Robot dynamics identification method considering load static separation and dynamic friction unified prediction

Abstract

The invention belongs to the technical field of robots, and discloses a robot dynamics identification method considering load static separation and dynamic friction unified prediction. A load dependent friction static separation method (LFSS); carrying out dynamic friction modeling based on HTFO; designing a PRF Fourier series excitation trajectory; and verifying the model. According to the method, a set of complete robot dynamics identification technology system is constructed, a convenient tool and platform are provided for popularization and use of the robot dynamics identification technology system in practical application, and development and application of the robot technology under complex working conditions are greatly promoted.

Description

Technical Field

[0001] The present invention belongs to, but is not limited to, the field of robotics technology, and particularly relates to a method for identifying robot dynamics by considering unified prediction of load static separation and dynamic friction. Background Art

[0002] During the operation of a robot, significant changes in load, friction, and interference pose high requirements for its dynamic adaptability. However, there are many challenges in determining the dynamic parameters of a robot in the prior art:

[0003] 1. In traditional model-based experimental systems during dynamic identification, it is usually difficult to accurately consider the influence of load force and joint friction on system characteristics because their mechanisms are unclear, resulting in difficulty in accurately determining the dynamic parameters of the robot.

[0004] 2. In actual drive systems, common linear friction models (such as Coulomb-viscous friction models) cannot accurately describe joint friction due to the non-linear characteristics of friction. Especially at low speeds, large identification errors will be caused.

[0005] 3. Although there are some non-linear friction models, these models are difficult to integrate into dynamic identification systems and often have large deviations in the zero-speed range. In addition, there is a strong correlation between the load and friction of robot joints. External loads will couple friction with inertial forces, resulting in inaccurate identification of dynamic parameters. Moreover, existing methods for separating non-linear friction usually rely on initially identified inertial parameters or multiple iterations, making it difficult to effectively decouple inertial and friction parameters. Summary of the Invention

[0006] Aiming at the problems existing in the prior art, the present invention provides a method for identifying robot dynamics by considering unified prediction of load static separation and dynamic friction.

[0007] The present invention is implemented as follows. A method for identifying robot dynamics by considering unified prediction of load static separation and dynamic friction, the system comprising:

[0008] S1: Non-linear friction pre-identification;

[0009] S2: Load-related friction static separation method (LFSS);

[0010] S3: Dynamic friction modeling based on HTFO;

[0011] S4: PRF Fourier series excitation trajectory design;

[0012] S5: Model verification.

[0013] Furthermore, the specific content of S1 includes:

[0014] In view of the fact that in the actual operation of industrial robots, especially during low-speed movement, the friction characteristics have a crucial impact on their performance, and traditional linear friction models, such as the Coulomb-viscous friction model, cannot accurately depict the actual friction behavior, so it is modified as follows:

[0015] The original linear friction model is:

[0016]

[0017] Among them, F c is the Coulomb friction, and F s represents the static friction force. At the moment when the robot starts and during the low-speed operation stage, the static friction force plays an important role. This model only simply considers the linear relationship between Coulomb friction and viscous friction with velocity and does not cover complex friction phenomena such as the Stribeck effect. The revised model is:

[0018]

[0019] Among them, q s is the Stribeck velocity, which determines the key turning point of the friction change with velocity; b is the damping decay coefficient, which reflects the decay characteristics of the non-linear viscous force at different velocities; m is an empirical constant (m = 1 corresponds to the Tustin model, m = 2 corresponds to the Gauss model), and different m values correspond to different friction velocity change laws.

[0020] To accurately identify the friction parameters, a special joint trajectory is designed. In these trajectories, the joint acceleration is strictly guaranteed to be zero, and the friction is induced by changing the joint velocity direction. The friction force data of the joint at different velocities are collected by using high-precision sensors, and then the particle swarm optimization (PSO) method is used for parameter pre-identification. For example, for the HS-CO610 robot, through a large number of experiments and complex calculations, the detailed non-linear friction parameters of each joint are obtained. Taking Joint1 as an example, its γ4(F c ) = 1.200, which shows the characteristics of this joint in terms of Coulomb friction; γ1(F s -F c ) = 5.848, which reflects the difference relationship between the static friction force and the Coulomb friction force. These parameters provide an important basis for subsequent accurate friction modeling.

[0021] Furthermore, the specific content of S2 includes:

[0022] (1) Construction of the static separation method: By deeply studying the characteristics of the robot in a high-load and high-dynamic operation environment, it can be known that the Coulomb friction coefficient F cIt will change significantly with the transmission power and external forces; in precise robot dynamic analysis, it is necessary to accurately identify the friction model and inertial parameters separately, and subtract the joint friction from the joint torque to determine the inertial torque; however, due to the mutual coupling of the forces / torques generated by inertial motion and joint friction, the inertial parameter model deviates significantly from the actual value, greatly affecting the identification accuracy;

[0023] For this reason, an LFSS method is proposed; this method models F c as a polynomial function of the four-quadrant dynamics and external forces. Through the analysis of a large amount of experimental data and theoretical derivation, the key expression is obtained:

[0024] F ci (q,τ outi )=α i |τ outi |+β i

[0025] where α i ,β i are coefficients related to joint i, which reflect the linear relationship between joint friction and load torque; for the rotating joints of serial robots, through a large number of experiments and theoretical analyses, it is found that only the load torque aligned with the joint rotation axis will affect the joint friction, which mainly includes the torque generated by gravity and the load torque externally applied to the robot; when the robot is in a static equilibrium state, since the angular velocity can be approximately regarded as zero, that is at this time and τ dyni ≈G(q), thus obtaining the static model of the robot considering load-related static friction:

[0026] τ=KI=(1±α)·(G(q)+τ ext )+β

[0027] where K is the current constant and I is the motor input current. On this basis, by designing experiments, only by obtaining the data of the external load change and current change of the robot under static conditions, the parameters of the load-related friction model can be efficiently identified using the linear regression method;

[0028] (2) Implementation of the identification process: First, according to the structure and motion characteristics of the robot, the static loading positions q s are carefully designed; in this process, the influence of coordinate system changes on the experimental results is fully considered, through the coordinate transformation formulas:

[0029]

[0030] and

[0031]

[0032] Accurately calculate the representation of the external force in the joint space; where, F ext represents the generalized external force, which consists of the force f ext and the torque m ext and is composed of, J T is the robot Jacobian matrix, obtained by the vector product method with respect to the base coordinate G;

[0033] Next, using the above formula, by converting the force coordinates from T to G, the external force change Δ S F ext and the exact relationship with the joint torque change Δτ ext are obtained:

[0034]

[0035] Based on this relationship, a series of static load experiments are designed by optimizing the conditions:

[0036] q=arg{q1,q2Lq n}max(‖Δτ ext ‖ 2

[0037] subject to:{‖Δ S f ext ‖2=1,Δ S m ext =1}

[0038] And the particle swarm optimization (PSO) algorithm is used to solve it, and the optimal posture of the robot in the static loading experiment is determined to maximize the loading effect of the robot and obtain sufficiently accurate experimental data;

[0039] Then, based on the data collected in the experiment, the parameters of the load-related friction model are accurately fitted using the least squares method (LS); through careful analysis and complex calculations of a large amount of data, the parameter fitting results of each joint are obtained; the recognition results are evaluated by calculating the normalized root mean square error, and the results show that this method can accurately identify the friction load-related coefficients, with an overall error of about 10%, effectively avoiding the coupling effect of inertia and friction parameters, and providing a reliable basis for subsequent dynamic modeling.

[0040] Furthermore, the specific content of S3 includes:

[0041] Deeply study the existing dynamic friction models and find that although the LuGre model can describe the dynamic characteristics of friction to a certain extent, there are still some deficiencies in robot applications; in order to better adapt to the actual operation of the robot, the LuGre model is simplified and improved; the original LuGre model is:

[0042]

[0043] Among them, σ is the dynamic friction coefficient and z is the internal variable. This model has certain complexity and computational difficulty in describing the friction of robot joints, and its performance in key regions such as zero-speed crossing is not ideal. By introducing the Laplace operator s and the hyperbolic tangent function, it is transformed into a dynamic friction model based on the first-order hyperbolic tangent HTFO:

[0044]

[0045] After such improvement, this model has significant advantages in describing the dynamic friction behavior of robot joints. Especially within the zero-speed crossing range, it can more accurately reflect the continuous characteristics of actual friction, effectively reducing the complexity of dynamic parameter identification. Through in-depth analysis and derivation of the robot motion equation, combined with the improved friction model above, a robot dynamic model is established, realizing the effective decoupling of friction-inertia, and determining the key parameters to be identified, providing a theoretical basis for subsequent parameter identification.

[0046] Furthermore, the specific steps of S4 include:

[0047] Aiming at the limitations of traditional time-domain excitation methods in identifying dynamic friction, a pseudo-random finite PRF Fourier series excitation trajectory with time-frequency constraints is proposed. Based on the traditional synthesis of periodic excitation references by finite Fourier series, such as:

[0048]

[0049] Deeply considering the dynamic characteristics of robot joints, the excitation current expression is designed:

[0050]

[0051] Among them, is used to implement the finite Fourier excitation trajectory, and its basic shape and parameters of the trajectory are determined by its calculation method; the PRBS pseudo-random binary sequence is used to effectively stimulate friction, and it generates a series of pulse signals through specific coding rules to stimulate different friction responses of robot joints. By deeply analyzing the frequency response of robot joints, using the expression of the linearized HTFO friction model in the frequency domain:

[0052]

[0053] By estimating the frequency response function, the accurate identification of the dynamic friction parameter σ0 is achieved; on the HS-CO610 robot experimental platform, through careful experimental design and data analysis, it is verified that this excitation design can effectively capture joint friction dynamics during the standard dynamic parameter excitation and identification process.

[0054] Furthermore, the S5 specifically includes:

[0055] A complete experimental system is carefully built on the HS-CO610 robot experimental platform. This platform includes the robot body, an adjustable load device that can accurately load within the range of 1 - 4 kg, a high-performance controller, high-precision sensors and other key components; through the collaborative work of these components, precise control and data acquisition of the robot under different loads and motion conditions can be achieved;

[0056] During the verification process, a variety of evaluation indicators are used to comprehensively analyze the experimental results; for the verification of the load-related friction model, the dynamic identification errors with and without the load-dependent model are mainly compared, and the e NRMS normalized root mean square error and e PE peak error and other indicators are adopted; the experimental results show that the LFSS method can significantly improve the identification accuracy; for example, in the case of the minimum load VT1, the sum of the values of the LFSS algorithm is reduced by 8.21% compared with the standard system without load separation, and the peak error and the sum of the values are reduced by 44.21%; at the maximum load VT4, there are improvements of 30.77% and 62.27% respectively; moreover, the fitting results of the LFSS method for each axis under different loads are more consistent, especially for joints with obvious friction-load correlation such as Joint4, the e NRMS value is reduced by more than 49%, effectively reducing the error;

[0057] In terms of the verification of dynamic friction separation, the performances of the Coulomb, tanh and HTFO friction models in predicting joint friction torque are compared; through the analysis and calculation of a large amount of experimental data, the comparison results of the predicted torque and the actual torque of each model, as well as the relative error and time cost, are obtained; the results show that the HTFO model has the best comprehensive performance under different working conditions.

[0058] Furthermore, the experimental process of the S5 specifically includes:

[0059] (1) Under different load conditions, the LFSS method is used to accurately obtain the parameters; by carefully analyzing the force and current changes of the robot joints under different loads, combined with the previously established models and algorithms, accurate parameter values are obtained; for example, under a 1 kg load, the α1 and β1 values of Joint1 are significantly different from those under a 4 kg load, and these differences reflect the influence law of the load on the friction parameters;

[0060] (2) According to the continuous friction model, by collecting joint trajectory data at different speeds, accurately calculate the non-linear friction, and obtain the parameter [γF v b]; in this process, using high-precision sensors and advanced data acquisition systems, the motion data of the robot joints under different speed and load combinations was recorded. Through complex calculations and analyses, the non-linear friction parameters of each joint were obtained;

[0061] (3) Based on the PRF Fourier series excitation trajectory and frequency domain separation method, construct a regression matrix for identifying dynamic friction and inertial characteristics Through the processing and analysis of a large amount of experimental data, using mathematical methods and algorithm optimization, a regression matrix that can accurately reflect the dynamic characteristics of the robot was obtained;

[0062] (4) Using the parameters and matrices obtained above, through a synchronous identification algorithm such as the least squares method, accurately identify the dynamic friction parameter σ0 and inertial parameter δ of the HTFO model l , thus obtaining the final robot dynamics model δ l .

[0063] Another object of the present invention is to provide a computer device, a computer-readable storage medium, and an information data processing terminal, characterized in that the computer device includes a memory and a processor, the memory stores a computer program, and when the computer program is executed by the processor, the processor executes the steps of the robot dynamics identification method considering load static separation and dynamic friction unified prediction.

[0064] Another object of the present invention is to provide a computer-readable storage medium storing a computer program, and when the computer program is executed by a processor, the processor executes the steps of the robot dynamics identification method considering load static separation and dynamic friction unified prediction.

[0065] Another object of the present invention is to provide an information data processing terminal, which integrates a system for the steps of the robot dynamics identification method considering load static separation and dynamic friction unified prediction; in this terminal, the robot system can collect various data during its operation in real time, including joint motion data, load data, friction force data, etc., and transmit these data to the processing module of the terminal.

[0066] Combined with the above technical solutions and the technical problems solved, the advantages and positive effects of the technical solutions to be protected by the present invention are:

[0067] By providing computer devices and information data processing terminals, the present invention not only constructs a complete set of robot dynamics identification technology systems, but also provides convenient tools and platforms for its popularization and use in practical applications, greatly promoting the development and application of robot technology under complex working conditions.

[0068] 1. Through the innovative LFSS method, the high-precision static separation of load-related friction is successfully achieved. Thus, the dilemma of friction-inertia coupling is effectively broken, and the accuracy of the robot dynamics regression model is greatly improved. Verified by experiments, the fitting results of each axis under different loads are more consistent compared with traditional methods.

[0069] 2. The adopted HTFO dynamic friction model combined with pre-identified parameters can extremely accurately describe the dynamic friction behavior of robot joints under various working conditions. During the movement of the robot, especially in critical and complex areas such as zero-speed crossing, it can more accurately reflect the continuous characteristics of actual friction, effectively reducing the complexity of dynamic parameter identification and significantly improving the accuracy of robot dynamics parameter identification.

[0070] 3. In the non-linear friction pre-identification link, a special joint trajectory is designed to ensure that the joint acceleration is zero and the joint velocity direction is changed to measure friction. Then, high-precision sensors are used to collect data and the particle swarm optimization (PSO) method is adopted for parameter pre-identification. This method can quickly and accurately obtain non-linear friction parameters. Compared with traditional identification methods, it greatly improves the identification efficiency, saves time and computing resources for subsequent overall dynamics parameter identification. At the same time, it overcomes the limitations of traditional algorithms, making the whole identification process more efficient.

[0071] The expected benefits and commercial values after the transformation of the technical solution of the present invention are:

[0072] The present invention effectively improves the accuracy and efficiency of parameter identification of robots under load and friction change conditions through innovative methods, thereby enhancing the motion control stability and operation accuracy of robots. By accurately identifying dynamics parameters, errors and equipment losses in the production process are effectively reduced. For example, in high-precision manufacturing links, the product qualification rate can be greatly improved. With the increasingly wide application of robots, its role in enhancing the added value of equipment performance will bring continuous economic benefits. Whether through technology transfer or equipment upgrade services, it can occupy an important position in the market and has extremely high commercial value.

[0073] The technical solution of the present invention fills the domestic and foreign industry technical gaps:

[0074] In the field of robot dynamics research, despite numerous related studies, there has been a lack of effective solutions to key issues related to load and friction. The unique identification system proposed in this invention successfully fills the long-standing technical gap in this field by relying on the LFSS method to accurately handle friction-inertia coupling, the HTFO model to accurately describe dynamic friction, and the PRF excitation trajectory to achieve efficient identification.

[0075] The technical solution of the present invention solves the technical problems that people have been eager to solve but have never been able to solve successfully:

[0076] For a long time, the problem of dynamic performance fluctuation caused by load and friction changes faced by robots during actual operation has been a key technical problem that needs to be solved in the industry. This invention deeply analyzes the load and friction mechanism, constructs an innovative model and identification method, and can achieve high-precision dynamic parameter identification under different loads (covering 1-4kg and a wider range) and speed conditions, effectively suppressing friction torque and joint reversal errors, greatly improving the robot's motion accuracy and stability, and meeting the industry's long-standing technical expectations.

[0077] The technical solution of the present invention overcomes technical prejudice:.

[0078] Previous studies have been limited to a certain extent by their reliance on traditional linear friction models, and there are difficulties in integrating dynamic friction models with the overall dynamics system of robots, which has led to technical bias. Many researchers believe that traditional linear friction models have certain applicability in robot dynamics analysis, but they lack understanding of their limitations under complex working conditions. This invention breaks through the constraints of these traditional concepts, and through innovative revisions of traditional friction models and effective improvements and integration of dynamic friction models, it proves the feasibility and superiority of new technical paths, brings new ideas and methods to robot dynamics research, and promotes the innovative development of industry technology. BRIEF DESCRIPTION OF THE DRAWINGS

[0079] Figure 1 This is the process of the robot dynamics identification system provided by the embodiment of the present invention;

[0080] Figure 2 is a schematic diagram of a robot pre-identification method provided by an embodiment of the present invention;

[0081] Figure 3 It is a method for joint force and coordinate system transformation in the LFSS method provided by an embodiment of the present invention;

[0082] Figure 4 The embodiment of the present invention provides a comparison of the performance of HTFO and the traditional friction model under different conditions;

[0083] Figure 5It is the joint frequency response result of the PRF excitation trajectory design provided by the embodiments of the present invention;

[0084] Figure 6 It is the hardware architecture of the robot system provided by the embodiments of the present invention;

[0085] Figure 7 It is the comparison of the dynamic parameters before and after the present invention under different friction models provided by the embodiments of the present invention;

[0086] Figure 8 It is to compare the parameter identification accuracy before and after the present invention under different loads provided by the embodiments of the present invention. Detailed implementation manners

[0087] In order to make the objectives, technical solutions and advantages of the present invention clearer, the present invention will be further described in detail below with reference to embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention, but not to limit the present invention.

[0088] As Figure 1 shown, the embodiments of the present invention provide a robot dynamics identification method considering load static separation and unified prediction of dynamic friction. The system includes:

[0089] S1: Nonlinear friction pre-identification;

[0090] S2: Load-related friction static separation method (LFSS);

[0091] S3: Dynamic friction modeling based on HTFO;

[0092] S4: PRF Fourier series excitation trajectory design;

[0093] S5: Model verification.

[0094] Furthermore, the specific content of S1 includes:

[0095] As Figure 2 shown, perform nonlinear friction pre-identification. Considering that in the actual operation of industrial robots, especially at low speeds, the friction characteristics have a key impact on their performance, and traditional linear friction models, such as the Coulomb-viscous friction model, cannot accurately describe the actual friction behavior, so it is modified:

[0096] The original linear friction model is:

[0097]

[0098] This model only simply considers the linear relationship between Coulomb friction and viscous friction and velocity, and does not cover complex friction phenomena such as the Stribeck effect; the revised model is:

[0099]

[0100] Among them, F s represents the static friction force, which plays an important role at the moment when the robot starts and during the low-speed operation stage; q s is the Stribeck speed, which determines the critical turning point of the friction change with speed; b is the damping decay coefficient, reflecting the decay characteristics of the non-linear viscous force at different speeds; m is an empirical constant (m = 1 corresponds to the Tustin model, (m = 2) corresponds to the Gauss model), and different m values correspond to different friction speed change laws;

[0101] To accurately identify the parameters in Equation 2, a special joint trajectory is designed; in these trajectories, the joint acceleration is strictly ensured to be zero, and the friction is induced by changing the joint speed direction; the friction force data of the joint at different speeds are collected by using a high-precision sensor, and then the particle swarm optimization (PSO) method is used for parameter pre-identification; for example, for the HS-CO610 robot, through a large number of experiments and complex calculations, the detailed non-linear friction parameters of each joint are obtained; taking Joint1 as an example, its γ4(F c ) = 1.200, indicating the characteristics of this joint in terms of Coulomb friction; γ1(F s -F c ) = 5.848, reflecting the difference relationship between the static friction force and the Coulomb friction force, and these parameters provide an important basis for subsequent accurate friction modeling.

[0102] The S2 specifically includes:

[0103] (1) Construction of the static separation method: Through in-depth research on the characteristics of the robot in a high-load and high-dynamic operation environment, it can be known that the Coulomb friction coefficient F c will change significantly with the driving force and external force; in accurate robot dynamics analysis, it is necessary to accurately identify the friction model and inertial parameters respectively, and subtract the joint friction from the joint torque to determine the inertial torque; however, due to the mutual coupling of the forces / torques generated by inertial motion and joint friction, the inertial parameter model deviates seriously from the actual value, greatly affecting the identification accuracy;

[0104] Therefore, as Figure 3 shown, load-related friction static separation is carried out. A LFSS method is proposed; this method models F c as a polynomial function of the four-quadrant dynamics and external force, and through the analysis of a large amount of experimental data and theoretical derivation, the key expression is obtained:

[0105] F ci (q, τ outi ) = α i |τouti | + β i

[0106] where α i , β i are coefficients related to joint i, which reflect the linear relationship between joint friction and load torque. For the rotating joints of a serial robot, through a large number of experiments and theoretical analyses, it is found that only the load torque aligned with the joint rotation axis will affect joint friction, which mainly includes the torque generated by gravity and the load torque externally applied to the robot. When the robot is in a static equilibrium state, since the angular velocity can be approximately regarded as zero, that is at this time and τ dyni ≈ G(q), thus obtaining the static model of the robot considering load-related static friction:

[0107] τ = KI = (1 ± α)·(G(q) + τ ext ) + β

[0108] On this basis, by designing experiments, only by obtaining the data of external load changes and current changes of the robot under static conditions, the parameters of the load-related friction model can be efficiently identified using the linear regression method;

[0109] (2) Implementation of the identification process: First, according to the structure and motion characteristics of the robot, carefully design the static loading position q s ; in this process, fully consider the influence of coordinate system changes on the experimental results, through the coordinate transformation formulas:

[0110]

[0111] and

[0112]

[0113] accurately calculate the representation of the external force in the joint space; where F ext represents the generalized external force, which is composed of the force f ext and the moment m ext , and J T is the Jacobian matrix of the robot, obtained by the vector product method with respect to the base coordinate G;

[0114] Then, using the above formulas, by converting the force coordinates from T to G, obtain the exact relationship between the external force change Δ S F ext and the joint moment change Δτ ext :

[0115]

[0116] Based on this relationship, a series of static load experiments were designed by optimizing the conditions:

[0117] q=arg{q1,q2Lq n}max(‖Δτ ext ‖ 2

[0118] subjectto:{‖Δ S f ext ‖2=1,Δ S m ext =1}

[0119] And the particle swarm optimization (PSO) algorithm was used to solve it, and the optimal posture of the robot in the static loading experiment was determined to maximize the loading effect of the robot and obtain sufficiently accurate experimental data;

[0120] Then, based on the data collected in the experiment, the least squares method (LS) was used to accurately fit the parameters of the load-related friction model; through detailed analysis and complex calculations of a large amount of data, the parameter fitting results of each joint were obtained; the recognition results were evaluated by calculating the normalized root mean square error. The results show that this method can accurately identify the friction load-related coefficient, and the overall error is about 10%, effectively avoiding the coupling effect of inertia and friction parameters, providing a reliable basis for subsequent dynamic modeling.

[0121] The specific content of S3 includes:

[0122] Based on the dynamic friction modeling of HTFO (refer to Figure 4 ), the existing dynamic friction models were studied in depth. It was found that although the LuGre model can describe the dynamic characteristics of friction to a certain extent, there are still some deficiencies in robot applications; in order to better adapt to the actual operation of the robot, the LuGre model was simplified and improved; the original LuGre model is:

[0123]

[0124] This model has certain complexity and computational difficulty in describing the friction of robot joints, and its performance in key regions such as zero-speed crossing is not ideal; by introducing the Laplace operator s and the hyperbolic tangent function, it was transformed into a dynamic friction model based on the first-order HTFO of hyperbolic tangent:

[0125]

[0126] With such improvements, the model has significant advantages in describing the dynamic friction behavior of robot joints. Especially in the zero-velocity crossing range, it can more accurately reflect the continuous characteristics of actual friction, effectively reducing the complexity of dynamic parameter identification. Through in-depth analysis and derivation of the robot's motion equations, combined with the improved friction model above, a robot dynamic model is established, achieving effective decoupling of friction and inertia, and determining the key parameters to be identified, providing a theoretical basis for subsequent parameter identification.

[0127] The specific content of S4 includes:

[0128] As Figure 5 shown, a PRF Fourier series excitation trajectory is designed. Aiming at the limitations of traditional time-domain excitation methods in identifying dynamic friction, a pseudo-random finite PRF Fourier series excitation trajectory with time-frequency constraints is proposed. Based on the synthesis of periodic excitation references using traditional finite Fourier series, such as:

[0129]

[0130] Deeply considering the dynamic characteristics of robot joints, an excitation current expression is designed:

[0131]

[0132] where is used to implement the finite Fourier excitation trajectory, and its basic shape and parameters of the trajectory are determined by its calculation method; the PRBS pseudo-random binary sequence is used to effectively excite friction, and it generates a series of pulse signals through specific coding rules to stimulate different friction responses of the robot joints. By deeply analyzing the frequency response of the robot joints, using the expression of the linearized HTFO friction model in the frequency domain:

[0133]

[0134] By estimating this frequency response function, accurate identification of the dynamic friction parameter σ0 is achieved. On the HS-CO610 robot experimental platform, through careful experimental design and data analysis, it is verified that this excitation design can effectively capture the joint friction dynamics during the standard dynamic parameter excitation and identification process.

[0135] The specific content of S5 includes:

[0136] A complete experimental system was carefully built on the HS-CO610 robot experimental platform. The platform includes the robot body, an adjustable load device that can accurately load within the range of 1 - 4 kg, a high-performance controller, high-precision sensors and other key components. Through the collaborative work of these components, precise control and data acquisition of the robot under different loads and motion conditions can be achieved.

[0137] During the verification process, a variety of evaluation indicators were used to comprehensively analyze the experimental results. For the verification of the load-related friction model, the dynamic identification errors with and without the load-dependent model were mainly compared, such as Figure 8 shown, the dynamic identification errors were calculated and compared with and without the LFSS method, and the e NRMS normalized root mean square error and e PE peak error and other indicators were used. The experimental results show that the LFSS method can significantly improve the identification accuracy. For example, in the case of the minimum load VT1, the sum of the values of the LFSS algorithm decreased by 8.21% compared with the standard system without load separation, and the peak error and the sum of the values decreased by 44.21%. At the maximum load VT4, there were increases of 30.77% and 62.27% respectively. Moreover, the fitting results of the LFSS method for each axis under different loads are more consistent. Especially for joints with obvious friction-load correlation such as Joint4, the e NRMS value decreased by more than 49%, effectively reducing the error.

[0138] Referring to Figure 7 , in the verification of dynamic friction separation, the performances of the Coulomb, tanh and HTFO friction models in predicting joint friction torque were compared. Through the analysis and calculation of a large amount of experimental data, the comparison results of the predicted torque and the actual torque of each model, as well as the relative error and time cost, were obtained. The results show that the HTFO model has the best comprehensive performance under different working conditions.

[0139] In summary, through the above specific implementation manners, the present invention can effectively achieve the accurate identification of robot dynamic parameters and greatly improve the performance of the robot under complex working conditions. In practical applications, the experimental parameters and methods can be appropriately adjusted and optimized according to the characteristics and application scenarios of different robots to further improve the accuracy and reliability of robot dynamic identification.

[0140] The S5 experimental process specifically includes:

[0141] (1)Under different load conditions, the LFSS method is used to accurately obtain parameters; by carefully analyzing the force and current changes of the robot joints under different loads and combining the previously established models and algorithms, accurate parameter values are obtained; for example, under a 1 kg load, the α1 and β1 values of Joint1 are significantly different from those under a 4 kg load, and these differences reflect the influence law of the load on the friction parameters;

[0142] (2)According to the continuous friction model, by collecting joint trajectory data at different speeds, the non - linear friction is accurately calculated to obtain the parameter [γF v b]; in this process, high - precision sensors and advanced data acquisition systems are used to record the motion data of the robot joints under different speed and load combinations, and through complex calculations and analyses, the non - linear friction parameters of each joint are obtained;

[0143] (3)Based on the PRF Fourier series excitation trajectory and frequency - domain separation method, a regression matrix for identifying dynamic friction and inertial characteristics is constructed Through the processing and analysis of a large amount of experimental data and the optimization of mathematical methods and algorithms, a regression matrix that can accurately reflect the dynamic characteristics of the robot is obtained;

[0144] (4)Using the parameters and matrices obtained above, through a synchronous identification algorithm such as the least - squares method, the dynamic friction parameter σ0 and inertial parameter δ of the HTFO model are accurately identified l , thus obtaining the final robot dynamics model δ l .

[0145] An embodiment of the present invention provides a computer device, a computer - readable storage medium, and an information data processing terminal, characterized in that the computer device includes a memory and a processor, the memory stores a computer program, and when the computer program is executed by the processor, the processor executes the steps of the robot dynamics identification method considering load static separation and dynamic friction unified prediction.

[0146] An embodiment of the present invention provides a computer - readable storage medium storing a computer program, and when the computer program is executed by a processor, the processor executes the steps of the robot dynamics identification method considering load static separation and dynamic friction unified prediction.

[0147] An embodiment of the present invention provides an information data processing terminal, and this terminal integrates a system with the steps of the robot dynamics identification method considering load static separation and dynamic friction unified prediction; in this terminal, the robot system can collect various data during its operation in real - time, including joint motion data, load data, and friction force data, etc., and transmit these data to the processing module of the terminal.

[0148] The robot dynamics identification system proposed by the present invention has a wide range of application fields, can significantly improve the performance and working efficiency of related products, and provide key technical support for the intelligent development of many industries.

[0149] (1) Grinding and machining of complex metal parts

[0150] In the field of precision metal processing, for the grinding and machining of parts with complex shapes, traditional robots are difficult to balance machining accuracy and efficiency. The robot system empowered by the robot dynamics identification system of the present invention can accurately control the grinding process. First, according to the three-dimensional model of the metal part, the motion path of the robot is planned, and the path is optimized in combination with the dynamics identification results. During the machining process, the load-related friction static separation (LFSS) method is used to adjust the driving force of the robot joints in real time to adapt to different grinding forces and workpiece weight changes, ensuring stable contact between the milling cutter and the workpiece. Based on the hyperbolic tangent first order (HTFO) dynamic friction modeling, the friction influence of the robot during high-speed movement and direction switching is effectively reduced, improving the grinding accuracy.

[0151] (3) Manufacturing of parts for medical rehabilitation equipment

[0152] In the manufacturing of medical rehabilitation equipment, for some precision joint parts and transmission parts, their manufacturing accuracy directly affects the performance of the equipment and the user experience of patients. By applying the robot dynamics identification system of the present invention, during the manufacturing process, the robot can accurately position and operate. For example, when producing the key parts of knee joint rehabilitation equipment, through the design of pseudo-random finite (PRF) Fourier series excitation trajectories, the robot movement becomes more stable, reducing the influence of vibration and impact on machining accuracy. Using high-precision dynamics parameter identification, it is ensured that at different machining stages, the robot can accurately apply appropriate forces and movements, guaranteeing the dimensional accuracy and surface quality of the parts.

[0153] It should be noted that the embodiments of the present invention can be implemented by hardware, software, or a combination of software and hardware. The hardware part can be implemented using dedicated logic; the software part can be stored in a memory and executed by a suitable instruction execution system, such as a microprocessor or dedicated designed hardware. Those of ordinary skill in the art can understand that the above devices and methods can be implemented using computer-executable instructions and / or included in processor control code, for example, such code is provided on a carrier medium such as a disk, CD, or DVD-ROM, a programmable memory such as a read-only memory (firmware), or a data carrier such as an optical or electronic signal carrier. The devices and modules of the present invention can be implemented by hardware circuits of programmable hardware devices such as very large scale integrated circuits or gate arrays, semiconductors such as logic chips, transistors, etc., or field programmable gate arrays, programmable logic devices, etc., can also be implemented by software executed by various types of processors, or can be implemented by a combination of the above hardware circuits and software such as firmware.

[0154] As described above, the above are only specific embodiments of the present invention, but the protection scope of the present invention is not limited thereto. Any modification, equivalent replacement, and improvement made within the spirit and principle of the present invention by those skilled in the art within the technical scope disclosed by the present invention should be covered by the protection scope of the present invention.

Claims

1. A robot dynamics identification method that considers unified prediction of load static separation and dynamic friction, characterized in that The system includes: A non-linear friction pre-identification unit, which is used to identify non-linear friction parameters by using the Particle Swarm Optimization algorithm (PSO) based on joint friction force data; A load-related friction static separation unit, which adopts the load-related friction static separation method (LFSS), constructs a load-related friction model based on a polynomial function, and identifies parameters through a static load experiment; A dynamic friction modeling unit, which establishes a dynamic friction model based on the HTFO (High-Order Hyperbolic Tangent) friction model and combines the Laplace operator s; An excitation trajectory design unit, which adopts the pseudo-random finite Fourier series (PRF) excitation trajectory design method to construct an excitation current signal to achieve the separation of dynamic friction parameters; A model verification unit, which evaluates the accuracy and stability of the identified dynamic model based on the robot experimental platform through the Normalized Root Mean Square Error (NRMSE) and peak error; 2. The system according to claim 1, wherein The non-linear friction pre-identification unit uses a high-precision sensor to collect joint friction force data at different speeds, and combines the Particle Swarm Optimization method (PSO) to identify the parameters of the revised friction model, including static friction force, Stribeck speed, damping decay coefficient, and friction speed change law coefficient.

3. The system according to claim 1, wherein, The load-related friction static separation unit collects external force and current data through a static loading experiment, fits the parameters of the load-related friction model by using a linear regression method, calculates the representation of the external force in the joint space by using a coordinate transformation formula, and determines the optimal loading posture based on the optimization conditions.

4. The system according to claim 1, wherein The dynamic friction modeling unit is based on the LuGre friction model, and uses the Laplace operator s and the hyperbolic tangent function to improve the friction model to describe the friction change in the zero-speed crossing region, constructs a dynamic friction model based on HTFO, and realizes the decoupling of friction-inertia in the robot motion equation.

5. The system according to claim 1, wherein The excitation trajectory design unit adopts the pseudo-random finite Fourier series (PRF) excitation method, synthesizes a periodic excitation signal through Fourier series, combines the pseudo-random binary sequence (PRBS), stimulates the friction characteristics of the robot joints, realizes the estimation of joint frequency response, and identifies friction parameters based on the HTFO friction model.

6. The system according to claim 1, characterized in that, The model verification unit evaluates the accuracy of the load-related friction model through the Normalized Root Mean Square Error (NRMSE) and peak error on the robot experimental platform, and compares the torque prediction errors of the Coulomb friction, tanh friction, and HTFO friction models to verify the dynamic friction identification ability of the HTFO model.

7. The system according to claim 1, characterized in that, The model verification unit accurately identifies the dynamic friction parameters and inertia parameters of the HTFO model based on experimental data through the Least Squares method (LS) and the synchronous identification algorithm, constructs the final robot dynamic model, and evaluates the identification error of the load-related friction model under different load conditions.

8. A computer device, a computer-readable storage medium, and an information data processing terminal, characterized in that The computer device includes a memory and a processor. When the computer program stored in the memory is executed by the processor, the processor executes the steps of the robot dynamic identification method for considering load static separation and dynamic friction unified prediction as described in any one of claims 1-7.

9. A computer-readable storage medium stores a computer program. When the computer program is executed by a processor, the processor is caused to execute the steps of the robot dynamics identification method considering static load separation and unified prediction of dynamic friction according to any one of claims 1-7.

10. An information data processing terminal integrates a system for executing the steps of the robot dynamics identification method considering static load separation and unified prediction of dynamic friction according to any one of claims 1-7. In this terminal, the robot system can collect various data during its operation in real time, including joint motion data, load data, friction force data, etc., and transmit these data to the processing module of the terminal.

Images