Real-time estimation and optimization analysis method for joint torque of operation mechanical arm

Abstract

The invention discloses a real-time estimation and optimization analysis method for the joint torque of an operation mechanical arm. The method comprises the following steps that S1, joint position data and torque data of the mechanical arm are collected; s2, a kinetic model and a torque calculation model of the mechanical arm are constructed based on a Lagrange kinetic equation, and a system error equation is obtained by combining the kinetic model and the torque calculation model; s3, constructing a fuzzy neural network model based on a fuzzy control rule and a neural network, and estimating and compensating the moment of force of the mechanical arm by taking the joint position error and the speed error of the mechanical arm as input; s4, a mechanical arm self-adaptive control strategy is constructed based on the torque calculation model and the fuzzy neural network model, and mechanical arm joint torque fluctuation is optimized; according to the method, the fuzzy neural network model is constructed on the basis of the mechanical arm dynamic model, the uncertain part of the mechanical arm is optimized and compensated, and high-precision control and stable operation of the mechanical arm are achieved.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of mechanical arm joint torque optimization, and particularly relates to a method for real-time estimation and optimization analysis of operation mechanical arm joint torque. BACKGROUND

[0002] With the rapid development of modern science and technology, the mechanical arm plays an increasingly important role in many fields. Due to the characteristics of nonlinearity, strong coupling and variable pose of the mechanical arm, the traditional control mode is often difficult to meet the requirement of high control precision due to the influence of external uncertain factors, and is prone to cause problems such as poor robustness of the control system and difficulty in realizing global stable progression, resulting in unsatisfactory control effect.

[0003] Therefore, it is necessary to develop a method for real-time estimation and optimization analysis of operation mechanical arm joint torque to solve the above problems. SUMMARY

[0004] The application aims to overcome the shortcomings of the prior art and provide a method for real-time estimation and optimization analysis of operation mechanical arm joint torque. The fuzzy neural network model is constructed on the basis of the dynamic model of the mechanical arm to optimize and compensate the uncertain part of the mechanical arm, so as to realize high-precision control and stable operation of the mechanical arm.

[0005] The application is achieved in the following manner: a method for real-time estimation and optimization analysis of operation mechanical arm joint torque, comprising the following steps:

[0006] S1, collecting joint position data and torque data of the mechanical arm based on the encoder and torque sensor assembly arranged at the joint of the mechanical arm;

[0007] S2, constructing a dynamic model and a torque calculation model of the mechanical arm based on the Lagrange dynamics equation, and obtaining a system error equation by combining the dynamic model and the torque calculation model;

[0008] S3, constructing a fuzzy neural network model based on fuzzy control rules and neural networks, taking the joint position error and velocity error of the mechanical arm as inputs, and estimating and compensating the torque of the mechanical arm; specifically comprising the following steps:

[0009] S31, constructing an input layer, and taking the joint position error and velocity error of the mechanical arm as inputs;

[0010] S32, setting a Gaussian fuzzy unit to calculate the fuzzy membership degree of the input signal through the Gaussian function;

[0011] S33, performing fuzzy reasoning according to the fuzzy control rules to obtain a feasible solution of the fuzzy control rules;

[0012] S34, optimizing the feasible solution of the fuzzy control rule by a genetic algorithm to obtain an optimal solution of the fuzzy control rule;

[0013] S35, de-fuzzing and obtaining the output of the fuzzy neural network model;

[0014] S4, constructing a mechanical arm adaptive control strategy based on the moment calculation model and the fuzzy neural network model, and optimizing the joint moment fluctuation of the mechanical arm.

[0015] Further, the dynamics model in step S2 is represented as: , wherein, is the joint angle position vector; is the joint velocity vector; is the joint acceleration vector; is the inertia matrix; is the vector containing the Coriolis force and the centripetal force; is the gravity vector; is the external disturbance term vector; is the joint driving moment vector.

[0016] Further, the moment calculation model is constructed based on the dynamics model in step S2 to obtain:

[0017]

[0018] , wherein, , , are the estimated values of , , respectively; is the joint desired acceleration vector; is the error vector, , is the joint desired position vector; , are the velocity and position feedback gain coefficient matrices respectively; is the system uncertainty term.

[0019] Further, the error equation is obtained by combining the dynamics model and the moment calculation model in step S2, represented as:

[0020]

[0021] , wherein, is the estimated value of ; , are the error terms of the Coriolis force and the centripetal force, the gravity, , ; This represents the system's uncertainties.

[0022] Furthermore, in step S32, fuzzy membership degree calculation is performed using the Gaussian function, specifically as follows:

[0023]

[0024] In the formula, The input to the fuzzy neural network model, Enter the number; , These are the first and second errors in the column matrix of the error equations. The element of the first The center value and width of a set of membership function language words, a vector sum vector These represent the center value and width of all membership functions, respectively; This represents the output value corresponding to each set of membership function language terms.

[0025] Furthermore, in step S33, fuzzy inference is performed according to the fuzzy control rules, and the influence of each node rule on the output is represented by a weighted multiplication method:

[0026]

[0027] In the formula, Indicates the first The degree of influence of each language set on the output in the fuzzy control rule; Indicates the first The output corresponding to the fuzzy control rule; This indicates the number of fuzzy control rules.

[0028] Furthermore, step S34, which optimizes the feasible solution of the fuzzy control rule based on a genetic algorithm, specifically includes the following steps: ① Encoding the feasible solution of the fuzzy control rule; ② Initializing the population by randomly generating several populations from the feasible solution vector; ③ Calculating the fitness value; ④ Performing individual selection, where the probability of an individual being selected is expressed as: In the formula, Indicates population size, Represents an individual ⑤ Crossover mutation, where the mutation rate is set between 0.001 and 0.1; ⑥ Obtain the new population; ⑦ Determine whether the termination condition is met. If not, return to step ③. If it is met, decode and output the optimal solution.

[0029] Furthermore, the fitness function for calculating the fitness value in step ③ is expressed as:

[0030]

[0031] , wherein, represents a target function; represents the weight of each target function; specifically, the target function adopts 3 evaluation criteria, i.e. , respectively adopts the average percentage error , the root mean square error and represents the coefficient is represented, wherein, is a predicted value, is an actual value, is the center value of .

[0032] Further, the output of the deblurring and obtaining of the fuzzy neural network model in the step S35 is specifically represented as:

[0033]

[0034] , wherein, represents the i-th fuzzy control rule, represents the number of fuzzy control rules; is the weight of the output layer; represents the output corresponding to the i-th fuzzy control rule. Further, in the step S3, after the fuzzy neural network model is constructed, the model is trained, the historical data of the mechanical arm is divided into a training set, a verification set and a test set, the data in the training set is input into the fuzzy neural network model for training, after the training is completed, the data in the verification set is used to verify the accuracy of the current model, whether the accuracy requirement is met is judged, if not met, the training is continued, if met, the training is completed, after the training is completed, the model is tested by using the data in the test set.

[0035] Due to the adoption of the above technical solutions, the beneficial effects of the present application are:

[0036] (1) Based on the mechanical arm dynamics model and the torque calculation model, by constructing the fuzzy neural network model, the ability of the fuzzy logic system to approximate any nonlinear function with arbitrary precision, and the self-learning nature of the neural network to continuously adjust its own parameters, the nonlinear link of the mechanical arm system is approximated, the uncertain part of the mechanical arm is compensated, the torque fluctuation is optimized, and high-precision control and stable operation of the mechanical arm are realized;

[0037] (1) Based on the mechanical arm dynamics model and the torque calculation model, by constructing the fuzzy neural network model, the ability of the fuzzy logic system to approximate any nonlinear function with arbitrary precision, and the self-learning nature of the neural network to continuously adjust its own parameters, the nonlinear link of the mechanical arm system is approximated, the uncertain part of the mechanical arm is compensated, the torque fluctuation is optimized, and high-precision control and stable operation of the mechanical arm are realized;

[0038] ​(2) In the construction of fuzzy neural network model, the feasible solution of fuzzy control rule is optimized by adopting genetic algorithm, the response speed of the model is improved, and the efficiency and accuracy of the model are effectively improved. BRIEF DESCRIPTION OF DRAWINGS

[0039] Figure 1 is the flow chart of the present application.

[0040] Figure 2 is the flow chart of the fuzzy neural network model constructed in the present application.

[0041] Figure 3 is the principle diagram of the genetic algorithm for optimizing the feasible solution of fuzzy control rule in the present application.

[0042] Figure 4 is the principle diagram of the adaptive control strategy in the present application. DETAILED DESCRIPTION

[0043] The technical solutions of the present application will be further specifically described below by examples in combination with the drawings.

[0044] As shown in Figure 1 , Figure 2 , Figure 3 , Figure 4 , a method for real-time estimation and optimization analysis of joint torque of a mechanical arm, comprising the following steps:

[0045] S1, collecting joint position data and torque data of the mechanical arm based on the encoder and torque sensor components arranged at the joints of the mechanical arm.

[0046] S2, constructing a dynamics model and a torque calculation model of the mechanical arm based on Lagrange dynamics equation, and obtaining a system error equation by combining the dynamics model and the torque calculation model.

[0047] Specifically, the dynamics model in the step S2 is represented as: , wherein, is the joint angle position vector; is the joint velocity vector; is the joint acceleration vector; is the inertia matrix; is the vector containing the Coriolis force and the centripetal force; is the gravity vector; is the external disturbance term vector; is the joint driving torque vector.

[0048] Specifically, the torque calculation model is constructed based on the dynamics model in the step S2 to obtain:

[0049]

[0050] , wherein, , , are respectively , , estimated values of is a joint desired acceleration vector; is an error vector, , is a joint desired position vector; , are respectively velocity and position feedback gain coefficient matrices; is a system uncertainty term.

[0051] Specifically, the step S2 combines the dynamic model and the torque calculation model to obtain an error equation, which is expressed as:

[0052]

[0053] , wherein, is an estimated value of ; , are respectively error terms of coriolis force and centripetal force, gravity, , ; is a system uncertainty term.

[0054] S3, a fuzzy neural network model is constructed based on fuzzy control rules and neural networks, taking joint position error and velocity error of the robot arm as input to estimate and compensate the torque of the robot arm; specifically comprising the following steps:

[0055] S31, an input layer is constructed, taking joint position error and velocity error of the robot arm as input.

[0056] S32, a Gaussian base fuzzy ware is set, and the input signal is subjected to fuzzy membership calculation through Gaussian function.

[0057] Specifically, the step S32 performs fuzzy membership calculation through Gaussian function, which is specifically expressed as:

[0058]

[0059] , wherein, is input of the fuzzy neural network model, is the number of inputs; , are respectively center value and width of the first membership function language word set of the first element in the error equation column matrix, is the number of elements in the error equation column matrix, is the number of elements in the error equation column matrix. and vector respectively represent the center value and width of all membership functions; represent the output value corresponding to each membership function language set.

[0060] S33, fuzzy inference is performed according to the fuzzy control rules to obtain a feasible solution of the fuzzy control rules.

[0061] Specifically, the step S33 of performing fuzzy inference according to the fuzzy control rules adopts a weighted multiplication method to represent the influence of each node rule of the current layer on the output:

[0062]

[0063] , wherein, represents the influence degree of each language set in the i th fuzzy control rule on the output; represents the output corresponding to the i th fuzzy control rule; represents the number of fuzzy control rules.

[0064] S34, the feasible solution of the fuzzy control rules is optimized by a genetic algorithm to obtain an optimal solution of the fuzzy control rules.

[0065] Specifically, the step S34 of optimizing the feasible solution of the fuzzy control rules based on the genetic algorithm specifically includes the following steps: ① encoding the feasible solution of the fuzzy control rules; ② initializing the population and randomly generating a plurality of populations of the feasible solution vectors; ③ calculating the fitness value; ④ performing individual selection, and the probability of individual selection is represented as: , wherein, represents the population size, represents the fitness value of the individual; ⑤ crossover and mutation, wherein the mutation rate is set to be between 0.001 and 0.1; ⑥ obtaining a new population; ⑦ judging whether the termination condition is met, if not, returning to step ③, if yes, decoding and outputting the optimal solution.

[0066] Specifically, the fitness function for calculating the fitness value in the step ③ is represented as:

[0067]

[0068] , wherein, represents the objective function; represents the weight of each objective function; specifically, the objective function adopts three evaluation criteria, i.e. , respectively adopting the average percentage error , the root mean square error​​ and Represents coefficients Let represent, in the formula, For predicted values, This is the actual value. for The central value of .

[0069] S35. Defuzzify and obtain the output of the fuzzy neural network model.

[0070] Specifically, the deblurring and obtaining the output of the fuzzy neural network model in step S35 is specifically represented as follows:

[0071]

[0072] In the formula, Indicates the first Fuzzy control rules, Indicates the number of fuzzy control rules; The weights of the output layer; Indicates the first The output corresponding to each fuzzy control rule.

[0073] Specifically, in step S3, after constructing the fuzzy neural network model, the model is trained. The historical data of the robotic arm is divided into a training set, a validation set, and a test set. The data in the training set is input into the fuzzy neural network model for training. After training, the accuracy of the current model is verified using the data in the validation set to determine whether it meets the accuracy requirements. If it does not meet the requirements, training continues; if it does meet the requirements, training is complete. After training is completed, the model is tested using the data in the test set.

[0074] S4. Based on the torque calculation model and the fuzzy neural network model, construct an adaptive control strategy for the robotic arm to optimize the torque fluctuation of the robotic arm joints.

[0075] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit it. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that modifications or equivalent substitutions can still be made to the specific implementation of the present invention. Any modifications or equivalent substitutions that do not depart from the spirit and scope of the present invention should be covered within the scope of the claims of the present invention.

Claims

1. A method for real-time estimation and optimization analysis of joint torques of a robotic arm, characterized in that: Includes the following steps: S1. Collect joint position data and torque data of the robotic arm based on encoder and torque sensor components installed at the joints of the robotic arm; S2. Construct a dynamic model and a torque calculation model for the robotic arm based on the Lagrange dynamics equations, and obtain the system error equation by combining the dynamic model and the torque calculation model; S3. Construct a fuzzy neural network model based on fuzzy control rules and neural networks, using the joint position error and speed error of the robotic arm as inputs, to estimate and compensate for the torque of the robotic arm; specifically including the following steps: S31. Construct the input layer, taking the joint position error and speed error of the robotic arm as input; S32. Set up the Gaussian fuzzy converter and calculate the fuzzy membership degree of the input signal through the Gaussian function; S33. Perform fuzzy reasoning based on the fuzzy control rules to obtain feasible solutions for the fuzzy control rules; S34. Optimize the feasible solutions of the fuzzy control rules using a genetic algorithm to obtain the optimal solution of the fuzzy control rules; S35. Deblurring and obtaining the output of the fuzzy neural network model; S4. Based on the torque calculation model and the fuzzy neural network model, construct an adaptive control strategy for the robotic arm to optimize the torque fluctuation of the robotic arm joints.

2. The method for real-time estimation and optimization analysis of joint torque of a robotic arm according to claim 1, characterized in that: The dynamic model in step S2 is expressed as follows: In the formula, This is the joint angle position vector; This is the joint velocity vector; This is the joint acceleration vector; The inertia matrix; It is a vector containing the Coriolis force and the centripetal force; It is the gravity vector; The vector of external perturbation terms; This is the joint driving torque vector.

3. The method for real-time estimation and optimization analysis of joint torque of a robotic arm according to claim 2, characterized in that: In step S2, the torque calculation model is constructed based on the dynamic model to obtain: ， 4. In the formula, , , They are respectively , , The estimated value; Let the joint be the desired acceleration vector; For the error vector, , Let the joint be the desired position vector; , These are the velocity and position feedback gain coefficient matrices, respectively; This represents the system's uncertainties.

5. The method for real-time estimation and optimization analysis of joint torque of a robotic arm according to claim 3, characterized in that: In step S2, the error equation is obtained by combining the dynamic model and the torque calculation model, and is expressed as follows: ， 6. In the formula, for The estimated value; , These are the error terms for Coriolis force, centripetal force, and gravity, respectively. , ; This represents the system's uncertainties.

7. The method for real-time estimation and optimization analysis of joint torque of a robotic arm according to claim 1, characterized in that: In step S32, the fuzzy membership degree is calculated using the Gaussian function, specifically as follows: ， 8. In the formula, The input to the fuzzy neural network model, Enter the number; , These are the first and second errors in the column matrix of the error equations. The element of the first The center value and width of a set of membership function language words, a vector sum vector These represent the center value and width of all membership functions, respectively; This represents the output value corresponding to each set of membership function language terms.

9. The method for real-time estimation and optimization analysis of joint torque of a robotic arm according to claim 1, characterized in that: In step S33, fuzzy inference is performed based on fuzzy control rules, and the influence of each node rule on the output is represented by a weighted multiplication method: ， 10. In the formula, Indicates the first The degree of influence of each language set on the output in the fuzzy control rule; Indicates the first The output corresponding to the fuzzy control rule; This indicates the number of fuzzy control rules.

11. The method for real-time estimation and optimization analysis of joint torque of a robotic arm according to claim 1, characterized in that: Step S34, which optimizes the feasible solution of the fuzzy control rule based on a genetic algorithm, specifically includes the following steps: ① Encoding the feasible solution of the fuzzy control rule; ② Initializing the population by randomly generating several populations from the feasible solution vector; ③ Calculating the fitness value; ④ Performing individual selection, where the probability of an individual being selected is expressed as: In the formula, Indicates population size, Represents an individual ⑤ Crossover mutation, where the mutation rate is set between 0.001 and 0.1; ⑥ Obtain the new population; ⑦ Determine whether the termination condition is met. If not, return to step ③. If it is met, decode and output the optimal solution.

12. The method for real-time estimation and optimization analysis of joint torque of a robotic arm according to claim 7, characterized in that: The fitness function used to calculate the fitness value in step ③ is expressed as follows: ， 13. In the formula, express One objective function; This represents the weight of each objective function; Specifically, the objective function is represented by three evaluation criteria, namely... The average percentage error was used respectively. Root mean square error and Represents coefficients Let represent, in the formula, For predicted values, This is the actual value. for The central value of .

14. The method for real-time estimation and optimization analysis of joint torque of a robotic arm according to claim 1, characterized in that: The deblurring and obtaining the output of the fuzzy neural network model in step S35 is specifically represented as follows: ， 15. In the formula, Indicates the first Fuzzy control rules, Indicates the number of fuzzy control rules; The weights of the output layer; Indicates the first The output corresponding to each fuzzy control rule.

16. The method for real-time estimation and optimization analysis of joint torque of a robotic arm according to claim 1, characterized in that: In step S3, after constructing the fuzzy neural network model, the model is trained. The historical data of the robotic arm is divided into a training set, a validation set, and a test set. The data in the training set is used to input the fuzzy neural network model for training. After training, the data in the validation set is used to verify the accuracy of the current model and determine whether it meets the accuracy requirements. If it does not meet the requirements, training continues. If it does meet the requirements, training is completed. After training is completed, the data in the test set is used to test the model.

Images