number of tests are required, so the optimization efficiency is low.

[0004] Chinese Patent CN108227479B discloses a PID control method and a PID control system for an articulated robot that enable faster response of the system and faster stabilization of the device through comprehensive adjustment of the overall gain, proportional gain, integral gain, and derivative gain. However, this reference does not consider the influence of PID control parameters on its control performance at different speeds and loads, and has drawbacks in terms of parameter selection and abnormal value processing, and cannot meet the requirements for high-precision control.

Summary of the Invention

Problems to be Solved by the Invention

[0005] The object of the present invention is to provide an optimization method for a robot joint module control system for the problems existing in the prior art. This method can realize the prediction of the response performance of the robot joint module by combining the response performance test of the robot joint module and an intelligent model, and can optimize the control parameters of the joint module.

Means for Solving the Problems

[0006] Performing a response performance test on the robot joint module, designing a five-factor response performance test with the copper mass m representing the load, the output speed v of the joint module, the proportional coefficient P of the joint module control system, the integral coefficient I of the joint module control system, and the derivative coefficient D of the joint module control system as test variables, and obtaining a joint module response performance dataset in step S1; Performing a response performance prediction on the robot joint module, constructing a prediction model, and training the prediction model with the joint module response performance dataset obtained in step S1 in step S2; Optimize the response performance of the joint module, establish the response performance index of the robot joint module, uniformly take values within the range of five factors in step S1 to form parameter combinations, substitute each parameter combination into the prediction model trained in step S2 to obtain the minimum value of the response performance index, and use it as the optimized control parameter in step S3. This is an optimization method for the robot joint module control system.

[0007] The present invention can realize the prediction of the response performance of the robot joint module and optimize the control parameters of the joint module by combining the response performance test of the robot joint module with an intelligent model while considering the influence of PID control parameters on its control performance at different speeds and loads.

[0008] Preferably, in order to study the influence of PID control parameters on its control performance at different speeds and loads in the joint module, the specific response performance test steps of the robot joint module in step S1 are as follows: In step S1.1, construct a response performance test bench for the robot joint module, select the robot joint module 1 to be tested and fix it to the base 2 by the fixing flange 7. The base 2 is fixedly attached to the plate 5. The output flange 6 of the robot joint module 1 is provided with a link 3, and a weight 4 with a mass of m is arranged at the other end of the link 3. The mass m of the weight 4 is exchanged according to the test requirements. Connect the robot joint module 1 to the host computer debugger. The host computer controls the robot joint module 1 according to the speed and load designed for the test and adjusts the parameter values of PID. In step S1.2, conduct test design, design an orthogonal test table, and design an orthogonal test table with n (n≥5) levels of five factors using the weight mass m representing the load, the output speed v of the joint module, the proportional coefficient P, the integral coefficient I, and the differential coefficient D as test variables. In step S1.3, using the joint module response performance dataset, based on the orthogonal test table in step S1.2, conduct tests on the robot joint module response performance test bench constructed in step S1.1 to obtain the robot joint module speed response curves with different test parameters, and extract from the speed response curves the overshoot amount O shoot and response time R time which are the response performance indicators of the joint module, extract, record, and store them to form the joint module response performance dataset.

[0009] The joint module is an integrated mechanical device that can realize movements such as rotation, swinging, and translation in different directions of the mechanical structure to complete specific functions and tasks. The joint module is widely used in various mechanical devices and apparatuses such as assembly industrial robots and can have different designs depending on the use case. Conduct joint module operation tests according to the orthogonal test table designed using the constructed test bench, and relevant data such as the response time and overshoot amount of the joint module under different influencing factors are obtained in the tests. By organizing the data, a dataset can be provided for training the prediction model, and by analyzing, the relationship between the overshoot amount and response speed of the system at different loads and different operating speeds can be obtained.

[0010] Preferably, to solve the problem that the model may fall into a local optimum solution, the specific steps of optimizing the joint module response performance in step S2 are as follows. In step S2.1, construct a MA - BPNN moth - flame neural network prediction model, where the prediction model includes a MA moth - flame model and a BPNN backpropagation neural network model. The MA moth - flame model provides the optimized weights and thresholds for the BPNN backpropagation neural network model. In step S2.2, train the MA-BPNN mayfly-neural network prediction model. First, introduce the joint module response performance dataset obtained in step S1.2 into the model. The number of mayfly individuals is initially N. Subsequently, update the positions of male and female mayflies, calculate the fitness values of male and female mayflies, and then repeatedly generate offspring mayflies and update their positions. When the fitness value of the offspring mayfly is less than F v stop the iteration; otherwise, continue to generate offspring mayflies and iterate. Use the global optimal parameters generated by the mayfly as the weights and thresholds of the neural network model for prediction, calculate the error, and when the error after iteration is less than E u output the prediction results of the overshoot amount O shoot and the response time R time , and the training of the MA-BPNN prediction model is completed.

[0011] The optimization of the mechanical joint module control parameters is a very complex non-linear problem, which is affected by multiple factors such as load, operating speed, and PID control parameters. It is very difficult to establish an effective mathematical relationship to describe the relationship between them. Therefore, exploring the relationship between these influencing factors and the response performance of its control system by means of machine learning is an effective means. Currently, various machine learning methods have been incorporated into the research on joint module motion control. Among them, it has been proven that the artificial neural network has achieved good results in terms of control parameter optimization. However, the above model has the problem that it may fall into a local optimal solution. To solve this problem, the control parameter optimization model using MA-BPNN comprehensively utilizes the strong non-linear mapping ability and elastic structure of the BP neural network, and the strong fault tolerance ability of the MA search model against errors and noises, so as to effectively avoid the inconvenience of falling into a local optimal solution in the calculation process.

[0012] Preferably, to realize the prediction of the robot joint module response performance, the training steps for the MA-BPNN mayfly-neural network prediction model in step S2.2 are as follows: In step S2.2.1, initialize the model, introduce the joint module response performance dataset obtained in step 1.3 into the model, and set the number of male and female mayflies in the mayfly model to N mayflies each, where the position of each mayfly represents a set of weight and threshold parameters of a BP neural network, the speed represents the step size in the update process of these parameters, initialize the weights and thresholds of the BP neural network, and initialize the position and speed of the mayfly. In step S2.2.2, update the speed and position of the male mayfly individuals, and update the speed v t+1 mij and position x t+1 ij in the j -th dimensional search space of the i -th male mayfly after the (t + 1)-th iteration according to equations (1) and (2).

Number

Number

Number

Equation

Equation

Number

Number

Number

[0013] The prediction of the response performance of the robot joint module is realized, and the problem that the optimization efficiency of the optimal control parameters of the robot joint module is low in the conventional method is solved. In addition, by adjusting the weight coefficients in the response performance index, the optimal control parameters that meet different response performance requirements can be obtained, effectively improving the optimization efficiency and accuracy of the robot joint module control parameters, and providing an effective theoretical basis for the optimization of the robot's performance.

[0014] Preferably, in order to optimize the control based on the optimization model, the specific steps of the joint module response performance optimization in step S3 are as follows: In step S3.1, establish the robot joint module response performance index K

Equation

[0015] By optimizing the control parameters of the joint module and obtaining the optimal values of the control parameters at different speeds and loads, the optimization problem of the joint module control parameters can be solved.

Advantages of the Invention

[0016] The optimization method of the robot joint module control parameters proposed in the present invention combines the orthogonal test of the robot joint module response performance and the intelligent model (MA-BPNN model) to realize the prediction of the robot joint module response performance, and solves the problem of low efficiency in optimizing the optimal control parameters of the robot joint module in the conventional method. In addition, by adjusting the weight coefficient in the response performance index, the optimal control parameters that meet different response performance requirements can be obtained, effectively improving the optimization efficiency and accuracy of the robot joint module control parameters, and providing an effective theoretical basis for the optimization of the robot performance.

Brief Description of the Drawings

[0017]

Figure 1

Figure 2

Figure 3

Figure 4

Figure 5

Figure 6

Mode for Carrying Out the Invention

[0018] Hereinafter, while referring to the drawings in the embodiments of the present invention, the technical solution means in the embodiments of the present invention will be clearly and completely described. Naturally, the described embodiments are only a part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments in the present invention, all other embodiments obtained by those skilled in the art without creative efforts shall fall within the protection scope of the present invention.

[0019] As shown in FIG. 1, the specific implementation procedure of the present invention includes a step S1 of performing a response performance test on a robot joint module, a step S2 of performing a response performance prediction on the robot joint module, and a step S3 of optimizing the response performance of the joint module.

[0020] Specifically, the S1 includes the following steps S1.1, step S1.2, and step 1.3. In step S1.1, a response performance test bench for a robot joint module is constructed. As shown in FIGS. 2 and 3, the robot joint module 1 is fixed to the base 2 by the fixing flange 7, the base 2 is attached to the plate 5 by bolts, the link 3 is attached to the output flange 6 of the robot joint module, the weight 4 with a weight of m is arranged at the other end of the link 3, and the weight m of the weight 4 can be exchanged according to the test conditions. In step S1.2, an orthogonal test table is designed. A total of 25 sets of orthogonal test tables with 5 levels of 5 factors are designed with the copper mass m, the output speed v of the joint module, the proportional coefficient P of the joint module control system, the integral coefficient I of the joint module control system, and the differential coefficient D of the joint module control system as test variables. In step 1.3, according to the orthogonal test table in step S1.2, 25 sets of tests are performed on the response performance test bench of the robot joint module constructed in step 1.1. The robot joint module speed response curves under different test parameters are obtained, and the overshoot amount O, which is the response performance index of the joint module, is obtained from the speed response curves. shootand response time R time are extracted, as shown in FIG. 4. The values of 25 sets of m, v, P, I, O shoot , R time are recorded and stored to form a joint module response performance dataset.

[0021] Specifically, the step S2 includes the following steps S2.1 and S2.2. In step S2.1, an MA-BPNN mayfly-neural network prediction model is constructed. This model mainly includes an MA mayfly model and a BPNN backpropagation neural network model, and the MA mayfly model provides the optimized weights and thresholds for the BPNN backpropagation neural network model. In step S2.2, the MA-BPNN prediction model is trained. As shown in FIG. 5, the training steps are as follows.

[0022] In step S2.2.1, the model is initialized. The joint module response performance dataset obtained in step S1.3 is introduced into the model. The number of male and female mayflies in the mayfly model is set to N mayflies = 25 respectively. The position of each mayfly represents a set of weight and threshold parameters of a BP neural network, and the speed represents the step size in the update process of these parameters. The weights and thresholds of the BP neural network are initialized. The position and speed of the mayfly are initialized. In step S2.2.2, the speed and position of the male mayfly individuals are updated. The speed v t+1 mij and position x t+1 ij in the j-th dimensional search space of the i-th male mayfly after the (t + 1)-th iteration are updated according to equations (1) and (2). [Number] However, t is the t-th iteration, and x t ij is the position of the i-th male mayfly in the j-th dimensional search space after the t-th iteration, and vt mij is the velocity of the $i$-th male mayfly in the $j$-th dimensional search space after the $t$-th iteration,

Number

Number

Number

Number

Equation

[0023] The comparison of the prediction results between the trained MA - BPNN prediction model and the conventional BP neural network model is as shown in Figure 6. Compared with the conventional BP neural network model, the coefficient of determination of the results of the MA - BPNN prediction model has increased by 2.06%, the mean squared error of O shoot has decreased by 46.92%, the mean squared error of R time has decreased by 43.75%, and the operation time has decreased by 0.33 s. The MA - BPNN prediction model proposed in the present invention is superior to the conventional BP neural network model in both the error of the prediction results and the operation efficiency.

[0024] Specifically, S3 includes the following steps S3.1, S3.2, and S3.3. In step S3.1, establish the response performance index K of the robot joint module.

Number

[0025] Based on the above description of the disclosed embodiments, those skilled in the art can implement or use the present invention. Various modifications to these embodiments will be apparent to those skilled in the art, and the general principles defined herein can be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention is not limited to these embodiments shown herein, but rather conforms to the broadest scope consistent with the principles and new features disclosed herein.

Claims

1. Performing a response performance test on a robot joint module, designing a five-factor response performance test with the copper mass m representing the load, the output speed v of the joint module, the proportional coefficient P of the joint module control system, the integral coefficient I of the joint module control system, and the differential coefficient D of the joint module control system as test variables, and obtaining a joint module response performance dataset in step S1; Performing a response performance prediction on the robot joint module, constructing a prediction model, and training the prediction model with the joint module response performance dataset obtained in step S1 in step S2; Optimizing the joint module response performance, establishing a robot joint module response performance index, uniformly taking values within the range of the five factors in step S1 to form parameter combinations, substituting the parameter combinations into the prediction models trained in step S2 respectively to obtain the minimum value of the response performance index, and using it as the optimized control parameter, including step S3. The optimization method of the robot joint module control system is characterized by this.

2. The specific steps of the response performance test of the robot joint module in step S1 are as follows: In step S1.1, construct a response performance test bench for the robot joint module, select the robot joint module 1 to be tested and fix it to the base 2 by the fixing flange 7. The base 2 is fixedly attached to the plate 5. The output flange 6 of the robot joint module 1 is provided with a link 3, and a copper weight 4 with a mass of m is arranged at the other end of the link 3. The mass m of the copper weight 4 is exchanged according to the test requirements. Connect the robot joint module 1 to the host computer debugger. The host computer controls the robot joint module 1 according to the designed speed and load in the test, and adjusts the PID parameter values. In step S1.2, perform test design, design an orthogonal test table, and design an orthogonal test table with n (n≥5) levels of five factors with the copper mass m representing the load, the output speed v of the joint module, the proportional coefficient P, the integral coefficient I, and the differential coefficient D as test variables. In step S1.3, using the joint module response performance dataset, and based on the orthogonal test table in step S1.2, conduct tests on the robot joint module response performance test bench constructed in step S1.1 to obtain the robot joint module speed response curves under different test parameters, and extract the overshoot amount O ｓｈｏｏｔ and response time R ｔｉｍｅ from the speed response curves, record and store them to form a joint module response performance dataset. The optimization method of the robot joint module control system according to claim 1, characterized in that.

3. The specific steps of the joint module response performance optimization in step S2 are as follows: In step S2.1, an MA-BPNN mayfly-neural network prediction model is constructed. The prediction model includes an MA mayfly model and a BPNN backpropagation neural network model. The MA mayfly model provides the optimized weights and thresholds for the BPNN backpropagation neural network model. In step S2.2, the MA-BPNN prediction model is trained. First, the joint module response performance dataset obtained in step S1.3 is introduced into the model. The number of mayfly individuals is initially N. Subsequently, the positions of male and female mayflies are updated, and the fitness values of male and female mayflies are calculated. Subsequently, offspring mayflies are repeatedly generated to update the positions of the offspring mayflies. When the fitness value of the offspring mayfly is less than F ｖ the iteration ends; otherwise, offspring mayflies continue to be generated and the iteration continues Using the global optimal parameters generated by the mayfly as the weights and thresholds of the neural network model, predictions are made, errors are calculated, and when the error after iteration is smaller than E ｕ the overshoot amount O ｓｈｏｏｔ and the response time R ｔｉｍｅ are output, and the training of the MA-BPNN prediction model is completed. The optimization method of the robot joint module control system according to claim 1, characterized in that.

4. The training steps for the MA-BPNN mayfly-neural network prediction model in step S2.2 are as follows: In step S2.2.1, initialize the model, introduce the joint module response performance dataset obtained in step 1.3 into the model, and set the number of male and female mayflies in the mayfly model to N ｍａｙｆｌｉｅｓ respectively. The position of each mayfly represents a set of weight and threshold parameters of a BP neural network, and the speed represents the step size in the update process of these parameters. Initialize the weights and thresholds of the BP neural network, and initialize the position and speed of the mayfly. In step S2.2.2, update the velocity and position of the male mayfly individuals, and use equations (1) and (2) to calculate the velocity v of the i-th male mayfly in the j-th dimensional search space after the (t + 1)-th iteration ｔ＋１ ｍｉｊ and the position x ｔ＋１ ｉｊ and update them 【Number 19】 However, t is the t-th iteration, and x ｔ ｉｊ is the position of the i-th male mayfly in the j-th dimensional search space after the t-th iteration, and v ｔ ｍｉｊ is the velocity of the i-th male mayfly in the j-th dimensional search space after the t-th iteration, 【Number 20】 da is the dance coefficient, which is used to describe the process of attracting female mayflies, and r ｃ is a random number, and r ｃ ∈ [−1, 1], and F ３ (x) is the fitness function, as shown in Equation 3, 【Number 21】 However, x is the position of the mayfly, and y ａｃｔｕａｌ is the actual output value, and y ｐｒｅｄｉｃｔｅｄ is the predicted output value, and N is the number of sample sets, In step S2.2.3, update the velocity and position of the female mayfly individuals, and use equations (4) and (5) to calculate the velocity v of the i-th female mayfly in the j-th dimensional search space after the (t + 1)-th iteration ｔ＋１ ｆｉｊ and the position y ｔ＋１ ｉｊ and update them 【Number 22】 However, y ｔ ｉｊ is the position of the i-th female mayfly in the j-th dimensional search space after the t-th iteration, and v ｔ ｆｉｊ is the velocity of the i-th female mayfly in the j-th dimensional search space after the t-th iteration, a ３ is the attraction coefficient of the female mayfly, r ｍｆ is the distance between the female mayfly and the male mayfly, and fl is the random flight coefficient In step S2.2.4, offspring mayflies are generated, and the mayfly population is updated by equations (6) and (7). 【No. 23】 However, offspring １ and offspring ２ represent two offspring mayflies, r ｌ is a random number, r ｌ ∈ [−1, 1], and male and female represent male mayflies and female mayflies, respectively, In step S2.2.5, the fitness value is judged, and F ｖ is the target fitness value, and F ３ (x) ≤ F ｖ If so, the iteration ends; otherwise, it proceeds to step S2.2.2 to repeat again. In step S2.2.6, the global optimal position of the mayfly is output, and the optimized weights and thresholds are obtained. In step S2.2.7, perform the forward propagation of the BPNN. The input neurons are the mass m of the weight, the output speed v of the joint module, the proportional coefficient P, the integral coefficient I, and the differential coefficient D of the joint module control system, and the output neurons are the overshoot amount O ｓｈｏｏｔ and the response time R ｔｉｍｅ That is, the data of the input layer is transmitted to the neurons of the hidden layer by the optimized weights and thresholds, and the excitation function F １ (x) = (e ｘ - e －ｘ ) / (e ｘ + e －ｘ ) is used to calculate the g-th neuron θ ｇ of the hidden layer according to Equation (8). The data of the hidden layer is transmitted to the neurons of the output layer by the weights and thresholds, and the excitation function F ２ (x) = (e ｘ ) / (e ｘ + e －ｘ ) is used to calculate the k-th neuron Y ｋ of the output layer according to Equation (9), [Number 24] However, X ｈ represents the h-th neuron in the input layer, and W ｈｇ is the weight between the h-th neuron in the input layer and the g-th neuron in the hidden layer, and T ｇ is the threshold of the g-th neuron in the hidden layer, and n ｈｉｄｅ is the number of nodes in the hidden layer, 【Number 25】 However, T ｋ is the threshold of the k-th neuron in the output layer, and W ｇｋ is the weight between the g-th neuron in the hidden layer and the k-th neuron in the output layer, and n ｏｕｔ is the number of nodes in the output layer, In step S2.2.8, the prediction error E ｕ is calculated, and the value of E ｕ is calculated according to Equation (10), [Number 26] In step S2.2.9, error backpropagation is performed, and the weights and thresholds are adjusted using the backpropagation model to reduce E ｕ thereby In step S2.2.10, by repeating the training and repeatedly executing steps S2.2.7, S2.2.8, and S2.2.9, until ｕ E ≤ ｖ the weights and thresholds are adjusted, where E ｖ is the target prediction error, and the prediction results of O ｓｈｏｏｔ and R ｔｉｍｅ are output, and the training of the MA - BPNN prediction model is completed. The optimization method of the robot joint module control system according to claim 1, characterized in that.

5. The specific steps for optimizing the joint module response performance in step S3 are as follows: In step S3.1, a robot joint module response performance index K is established. [Number 27] However, α １ is the overshoot amount weighting coefficient, and α ２ is the response time weighting coefficient, and α １ + α ２ = 1, In step S3.2, the test variables are the five factors that affect the robot joint module, namely the pendulum mass m, the output speed v of the joint module, the proportional coefficient P, the integral coefficient I, and the differential coefficient D. As each factor, n (n≥5) values are uniformly taken within their respective value ranges, and n ５ sets of parameter combinations are formed. In step S3.3, the n ５ sets of parameter combinations in step S3.2 are respectively substituted into the MA-BPNN model trained in step S2, the response performance index K is calculated for each parameter combination, and the parameter combination of the proportional coefficient P, integral coefficient I, and differential coefficient D with the minimum K value obtained is used as the optimized optimal control parameter of the robot joint module control system according to claim 1 or 4. An optimization method for the control system.

Images