A method and system for identifying dynamic parameters of a mechatronic system based on an inverse model

Abstract

The present invention belongs to the technical field related to mechatronic control, and discloses a method and system for identifying dynamic parameters of a mechatronic system based on an inverse model. The method includes the following steps: constructing an inverse model of the driving torque of the manipulator to be processed; collecting the actual motion parameters of the manipulator to be processed when moving under the expected motion parameters, and calculating the expected driving torque and the actual driving torque by using the inverse model; constructing an error function by using the expected driving torque and the actual driving torque, taking the actual moment of inertia of the manipulator to be processed as a variable, constructing an optimization model with the minimum error function as the goal, and solving the optimization model to obtain the actual moment of inertia corresponding to the minimum error function, so as to realize the identification of the dynamic parameters of the manipulator to be processed. Through the present invention, the problems that the existing parameter identification technology has low calculation efficiency, poor robustness when dealing with mechatronic systems with strong nonlinear characteristics, and cannot adapt to variable working condition scenarios are solved and overcome.

Description

Technical Field

[0001] The present invention belongs to the technical field related to mechatronic control, and more specifically, relates to a method and system for identifying dynamic parameters of a mechatronic system based on an inverse model. Background Art

[0002] A mechatronic system refers to a system composed of mechanical devices and electrical equipment, which is widely used in fields such as automation, robotics, aerospace, and intelligent manufacturing. The performance of a mechatronic system is closely related to its dynamic parameters (such as inertia, friction, stiffness, damping, etc.), which directly affect the motion response, stability, control accuracy, and anti-interference ability of the system. Identifying the dynamic parameters of a mechatronic system aims to identify the unknown dynamic parameters in the model by observing the input and output data of the system to improve the performance and reliability of the system. By accurately identifying the dynamic parameters of a mechatronic system, the control accuracy, response speed, and adaptability to different working conditions of the system can be effectively improved.

[0003] Traditional parameter identification methods estimate the values of dynamic parameters by constructing a mathematical model and combining observational data, mainly including the least squares method, maximum likelihood estimation method, etc. However, these methods are usually used when assuming the system model is linear or approximately linear, and only consider minimizing the error, ignoring the complex nonlinear characteristics of mechatronic systems, showing deficiencies such as strong dependence on the initial value, sensitivity to external disturbances of the system, and high computational complexity. To overcome the deficiencies of traditional parameter identification methods, parameter identification methods based on optimization algorithms use optimization algorithms such as genetic algorithms and particle swarm algorithms to find the optimal dynamic parameter values that can minimize the objective function. Although they can effectively handle nonlinear problems, they often have problems such as slow convergence speed and easy entrapment in local optima in practical applications; data-driven parameter identification methods fit the dynamic parameter values through machine learning algorithms, which can conveniently handle nonlinear problems, but the identification accuracy depends on the scale and quality of the observational data, and the generalization ability is poor, and the identification accuracy will drop significantly when the working conditions change. Summary of the Invention

[0004] Aiming at the above defects or improvement requirements of the prior art, the present invention provides a method and system for identifying dynamic parameters of a mechatronic system based on an inverse model, which solves the problems of low computational efficiency, poor robustness, and inability to adapt to variable working condition scenarios of existing parameter identification technologies when dealing with mechatronic systems with strong nonlinear characteristics.

[0005] To achieve the above object, according to one aspect of the present invention, a method for identifying dynamic parameters of a mechatronic system based on an inverse model is provided, and the method includes the following steps:

[0006] Simplify the manipulator to be processed into a single-axis manipulator, and thus construct an inverse model of the driving torque of the manipulator to be processed with respect to the moment of inertia;

[0007] Set the desired motion parameters of the robotic arm to be processed, collect the actual motion parameters of the robotic arm to be processed when it moves under the desired motion parameters, and calculate the desired driving torque and the actual driving torque using the inverse model;

[0008] Construct an error function using the desired driving torque and the actual driving torque. Taking the actual moment of inertia of the robotic arm to be processed as a variable, construct an optimization model with the minimum error function as the goal, and solve the optimization model to obtain the actual moment of inertia corresponding to the minimum error function, so as to realize the identification of the dynamic parameters of the robotic arm to be processed.

[0009] Further preferably, the inverse model of the driving torque is as follows:

[0010] ;

[0011] Where, and are the joint driving torque and the load torque of the robotic arm, J is the rated joint moment of inertia of the robotic arm, b is the joint friction coefficient of the robotic arm, m and l are the link mass and the link length of the robotic arm, g is the acceleration due to gravity, , and are the desired joint angle, the desired joint angular velocity, and the desired joint angular acceleration of the robotic arm, are set artificially, and are obtained by the difference method.

[0012] Further preferably, the method of simplifying the robotic arm to be processed into a single-axis robotic arm is: simplify the structure of the robotic arm to be processed into a single-axis robotic arm including a link and a rotary joint, and simplify the multi-axis to the load torque at the end of the link.

[0013] Further preferably, the desired motion parameters are the desired angle, the desired angular velocity, and the desired angular acceleration of the link, and the actual motion parameters are the actual angle, the actual angular velocity, and the actual angular acceleration of the link.

[0014] Further preferably, the optimization model is as follows:

[0015] ;

[0016] Where, and are the actual joint moment of inertia of the robotic arm and its estimated value, N is the number of sampling points, and the subscript iIndicates the i th sampling point, , and are the actual joint rotation angle, actual joint angular velocity, and actual joint angular acceleration of the robotic arm.

[0017] Further preferably, the gradient descent iteration method is used to solve the optimization model.

[0018] Further preferably, the gradient descent iteration method is performed according to the following steps:

[0019] (1) Set the initial value of the actual moment of inertia, and calculate the error function and gradient;

[0020] (2) Update the estimated value of the moment of inertia and the learning rate;

[0021] (3) Repeat step (2) until the preset number of iterations is reached.

[0022] Further preferably, the moment of inertia is updated according to the following formula:

[0023] ;

[0024] where α is the learning rate, which controls the step size of each iteration of the gradient descent method, k indicates the k th iteration, and respectively represent the estimated values of the actual joint moment of inertia of the robotic arm at the k +1 and k th iterations of is a simplified form of , indicating the gradient obtained by taking the partial derivative of the error function k with respect to at the th iteration.

[0025] According to another aspect of the present invention, a dynamic parameter identification system for an electromechanical system based on an inverse model is provided. The system includes an actuator, and the actuator is used to execute the above-mentioned method for identifying dynamic parameters of an electromechanical system based on an inverse model.

[0026] According to yet another aspect of the present invention, a computer-readable storage medium is provided, which includes a computer program, and when the computer program is executed, it implements the above-mentioned method for identifying dynamic parameters of an electromechanical system based on an inverse model.

[0027] Generally speaking, compared with the prior art through the above technical solutions conceived by the present invention, the following beneficial effects are achieved:

[0028] 1. The present invention applies the inverse model technology to model parameter identification, uses the inverse model to establish an optimization model with the minimum error, and conducts reverse modeling on the nonlinear relationship between the system output and input by combining the inverse model, effectively eliminating the influence of nonlinear error propagation in traditional model parameter identification methods. It is especially suitable for electromechanical systems with strong nonlinear characteristics, explicitly reflects the core physical laws of the system, helps reduce the interference of redundant parameters and external disturbances, has strong robustness and generalization ability, and especially has advantages in dealing with variable working condition scenarios.

[0029] 2. This method can effectively improve the identification accuracy and significantly reduce the calculation time. Traditional electromechanical system parameter identification methods usually rely on the quantity and quality of observed data and the setting of iterative initial values, and are easily affected by nonlinear error accumulation and external disturbances when dealing with systems with strong nonlinear characteristics. By combining the inverse model to conduct reverse modeling on the nonlinear relationship between the system output and input, a more intuitive and interpretable model framework is provided, thus reducing the influence of nonlinear error propagation in traditional methods and improving the accuracy of parameter identification. In addition, by combining numerical differentiation methods and the inverse model, the calculation process is optimized, significantly improving the calculation efficiency, avoiding unnecessary redundant calculations, and thus shortening the time of the entire parameter identification process.

[0030] 3. This method has strong robustness and adaptability, and is especially suitable for variable working condition scenarios. When traditional parameter identification methods face changes in the environment or working conditions, they often need to retrain the model or adjust the parameters, resulting in a decrease in identification accuracy or calculation efficiency. The inverse model only models the key parameters related to the system state feedback. The identification method based on the inverse model can eliminate the interference caused by working condition changes and external disturbances by constructing an inverse model. Through this method, good identification performance can be maintained even under complex and dynamically changing working conditions, ensuring that the dynamic parameters of electromechanical systems can be identified efficiently and accurately in different working environments. Description of the Drawings

[0031] Figure 1 It is a schematic diagram of the main process of the electromechanical system dynamic parameter identification method based on the inverse model constructed according to the preferred embodiment of the present invention.

[0032] Figure 2 It is a schematic diagram of the structure of a single-axis robotic arm system of the electromechanical system dynamic parameter identification method based on the inverse model constructed according to the preferred embodiment of the present invention.

[0033] Figure 3 It is a schematic diagram of the main process of using the gradient descent algorithm to perform iterative optimization and solution of dynamic parameters in an exemplary embodiment constructed according to the preferred embodiment of the present invention.

[0034] Figure 4 It is an experimental scenario diagram of a six-axis robotic arm system for the method of identifying dynamic parameters of an electromechanical system based on an inverse model constructed according to a preferred embodiment of the present invention.

[0035] Figure 5 It is a parameter identification result diagram of the moment of inertia of each joint of a six-axis robotic arm under different load conditions constructed according to a preferred embodiment of the present invention. Among them, (a) is the parameter identification result diagram of the moment of inertia of the 6th joint when the angles of the other joints of the six-axis robotic arm are 0°; (b) is the parameter identification result diagram of the moment of inertia of the 5th joint when the angles of the other joints of the six-axis robotic arm are 0°; (c) is the parameter identification result diagram of the moment of inertia of the 5th joint when the 6th joint of the six-axis robotic arm is 5° and the angles of the other joints are 0°; (d) is the parameter identification result diagram of the moment of inertia of the 5th joint when the 6th joint of the six-axis robotic arm is 45° and the angles of the other joints are 0°.

[0036] Figure 6 It is a parameter identification result diagram of the moment of inertia of the 4th joint of a six-axis robotic arm under different load conditions constructed according to a preferred embodiment of the present invention. Among them, (a) is the parameter identification result diagram of the moment of inertia of the 4th joint when the angles of the other joints of the six-axis robotic arm are 0°; (b) is the parameter identification result diagram of the moment of inertia of the 4th joint when the 5th joint of the six-axis robotic arm is 5° and the angles of the other joints are 0°; (c) is the parameter identification result diagram of the moment of inertia of the 4th joint when the 5th joint of a 6-degree-of-freedom robotic arm is 20° and the angles of the other joints are 0°; (d) is the parameter identification result diagram of the moment of inertia of the 4th joint when the 5th joint of the six-axis robotic arm is 45° and the angles of the other joints are 0°. Detailed implementation manners

[0037] In order to make the objectives, technical solutions and advantages of the present invention clearer and more understandable, the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention. In addition, the technical features involved in the various embodiments of the present invention described below can be combined with each other as long as they do not conflict with each other.

[0038] As Figure 1 shown, a method for identifying dynamic parameters of an electromechanical system based on an inverse model specifically includes the following steps:

[0039] S1. Analyze the physical structure of the electromechanical system, derive and establish a forward dynamics model of the electromechanical system based on Lagrange's dynamics equation, and on this basis, establish an inverse dynamics model of the electromechanical system.

[0040] Simplify the electromechanical system to be studied into a single-axis robotic arm. As Figure 1As shown, the single-axis robotic arm consists of a rigid link and a rotary joint. The joint position of the robotic arm, i.e., the joint rotation angle, is represented by θ The joint driving torque of the robotic arm is τ , and the load torque at the end of the robotic arm is τ L . The single-axis robotic arm is composed of the following basic elements:

[0041] Base: The fixed part at the bottom of the robotic arm, usually connected to a support platform or the ground.

[0042] Joint: The rotating axis of the robotic arm, which is its only degree of freedom. The joint is usually driven by a motor or a servo, enabling the robotic arm to rotate around the joint axis.

[0043] Link: The connecting part from the base to the end effector, usually a rigid rod that can bear torque and force.

[0044] End-effector: The component located at the end of the robotic arm, used to perform various tasks. In a single-axis robotic arm, the end effector is usually fixed at the end of the link and can be designed in different forms according to the task requirements.

[0045] Load: Usually an object mounted on the end effector, or the mass of the end effector itself.

[0046] As Figure 2 shown, the kinetic energy of the single-axis robotic arm mainly comes from the rotational motion of the robotic arm. The kinetic energy T of the robotic arm can be expressed as:

[0047] .

[0048] Where J is the moment of inertia of the robotic arm, and is the joint angular velocity of the robotic arm.

[0049] The potential energy of the single-axis robotic arm mainly comes from the gravity of the robotic arm. Assuming that the gravity acts on the center of mass of the robotic arm, the center of mass of the robotic arm is located at the midpoint of the link, and the distance from the rotation axis is l / 2. The potential energy V of the robotic arm can be expressed as:

[0050] .

[0051] Where m is the mass of the link of the robotic arm, g is the acceleration due to gravity, l is the length of the link of the robotic arm, and θ is the joint rotation angle of the robotic arm.

[0052] Lagrangian function L is the kinetic energy T minus the potential energy V :

[0053] .

[0054] The form of the Lagrangian equation is:

[0055] .

[0056] Wherein, F is the generalized force. and are calculated as follows:

[0057] ;

[0058] .

[0059] Then, the time derivative part of the Lagrangian equation is calculated according to Equation (5):

[0060] .

[0061] Substituting Equation (6) and Equation (7) into Equation (4), the Lagrangian equation of the uniaxial robotic arm is obtained as:

[0062] .

[0063] For Figure 2 the uniaxial robotic arm shown in F the generalized force

[0064] .

[0065] Wherein, is the joint angular acceleration of the robotic arm, b is the joint friction coefficient of the robotic arm, τ L is the load torque at the end of the robotic arm, which comes from the operation task of the robotic arm or an external load.

[0066] Through the Lagrangian method, the mathematical model of the uniaxial robotic arm is obtained. This model includes the effects of joint inertia, friction, gravity, and external load torque. This equation can be used to analyze and control the motion of the robotic arm, especially playing a key role in subsequent inverse dynamics solution and parameter identification.

[0067] For a single-axis robotic arm, the forward dynamics problem refers to: Given the joint driving torque applied to the robotic arm by the actuator at a certain moment τ , as well as the joint rotation angle θ and joint angular velocity at that instant, according to Equation (9), find the joint angular acceleration of the robotic arm at this time; the inverse dynamics problem refers to: Given the joint rotation angle θ , joint angular velocity and joint angular acceleration of the robotic arm at a certain moment, according to Equation (9), find the driving torque τ applied to the robotic arm by the actuator at this time.

[0068] Placing the unknown quantity to be solved on the left side of the equation and the known quantity on the right side, the forward model and inverse model of the single-axis robotic arm can be described as follows:

[0069] .

[0070] S2. Set different control signals for the electromechanical system, and collect and calculate experimental data based on sensors and numerical differentiation methods.

[0071] S2.1 Set different control signals (i.e., the desired rotation angle of the system) for the electromechanical system, calculate the desired angular velocity and desired angular acceleration of the system based on the numerical differentiation method, and substitute them into the inverse dynamics model established in step S1 to calculate the desired input of the system (i.e., the desired driving torque of the system). Apply different control signals to the single-axis robotic arm, calculate the required experimental data through the numerical differentiation method, and collect the actual output observation values of the system through sensors.

[0072] In this embodiment, it is the desired joint rotation angle of the single-axis robotic arm. Substitute the set control signal and the rated structural parameters of the single-axis robotic arm into the inverse model of the single-axis robotic arm system established in Equation (10) of step S1 to calculate the desired input of the system, that is, calculate the driving torque that the actuator needs to apply to the robotic arm joint in order to drive the single-axis robotic arm to reach the desired joint rotation angle. The specific implementation steps are as follows:

[0073] Set the control signal as the desired joint rotation angle θ d , and are the desired joint angular acceleration and desired joint angular acceleration respectively. In the motion analysis of the robotic arm, the numerical differentiation method can be used to approximately calculate the angular velocity and angular acceleration at each moment according to the rotation angle. The numerical differentiation methods include the forward difference method, the backward difference method, and the central difference method, among which the central difference method can achieve more accurate calculations. The formula for calculating using the central difference method is as follows:

[0074] 。

[0075] Assume that the total number of time steps in the time series is N , where i ∈ [0, N - 1] is the index of the time series, θ d,i is the expected joint angle at the corresponding moment, and Δ t = t i+1 - t i is the time interval at each moment.

[0076] After calculating the expected joint angular velocity, the expected joint angular acceleration can be further calculated using the central difference method:

[0077] 。

[0078] Similarly, is the expected joint angular velocity at the corresponding moment.

[0079] For the first moment ( i = 0) and the last moment ( i = N - 1), since there is no data for the two adjacent moments, the forward difference or backward difference method can be used to calculate the angular velocity and angular acceleration.

[0080] For the first time step, the forward difference method is used to calculate the angular acceleration based on the current value and the next value of the angular velocity:

[0081] 。

[0082] For the last time step, the backward difference method is used to calculate the angular acceleration based on the current value and the previous value of the angular velocity:

[0083] 。

[0084] Set the rated joint moment of inertia of the robotic arm to J , and substitute θ d and the calculated and into the inverse model of Equation (10) to calculate the system input, that is, the driving torque τ exerted by the actuator on the robotic arm joint is:

[0085] 。

[0086] S2.2 Apply the expected driving torque of the system calculated in step S2.1 to the electromechanical system through the actuator, and collect and record the rotation angles of the actual robotic arm of the system at each time step through the sensor.

[0087] According to the expected input of the system calculated in step S2.1, apply the joint driving torque to the real single-axis robotic arm system through the actuator τ , observe the response of the real robotic arm device, and record the relevant real observation values. The specific implementation steps are as follows:

[0088] Apply the joint driving torque calculated in step S2.1 τ to the rotating joint of the real single-axis robotic arm through the actuator. The real robotic arm responds to the input joint torque and generates a rotation around the joint axis. Collect the actual joint rotation angles of the single-axis robotic arm at each moment through the angle sensor θ a,i , and approximately solve the actual joint angular velocity of the single-axis robotic arm at each moment through numerical differentiation method with reference to equations (12) to (14) and the actual joint angular acceleration .

[0089] Record the expected joint rotation angles θ d,i , expected joint angular velocities and expected joint angular accelerations , actual joint rotation angles θ a,i , actual joint angular velocities and actual joint angular accelerations of the single-axis robotic arm at each moment through sensor collection and numerical differentiation calculation for subsequent parameter identification operations.

[0090] S3. Select the dynamic parameters of the electromechanical system to be identified, define the error function of the electromechanical system according to the parameters to be identified, and eliminate the operating condition terms based on the inverse model. On this basis, define the optimization objective function;

[0091] S3.1. Based on the system inverse model constructed in step S1, calculate the inverse model solution results of the expected system rotation angle and the actual system rotation angle respectively, analyze the reasons for the deviation between the two results, and thus select the dynamic parameters that need to be identified.

[0092] In the inverse model parameter identification, substitute the actual joint rotation angle θ a into the inverse model calculation formula of equation (10) to obtain:

[0093] .

[0094] Theoretically, both the left sides of equations (15) and (16) are τ , so the right sides of the two equations should be equal, that is

[0095] .

[0096] Solving gives θ d = θ a , but this result does not conform to the actual situation: in actual operation, the actual rotation angle θ a of the single-axis robotic arm joint is always θ d there is a certain deviation from the expected joint rotation angle, so equation (16) does not hold, that is .

[0097] It is inferred that this is due to the deviation between the moment of inertia J a of the actual equipment and the rated value J (that is J a ≠ J ). Therefore, the joint moment of inertia of the single-axis robotic arm is selected as the dynamic parameter to be identified.

[0098] The moment of inertia J a of the robotic arm joint of the actual equipment is an unknown parameter to be identified, and equation (16) is changed to

[0099] ;

[0100] S3.2 Define an error function based on the parameter to be identified selected in step S3.1, eliminate the working condition terms based on the inverse model, and determine the optimization objective and define the optimization objective function according to the error function.

[0101] Based on the dynamic parameter to be identified selected in step S3.1 J a , define the error function of the single-axis robotic arm related to J a , and eliminate the load term τ L to eliminate the influence of load condition changes, and determine the optimization objective and define the optimization objective function according to the error function. The following are the specific implementation steps:

[0102] Combining equations (15) and (18), define the error function E( J a ) related to J a to measure the parameter estimation error:

[0103] ;

[0104] wherein, E( J a ) represents the total error over all data points, N is the number of sampling points. In this error function, the desired joint angle i at time θ d,i is given manually, while the actual joint angle i at time θ a,i is obtained by the actual single-axis robotic arm system responding to θ d,i .

[0105] In addition, as can be seen from Equation (19), the error function E( J a ) finally eliminates the load term τ L , that is, the error function is not affected by changes in operating conditions, and the identification technology based on the inverse model can effectively realize the identification of the dynamic parameters of a single-axis robotic arm under variable load conditions.

[0106] To find the estimated value J a of the joint moment of inertia of the single-axis robotic arm that is closest to the actual moment of inertia , it is necessary to minimize the error function E( J a ), so the optimization objective function is defined as:

[0107] ;

[0108] Usually, the best estimated value J a of the actual moment of inertia J a can be obtained by solving the derivative of the error function E( J a ) with respect to and setting it to zero.

[0109] S4. Perform iterative optimization and solution of the parameters to be identified based on the gradient descent method to obtain the best estimated values of the parameters;

[0110] S4.1 Based on the error function constructed in step S3.2, calculate the gradient of the error function by taking the derivative of the parameters to be identified, and set the minimization condition to construct the minimization condition equation.

[0111] Iteratively solve the optimization objective function constructed in step S3.2 using the gradient descent method to obtain the best estimated value of the kinetic parameters to be identified selected in step S3.1. S4 contains two sub-steps S4.1 and S4.2. In step S4.1, the gradient of the error function is calculated by taking the first derivative of the error function defined in step S3 with respect to the kinetic parameters to be identified, and the minimization boundary condition is solved by setting the gradient to zero. The following are the specific implementation steps:

[0112] Calculate the gradient of the error function E( J a ) with respect to J a , which can be obtained by calculating the first derivative of E( ) with respect to J a : J a :

[0113] ;

[0114] Let

[0115] ;

[0116] Equation (21) can be simplified to:

[0117] ;

[0118] According to the chain rule, equation (23) can be expanded as:

[0119] ;

[0120] Among them, , substituting equation (22) into equation (24), the calculation shows:

[0121] ;

[0122] Set the gradient to zero to solve the minimization condition, that is, let ∇ Ja E( J a ) = ∂E( J a ) / ∂ J a = 0, substituting into equation (25) gives:

[0123] ;

[0124] By solving the minimization condition equation (26), find the best parameter estimate value that minimizes the error function E( J a ) 。

[0125] Specifically, under normal circumstances, the minimization condition equation can be solved by a linearization method. If there are multiple unknowns (such as b or l / 2 ·mg ), they can be regarded as constants to solve multiple parameters simultaneously.

[0126] However, in the problems of electromechanical systems with strong nonlinear characteristics, the minimization condition equation contains nonlinear terms, and it is not easy to achieve an analytical solution. Therefore, numerical optimization methods, such as the gradient descent method or the Newton method, are usually adopted to iteratively solve the optimal parameter estimation value of this equation. Among them, the gradient descent method is very suitable for dealing with such nonlinear optimization problems, and its main advantages include: the gradient descent method can handle complex dynamic equations, especially parameter identification problems; the gradient descent method only requires the gradient information of the objective function, which is easy to calculate and implement; the gradient descent method is easy to extend to problems with multiple parameters (multidimensional optimization).

[0127] S4.2 Use the gradient descent method to perform parameter iterative solution on the minimization condition equation constructed in step S4.1. Finally, when the error function converges or reaches the maximum number of iterations, the best estimated value of the parameter to be identified is obtained.

[0128] Use the gradient descent method to perform iterative optimization solution of the actual moment of inertia parameter estimate value based on the minimization condition equation (26) obtained in step S4.1. As Figure 3 shown, the following are the specific implementation steps:

[0129] Initialize the parameters: Start from an initial value (usually random or based on empirical estimation), that is, ;

[0130] Calculate the gradient: Calculate the gradient of the error function with respect to according to Equation (25), and use the simplified form for substitution;

[0131] Update the parameter estimate value: Update the parameter estimate value based on the gradient and the learning rate according to Equation (27):

[0132] ;

[0133] where α is the learning rate, which controls the step size of each iteration, and k represents the k th iteration;

[0134] Update the learning rate: Substitute the current parameter estimate value, and according to the error function Update the learning rate according to the value size;

[0135] Repeat the iteration: until the error function converges or reaches the set maximum number of iterations M ;

[0136] The beneficial effects of the present invention will be verified in combination with specific embodiments as follows:

[0137] Select a six-axis robotic arm system to verify the method for identifying the dynamic parameters of an electromechanical system based on the inverse model. The experimental scenario is set up as Figure 4 shown. Among them, the object to be parameter-identified is a serial robotic arm with six degrees of freedom. An inverse model of the robotic arm is built based on the Lagrangian dynamics equation, and a parameter identification device for estimating the actual rotational inertia of the robotic arm joints is built. J a The actuator is used to apply the joint driving torque obtained by solving the inverse model to the robotic arm, and an angle sensor is used to collect the actual joint rotation angles of the robotic arm at each moment τ . The control signal, that is, the desired joint rotation angle of the robotic arm θ a , is set through the controller. The control signal is set as a parabola θ d () = 1000 y ( t ) = 1000 ·t 2 for simulation.

[0138] When identifying the actual rotational inertia of one joint of the six-axis robotic arm, the structures serially connected to this joint later can be regarded as the load of this joint. At this time, the system can be simplified to a single-axis robotic arm with a load torque Figure 2 as shown. Different angles can be set for joints 1 to 6 of the robotic arm, that is, different load conditions for the robotic arm joints. τ L As shown in

[0139] and Figure 5 and 6 , according to the method for identifying the dynamic parameters of an electromechanical system based on the inverse model provided in the embodiments of the present application, after iterative optimization, the estimated values of the rotational inertia parameters of each joint can gradually converge to stable values when the six-axis robotic arm is under different load conditions .

[0140] It is easy for those skilled in the art to understand that the above are only preferred embodiments of the present invention, and are not used to limit the present invention. Any modifications, equivalent replacements, and improvements made within the spirit and principle of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for identifying dynamic parameters of an electromechanical system based on an inverse model, characterized in that, The method includes the following steps: Simplify the manipulator to be processed into a single-axis manipulator, and thus construct an inverse model of the driving torque with respect to the moment of inertia of the manipulator to be processed; Set the expected motion parameters of the manipulator to be processed, collect the actual motion parameters of the manipulator to be processed when it moves under the expected motion parameters, and calculate the expected driving torque and the actual driving torque by using the inverse model; Construct an error function by using the expected driving torque and the actual driving torque, construct an optimization model with the actual moment of inertia of the manipulator to be processed as a variable and the minimum of the error function as the goal, solve the optimization model to obtain the actual moment of inertia corresponding to the minimum of the error function, so as to realize the identification of the dynamic parameters of the manipulator to be processed.

2. The method for identifying dynamic parameters of an electromechanical system based on an inverse model according to claim 1, wherein, The inverse model of the driving torque is as follows: Among them, τ and are the joint driving torque and the load torque of the robotic arm, J is the rated joint moment of inertia of the robotic arm, b is the joint friction coefficient of the robotic arm, m and l are the link mass and the link length of the robotic arm, g is the acceleration due to gravity, θ d 、 and are the desired joint angle, the desired joint angular velocity and the desired joint angular acceleration of the robotic arm, which are set artificially, and are obtained by the difference method.

3. The method for identifying dynamic parameters of an electromechanical system based on an inverse model according to claim 2, characterized in that The method of simplifying the manipulator to be processed into a single-axis manipulator is: simplify the structure of the manipulator to be processed into a single-axis manipulator including a connecting rod and a rotating joint, and simplify the multi-axis to the load torque at the end of the connecting rod.

4. A method for identifying dynamic parameters of an electromechanical system based on an inverse model according to claim 1 or 2, characterized in that, The expected motion parameters are the expected rotation angle, expected angular velocity and expected angular acceleration of the connecting rod, and the actual motion parameters are the actual rotation angle, actual angular velocity and actual angular acceleration of the connecting rod.

5. A method for identifying dynamic parameters of an electromechanical system based on an inverse model according to claim 1 or 2, characterized in that The optimization model is as follows: Among them, J a and are the actual joint moment of inertia of the robotic arm and its estimated value, N is the number of sampling points, and the subscript i represents the i th sampling point, θ a 、 and are the actual joint angle, actual joint angular velocity, and actual joint angular acceleration of the robotic arm, m and l are the link mass and link length of the robotic arm, g is the acceleration due to gravity.

6. A method for identifying dynamic parameters of an electromechanical system based on an inverse model according to claim 1 or 2, characterized in that, The gradient descent iteration method is used to solve the optimization model.

7. A method for identifying dynamic parameters of an electromechanical system based on an inverse model according to claim 6, characterized in that, The gradient descent iteration method is carried out according to the following steps: (1) Set the initial value of the actual moment of inertia, and calculate the error function and the gradient; (2) Update the estimated value of the moment of inertia and the learning rate; (3) Repeat step (2) until the preset number of iterations is reached.

8. The method for identifying dynamic parameters of an electromechanical system based on an inverse model according to claim 7, characterized in that, The update of the moment of inertia is carried out according to the following formula: Among them, α is the learning rate, which controls the step size of each iteration of the gradient descent method. k represents the k th iteration. and respectively represent the estimated values of the actual joint moment of inertia of the robotic arm at the k +1 and k th iterations. J a is the estimated value. is in a simplified form, representing the gradient obtained by taking the partial derivative of the error function k with respect to at the th iteration.

9. An electromechanical system dynamic parameter identification system based on an inverse model, characterized in that The system includes an actuator, and the actuator is used to execute a method for identifying dynamic parameters of an electromechanical system based on an inverse model according to any one of claims 1-8.

10. A computer-readable storage medium, including a computer program, characterized in that, When the computer program is executed, it realizes a method for identifying dynamic parameters of an electromechanical system based on an inverse model according to any one of claims 1-8.

Images