A method and system for verifying the accuracy of human body mechanical impedance identification methods

Abstract

This invention provides a method and system for verifying the accuracy of human body mechanical impedance identification methods, belonging to the field of human body impedance measurement. The method includes: presetting a disturbance displacement / angle and calculating its velocity / angular velocity and acceleration / angular acceleration information; presetting an interaction force / torque according to the application scenario; presetting human body stiffness, damping, and inertia; establishing an impedance dynamics model and outputting preset values ​​of restoring force / torque; superimposing the preset values ​​of interaction force / torque and adding white noise to obtain simulated measured force / torque information; extracting simulated restoring force / torque from the simulated measured force / torque; using the human body mechanical impedance identification method to be verified, identifying the identification values ​​of each impedance parameter from the mapping relationship between displacement / angle and simulated restoring force / torque; and comparing the impedance identification values ​​with the corresponding preset values ​​to determine the accuracy of the impedance identification method. This invention is applicable to the verification of impedance parameter identification methods for time-varying or time-invariant, limb extremities, or joints under different interaction scenarios.

Description

Technical Field

[0001] This invention belongs to the field of human body mechanical impedance measurement, and more specifically, relates to a method and system for verifying the accuracy of human body mechanical impedance identification methods. Background Technology

[0002] With the development of sports science, bionics, and robotics, learning and understanding human's remarkable motor abilities and applying them to robots to achieve human-like human-computer interaction, human-computer collaboration, human-computer teleoperation, or to propose design methods for actuators and controllers with human-like compliant characteristics, is currently one of the research hotspots in the field of robotics. Among these, the human body's mechanical impedance characteristics, as one of the important characteristics of human's remarkable motor abilities, enable the human body to effectively adapt to complex environments, complete dynamic motion tasks, and maintain compliance and stability in the process of human-environment interaction.

[0003] Currently, human body mechanical impedance cannot be directly measured using specialized sensors like motion or mechanical signals. Instead, it requires collecting information such as displacement and restoring force through perturbation experiments, and then using parameter identification techniques to identify the human body mechanical impedance value from the experimental data, thus achieving indirect impedance measurement. From a purely mathematical perspective, regardless of the parameter identification method used, given the input and output information of the system to be identified, the method will always produce an identification result. However, the accuracy of the identification result needs to be judged by comparing it with the true value. For human body impedance parameters, since there are currently no sensors that can directly measure the true impedance value, a convenient and effective method is needed to verify the accuracy of the human body impedance identification method and its results while performing human body impedance parameter identification.

[0004] Although parameter identification technology is relatively advanced, and many methods exist for identifying human mechanical impedance, classic methods such as the least squares method can be directly used for identifying time-invariant parameters. However, for complex application scenarios, such as the identification of time-varying parameters, it is still necessary to develop appropriate methods for parameter identification. Specifically, when there is a time-varying interaction between the human body and the environment, human mechanical impedance exhibits time-varying characteristics; when the human posture changes, the human mechanical impedance also differs from that in a fixed posture; furthermore, the degree of muscle activation also affects the level of human mechanical impedance. For such complex scenarios, suitable parameter identification methods need to be proposed to achieve effective and accurate identification of complex impedances. Methods such as the short data segment method, the lumped data method, the linear variable parameter method, and the basis function expansion method have all been used for identifying time-varying parameters. However, the applicable conditions of these methods may differ, and the requirements for data acquisition experiments may also differ, making it difficult to judge the merits of different identification methods. Therefore, a universal method is also needed to verify the feasibility of different impedance identification methods and determine their merits. Summary of the Invention

[0005] In view of the unique nature of human mechanical impedance, which cannot be directly measured by sensors, and the lack of accuracy verification methods for existing impedance measurements based on parameter identification technology, the purpose of this invention is to provide a method and system for verifying the accuracy of human mechanical impedance identification methods. This aims to solve the problems of difficulty in measuring human mechanical impedance and unclear accuracy of impedance identification, especially the unclear accuracy of impedance identification methods for complex application scenarios.

[0006] To achieve the above objectives, in one aspect, the present invention provides a method for verifying the accuracy of human body mechanical impedance identification methods, comprising the following steps:

[0007] Step 1: Set the disturbance displacement / angle information, and calculate the first and second derivatives of the disturbance displacement / angle information in sequence to obtain the velocity / angular velocity information and acceleration / angular acceleration information during the disturbance process.

[0008] Step 2: Set the interactive force / torque level to be time-varying or time-invariant based on the characteristics of the interactive task;

[0009] Step 3: Preset the stiffness based on the square of the interaction force / torque and its linear mapping relationship with stiffness, preset the damping based on the linear mapping relationship between the interaction force / torque and damping, and randomly preset the inertia; among them, the linear parameters reflecting the linear mapping relationship are all randomly set.

[0010] Step 4: Establish an impedance dynamics model and obtain preset values ​​for restoring force / torque. Specifically, input displacement / angle into the stiffness dynamics model, velocity / angular velocity into the damping dynamics model, and acceleration / angular acceleration into the inertial dynamics model. Combine these with preset stiffness, damping, and inertia to obtain the stiffness component, damping component, and inertial component of the restoring force / torque. Sum these three components to obtain the preset value for the restoring force / torque.

[0011] Step 5: Superimpose the preset restoring force / torque with the set interaction force / torque to obtain a noise-free measured force / torque; add random white noise to the noise-free measured force / torque output signal to obtain simulated measured force / torque information;

[0012] Step Six: Based on the different characteristics of interactive forces / torques in different interactive tasks, extract simulated restoring forces / torques from simulated measured forces / torques using analytical or numerical methods, simulating the process of extracting restoring forces / torques from force / torque data measured by sensors; using perturbation displacement / angle, velocity / angular velocity, and acceleration / angular velocity as input signals, and simulated restoring forces / torques as output signals, use the human mechanical impedance identification method to be verified to identify the stiffness, damping, and inertia identification values ​​of human limb extremities / joints;

[0013] Step 7: Compare the identified values ​​of human body stiffness, damping, and inertia with the preset stiffness, damping, and inertia respectively, and verify the accuracy of the human body mechanical impedance identification method through the variance contribution rate and root mean square error evaluation index.

[0014] More preferably, in step two, the interaction force / torque f is set according to the interaction task type. v The magnitude of (t); in step three, the inertia I in the mechanical impedance is randomly preset, and the stiffness K(t) and damping B(t) in the mechanical impedance are preset;

[0015] Each component of stiffness has a linear relationship with the square of the interaction force / moment amplitude:

[0016]

[0017] Each component of the damping has a linear relationship with the absolute value of the interaction force / torque amplitude:

[0018]

[0019] Wherein, stiffness K(t) is a matrix containing i*j elements; K ij Let α be any element in the preset stiffness matrix; α0…α6 are a series of linear parameters that map the squares of the interaction force / moment amplitudes to the elements of the stiffness matrix; the damping B(t) is a matrix containing i*j elements, B ij (t) represents any element in the preset damping matrix; i = 1, 2, 3, j = 1, 2, 3; β0…β6 are a series of linear parameters that map the absolute values ​​of the interaction force / torque amplitudes to the elements of the damping matrix; f is the setpoint for the interaction force / torque. v A component of (t) has subscripts x, y and z that represent the interaction force / torque components along the X, Y and Z axes, respectively; + and - represent the two opposite directions of the interaction force / torque.

[0020] More preferably, in step four, an impedance dynamics model is established, with the preset value of the restoring force / torque being f. r (t)=f I (t)+f B (t)+f K (t);

[0021] Among them, the inertial dynamics model Describe the inertia component f in the restoring force / torque I (t) with inertia I and acceleration / angular acceleration The dynamic relationship between them;

[0022] Damped dynamic model Describe the damping component f in the restoring force / torque B (t) and time-varying damping value B(t) and velocity / angular velocity The dynamic relationship between them;

[0023] Stiffness dynamics model f K (t)=K(t)q(t) describes the stiffness component f in the restoring force / torque. K The dynamic relationship between (t) and time-varying stiffness K(t) and displacement / angle q(t);

[0024] In step five, set the white noise f. n (t), through the preset value f of the interaction force / torque v (t), preset value of restoring force / torque f r (t) and white noise setpoint f n By superimposing (t), the simulated measured force / torque f is obtained. o2 (t)=f v (t)+f r (t)+f n (t);

[0025] Where q(t) is the three-dimensional random perturbation displacement / angle set in step one, its first and second derivatives are calculated to obtain the perturbation velocity / angular velocity, respectively. With acceleration / angular acceleration

[0026] More preferably, the method for extracting the simulated restoring force / torque from the simulated measured force / torque in step six is ​​as follows: depending on the type of interaction force / torque, an analytical or numerical method is used to extract the simulated measured force / torque from the simulated measured force / torque f. o2 Extract the simulated restoring force / torque f′ used to identify impedance parameters from (t). r (t); where, when the interaction force / torque f v When (t) is 0, the simulated restoring force / torque f′ r (t) equals the simulated measured force / torque f o2 (t); when the interaction force / torque f v When (t) is a constant value that is not equal to 0, the simulated measured force / torque f o2 (t) Perform a mean-removal operation to obtain the simulated restoring force / torque f′ r (t); when the interaction force / torque f v (t) When it has time-varying characteristics, the simulated measured force / torque f is analyzed using numerical methods. o2 The restoring force / torque and the interaction force / torque in (t) are decoupled to obtain the simulated restoring force / torque f′. r (t).

[0027] More preferably, the disturbance displacement / angle q(t) and its velocity / angular velocity are... Acceleration / Angular Acceleration As input to the human body impedance identification model, the simulated restoring force / torque f′ is used. r (t) is the output of the human body impedance identification model. Using the human body impedance identification model and method to be verified, the inertia, damping, and stiffness of the human limb extremities / joints are identified to obtain the inertia identification value I′, damping identification value B′(t), and stiffness identification value K′(t); the following relationships are satisfied:

[0028]

[0029] The accuracy of the human body impedance identification method is verified by comparing the identified impedance parameter values ​​I′, B′(t), and K′(t) with the preset impedance parameter values ​​I, B(t), and K(t) using the variance contribution rate and root mean square error evaluation index.

[0030] On the other hand, the present invention provides a system for verifying the accuracy of human body mechanical impedance identification methods, comprising:

[0031] The simulation input module is used to set the disturbance displacement / angle information, and sequentially calculate the first and second derivatives of the disturbance displacement / angle information to obtain the velocity / angular velocity information and acceleration / angular acceleration information during the disturbance duration; and to set the time-varying or time-invariant interactive force / torque level according to the characteristics of the interactive task.

[0032] The impedance parameter preset module is used to preset the stiffness based on the square of the interaction force / torque and its linear mapping relationship with stiffness, preset the damping based on the linear mapping relationship between the interaction force / torque and damping, and randomly preset the inertia; among them, the linear parameters reflecting the linear mapping relationship are all randomly set.

[0033] The restoring force / torque preset module is used to establish an impedance dynamics model and obtain preset values ​​for the restoring force / torque. Specifically, displacement / angle information is input into the stiffness dynamics model, velocity / angular velocity information is input into the damping dynamics model, and acceleration / angular acceleration information is input into the inertial dynamics model. Combined with preset stiffness, damping, and inertia, the stiffness component, damping component, and inertial component of the restoring force / torque are obtained. The three components are then summed to obtain the preset value of the restoring force / torque.

[0034] The output module is used to superimpose the preset restoring force / torque with the set interaction force / torque to obtain a noise-free measured force / torque; and to add random white noise to the noise-free measured force / torque output signal to obtain simulated measured force / torque information.

[0035] The human impedance identification module is used to extract simulated restoring force / torque from simulated measured force / torque using analytical or numerical methods, based on the different characteristics of interactive forces / torques in different interactive tasks. It simulates the process of extracting restoring force / torque from force / torque data measured by sensors. It uses perturbation displacement / angle, velocity / angular velocity, and acceleration / angular acceleration as input signals and simulated restoring force / torque as output signals. It uses the human impedance identification method to be verified to identify the stiffness, damping, and inertia of human limb extremities / joints.

[0036] The impedance parameter identification accuracy verification module is used to compare the identified values ​​of human body stiffness, damping, and inertia with preset stiffness, damping, and inertia, respectively, and to verify the accuracy of the human body mechanical impedance identification method through variance contribution rate and root mean square error evaluation indicators.

[0037] More preferably, the impedance parameter preset module sets the interaction force / torque f according to the interaction task type. v The magnitude of (t); the inertia I in the randomly preset mechanical impedance, and the stiffness K(t) and damping B(t) in the preset mechanical impedance;

[0038] Each component of stiffness has a linear relationship with the square of the interaction force / moment amplitude:

[0039]

[0040] Each component of the damping has a linear relationship with the absolute value of the interaction force / torque amplitude:

[0041]

[0042] Wherein, stiffness K(t) is a matrix containing i*j elements; K ij Let α be any element in the preset stiffness matrix; α0…α6 are a series of linear parameters that map the squares of the interaction force / moment amplitudes to the elements of the stiffness matrix; the damping B(t) is a matrix containing i*j elements, B ij (t) represents any element in the preset damping matrix; i = 1, 2, 3, j = 1, 2, 3; β0…β6 are a series of linear parameters that map the absolute values ​​of the interaction force / torque amplitudes to the elements of the damping matrix; f is the setpoint for the interaction force / torque. v A component of (t) has subscripts x, y and z that represent the interaction force / torque components along the X, Y and Z axes, respectively; + and - represent the two opposite directions of the interaction force / torque.

[0043] More preferably, an impedance dynamics model is established in the restoring force / torque preset module, and the preset value of the restoring force / torque is f. r (t)=f I (t)+fB (t)+f K (t);

[0044] Among them, the inertial dynamics model Describe the inertia component f in the restoring force / torque I (t) with inertia I and acceleration / angular acceleration The dynamic relationship between them;

[0045] Damped dynamic model Describe the damping component f in the restoring force / torque B (t) and time-varying damping value B(t) and velocity / angular velocity The dynamic relationship between them;

[0046] Stiffness dynamics model f K (t)=K(t)q(t) describes the stiffness component f in the restoring force / torque. K The dynamic relationship between (t) and time-varying stiffness K(t) and displacement / angle q(t);

[0047] Where q(t) is the set random perturbation displacement / angle, its first and second derivatives are obtained to get the perturbation velocity / angular velocity, respectively. With acceleration / angular acceleration

[0048] More preferably, the method by which the human body impedance identification module extracts the simulated restoring force / torque from the simulated measured force / torque is as follows:

[0049] Output module setting white noise f n (t), through the preset value f of the interaction force / torque v (t), preset value of restoring force / torque f r (t) and white noise setpoint f n By superimposing (t), the simulated measured force / torque f is obtained. o2 (t)=f v (t)+f r (t)+f n (t);

[0050] Depending on the type of interaction force, analytical or numerical methods are used to obtain the simulated and measured force / torque f. o2 Extract the simulated restoring force / torque f′ used to identify impedance parameters from (t). r (t); where, when the interaction force / torque f v When (t) is 0, the simulated restoring force / torque f r (t) equals the simulated measured force / torque f o2 (t); when the interaction force / torque f vWhen (t) is a constant value that is not equal to 0, the simulated measured force / torque f o2 (t) Perform a mean-removal operation to obtain the simulated restoring force / torque f′ r (t); when the interaction force / torque f v (t) When it has time-varying characteristics, the simulated measured force / torque f is analyzed using numerical methods. o2 The restoring force / torque and the interaction force / torque in (t) are decoupled to obtain the simulated restoring force / torque f′. r (t).

[0051] More preferably, the human body impedance identification module will include the disturbance displacement / angle q(t) and its velocity / angular velocity. Acceleration / Angular Acceleration As input, simulate the restoring force / torque f′ r (t) is used as the output. The human body impedance identification method to be verified is used to identify the inertia, damping, and stiffness of the human limb extremities / joints, and to obtain the inertia identification value I′, the damping identification value B′(t), and the stiffness identification value K′(t); the following relationship is satisfied:

[0052]

[0053] The accuracy of the human body impedance identification method is verified by comparing the identified impedance parameter values ​​I′, B′(t), and K′(t) with the preset impedance parameter values ​​I, B(t), and K(t) using the variance contribution rate and root mean square error evaluation index.

[0054] In summary, the technical solutions conceived by this invention have the following beneficial effects compared with the prior art:

[0055] Human mechanical impedance cannot be directly measured using physical sensors; instead, it must be measured indirectly through parameter identification experiments and methods. Due to the lack of true measurement values, it is impossible to determine the accuracy of the human impedance identification results. For these reasons, this invention provides a method and system for verifying the accuracy of human mechanical impedance identification methods. Based on the time-invariant or time-varying characteristics of interaction forces / torques, muscle activation, and limb posture during human-environment interaction, the accuracy of time-invariant or time-varying impedance parameter identification methods can be verified accordingly. For tasks involving interaction at limb extremities or joints, the accuracy of human limb extremity impedance or joint impedance identification can be verified. For different impedance parameter identification methods, a comparative analysis of identification accuracy can be performed to determine their superiority or inferiority. Accurate human impedance parameter identification results can be used to analyze the dynamic interaction characteristics between the human body and the external environment, and can also be used to guide the anthropomorphic compliant design or anthropomorphic compliant interactive control of robots. Attached Figure Description

[0056] Figure 1 This is a system principle block diagram provided by an embodiment of the present invention for verifying the accuracy of the human body mechanical impedance identification method;

[0057] Figure 2 The impedance identification method provided in this embodiment of the invention verifies the preset values ​​of each torque in the system;

[0058] Figure 3 This is a comparison chart of the time-varying impedance identification value of the human ankle joint and the preset value provided in the embodiment of the present invention. Detailed Implementation

[0059] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be described in detail herein with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are for illustrative purposes only and are not intended to limit the invention.

[0060] On one hand, the present invention provides a system for verifying the accuracy of human body mechanical impedance identification methods, comprising:

[0061] The simulation input module provides input information for the verification system. This input includes displacement, interaction force / torque, and white noise. First, three-dimensional random perturbation displacement information is set, requiring its power spectrum to remain relatively flat within a certain range and the perturbation amplitude to be within a reasonable range. The first and second derivatives of this random perturbation displacement information are then calculated sequentially to obtain the velocity / angular velocity and acceleration / angular acceleration information during the perturbation process. Second, the magnitude of the interaction force / torque is set according to the specific interaction task. Finally, the white noise level is set while ensuring a reasonable signal-to-noise ratio.

[0062] The impedance parameter preset module is used to preset the mechanical impedance parameters of the verification system. The mechanical impedance parameters include three parameters: stiffness, damping, and inertia. Depending on whether the problem is planar or spatial, these parameters can be two-dimensional or three-dimensional, and each impedance parameter includes 2*2 or 3*3 components. The inertia value can be randomly preset, while the stiffness and damping are preset according to the following principles: each component of stiffness is linearly related to the square of the interaction force / torque amplitude, and each component of damping is linearly related to the absolute value of the interaction force / torque amplitude. Based on the interaction force / torque given in the simulation input module, and given linear parameters describing the linear mapping relationship, the preset values ​​of stiffness and damping can be obtained through the above linear relationships. The linear parameters can be randomly given, but the obtained stiffness and damping parameters must be within a reasonable range of human body impedance.

[0063] The restoring force / torque preset module establishes an impedance dynamics model, namely a stiffness dynamics model, a damping dynamics model, and an inertial dynamics model. These three models take displacement, velocity / angular velocity, and acceleration / angular acceleration from the input module as inputs, respectively, and combine them with the stiffness, damping, and inertia preset values ​​from the impedance parameter preset module to calculate the stiffness component, damping component, and inertial component of the restoring force / torque, respectively. The sum of these three force / torque components is the preset restoring force / torque.

[0064] The output module generates model output information for parameter identification; the preset value of the restoring force / torque output by the restoring force / torque preset module is added to the interactive force / torque set in the simulation input module to obtain a noise-free measured force / torque; this measured force / torque is then added to the random white noise in the input module to form the simulated measured force / torque for parameter identification;

[0065] The human body impedance parameter identification module is used to identify impedance parameters from the input and output information of the human body impedance system. To simulate the process of identifying human body impedance parameters using the identification method to be verified in actual applications, it is necessary to extract the simulated restoring force / torque values ​​for identification from the simulated measured force / torque. Depending on the type of interaction force / torque, there are three methods to achieve this goal: when the interaction force / torque is 0, the simulated restoring force / torque is equal to the preset restoring force / torque; when the interaction force / torque is a constant value that is not equal to 0, the simulated restoring force can be obtained by performing a mean-reduction operation on the preset restoring force / torque; when the interaction force / torque has time-varying characteristics, the simulated restoring force / torque and the interaction force / torque in the simulated measured force / torque are decoupled using numerical methods to obtain the simulated restoring force / torque. Using the displacement in the simulation input module as input and the simulated restoring force / torque extracted from the simulated measured force / torque as output, the human body impedance parameter identification method to be verified is used to identify the identified values ​​of the human body impedance parameters.

[0066] The impedance parameter identification accuracy verification module is used to verify the accuracy of the human body impedance identification method. It compares the impedance parameter identification value obtained by the human body impedance parameter identification module with the impedance parameter preset value set in the impedance parameter preset module. The accuracy of the human body impedance identification method is verified by quantitatively calculating the variance contribution rate (VAF) and root mean square error (RMSE) evaluation indicators.

[0067] More specifically, if the stiffness of the human limb extremity is identified, and a random three-dimensional perturbation displacement q(t) of the limb extremity is set as a function of time t, its first and second derivatives are calculated to obtain the velocity / angular velocity, respectively. and acceleration / angular acceleration The three-dimensional perturbation displacement and its velocity / angular velocity and acceleration / angular acceleration are all 3*1 matrices.

[0068] Set the interaction force / torque level f v (t), depending on the actual task, the interaction force / torque can be 0 or non-zero, can change with time or remain constant, and can also have directional information to simulate the application of interaction force / torque in different directions; when studying three-dimensional impedance, this interaction force / torque is a 3*1 matrix.

[0069] We assume that each component of the stiffness matrix K(t) has the following linear relationship with the square of the magnitude of the interaction force / moment:

[0070]

[0071] Wherein, the preset stiffness K(t) is a matrix containing i*j elements; K ij α0…α6 are any element in the preset stiffness matrix; α0…α6 are a series of linear parameters that map the squares of the interaction force / moment amplitudes to the elements of the stiffness matrix, and can be set randomly; f is the setpoint for the interaction force / moment. v A component of (t) has subscripts x, y, and z representing the interaction force / torque components along the X, Y, and Z axes, respectively; + and - represent the two opposite directions of the interaction force / torque, such as f. x+ Represents the interaction force / torque f v (t) Force / torque components in the positive X-axis direction;

[0072] Similarly, each component of the damping matrix B(t) is assumed to have the following linear relationship with the absolute value of the interaction force / torque amplitude:

[0073]

[0074] Wherein, the preset damping B(t) is a matrix containing i*j elements, B ij (t) represents any element in the preset damping matrix; i = 1, 2, 3, j = 1, 2, 3; β0…β6 are a series of linear parameters that map the absolute values ​​of the interaction force / torque amplitude to each element of the damping matrix, and can be set randomly; f and its subscripts are defined as above;

[0075] Set the inertia matrix I, which is generally a fixed value, but can be set randomly.

[0076] It should be noted that although the inertia matrix I and the linear parameters α0…α6, β0…β6 can be randomly set, the overall stiffness matrix K(t), damping matrix B(t), and all elements of the inertia matrix I should be within a reasonable range of human body impedance. When studying three-dimensional impedance, the inertia matrix, stiffness matrix, and damping matrix are all 3*3 matrices.

[0077] An impedance dynamics module is established using a second-order impedance model, and a preset value f for the restoring force / torque is established.r (t) and displacement q(t), velocity / angular velocity Acceleration / Angular Acceleration The relationship between them is shown below:

[0078]

[0079]

[0080] f K (t)=K(t)q(t)

[0081] f r (t)=f I (t)+f B (t)+f K (t)

[0082] This is an inertial dynamics module used to describe the restoring force / torque inertial components f. I (t) and acceleration / angular acceleration The dynamic relationship between them, where I is the preset inertia matrix, and it is generally assumed that the inertia does not change with time; This is a damping dynamics module used to describe the restoring force / moment damping component f. B (t) and velocity / angular velocity The dynamic relationship between them, where B(t) is a preset damping matrix that may vary with time; f K (t)=K(t)q(t) is the stiffness dynamics module, used to describe the restoring force / moment stiffness components f. K The dynamic relationship between K(t) and displacement q(t) is given by K(t), which is a preset stiffness matrix that may vary with time; the preset value of restoring force / moment is f. r (t) is the sum of the restoring force / moment inertial components, damping components, and stiffness components; when studying three-dimensional impedance, this restoring force / moment is a 3*1 matrix;

[0083] By using the preset value f of interactive force / torque v (t) and preset value of restoring force / torque f r The sum of (t) yields the noise-free preset value f for the measured force / torque. o1 (t); further compared with the preset white noise f n (t) Superimpose the measured force / torque simulation values ​​f for parameter identification. o2 (t); when studying three-dimensional impedance, f o1 (t) and f o2 (t) are all 3*1 matrices, as shown below:

[0084] f o1(t)=f r (t)+f v (t)

[0085] f o2 (t)=f o1 (t)+f n (t)

[0086] This completes the establishment of the simulation model of the entire human body mechanical impedance system, that is, simulating the external behavior of the limb end under the action of any disturbance q(t), due to its stiffness, damping and inertial characteristics. o2 (t) represents the force / torque;

[0087] Based on this simulation system, in order to perform subsequent impedance parameter identification, it is necessary to obtain the simulated values ​​of measured force / torque f. o2 Extract the simulated restoring force / torque value f′ from (t) to identify the impedance parameter. r (t); Based on the different types of interaction forces / torques between the limbs and the environment during the actual task, there are three methods to achieve this goal: when the interaction force / torque f v When (t) is 0, the restoring force / torque f′ r (t) equals the measured force / torque f o2 (t); when the interaction force / torque f v When (t) is a constant value that is not equal to 0, for the measured force / torque f o2 (t) Performing a mean-removal operation yields the restoring force / torque f′. r (t); when the interaction force / torque f v (t) When it has time-varying characteristics, numerical methods are used to analyze the measured force / torque f. o2 The restoring force / torque and the interaction force / torque in (t) are decoupled to obtain the restoring force / torque f′. r (t);

[0088] The displacement q(t) of the limb end and its velocity / angular velocity Acceleration / Angular Acceleration As input to the human body impedance system, the simulated value f′ of the restoring force / torque extracted from the measured force / torque is used. r (t) is taken as the output of the human body impedance system. Using the impedance parameter identification method to be verified, the inertia, damping, and stiffness of the human limb end are identified to obtain the inertia identification value I′, the damping identification value B′(t), and the stiffness identification value K′(t), which satisfy the following relationship:

[0089]

[0090] The accuracy of the human body impedance identification method was verified by comparing the identified impedance parameter values ​​I′, B′(t), and K′(t) with the preset impedance parameter values ​​I, B(t), and K(t) using the variance contribution rate (VAF) and root mean square error (RMSE) evaluation indicators.

[0091] On the other hand, as attached Figure 1 As shown, this invention provides a method for verifying the accuracy of human body mechanical impedance identification methods, specifically including the following steps:

[0092] Step 1: Set the disturbance displacement / angle information, and calculate the first and second derivatives of the disturbance displacement / angle information in sequence to obtain the velocity / angular velocity information and acceleration / angular acceleration information during the disturbance process.

[0093] Step 2: Set the interactive force / torque level to be time-varying or time-invariant based on the characteristics of the interactive task;

[0094] Step 3: Preset the stiffness based on the square of the interaction force / torque and its linear mapping relationship with stiffness, preset the damping based on the linear mapping relationship between the interaction force / torque and damping, and randomly preset the inertia; among them, the linear parameters reflecting the linear mapping relationship are all randomly set.

[0095] Step 4: Establish an impedance dynamics model and obtain preset values ​​for restoring force / torque. Specifically, input displacement / angle into the stiffness dynamics model, velocity / angular velocity into the damping dynamics model, and acceleration / angular acceleration into the inertial dynamics model. Combine these with preset stiffness, damping, and inertia to obtain the stiffness component, damping component, and inertial component of the restoring force / torque. Sum these three components to obtain the preset value for the restoring force / torque.

[0096] Step 5: Superimpose the preset restoring force / torque with the set interaction force / torque to obtain a noise-free measured force / torque; add random white noise to the noise-free measured force / torque output signal to obtain simulated measured force / torque information;

[0097] Step Six: Based on the different characteristics of interactive forces in different interactive tasks, extract simulated restoring force / torque from the simulated measured force / torque using analytical or numerical methods, simulating the process of extracting restoring force / torque from the force / torque data measured by sensors; using perturbation displacement / angle, velocity / angular velocity, and acceleration / angular acceleration as input signals, and simulated restoring force / torque as output signals, use the human mechanical impedance identification method to be verified to identify the stiffness, damping, and inertia identification values ​​of the human limb ends / joints;

[0098] Step 7: Compare the identified values ​​of human body stiffness, damping, and inertia with the preset stiffness, damping, and inertia respectively, and verify the accuracy of the human body mechanical impedance identification method through the variance contribution rate and root mean square error evaluation index.

[0099] Example

[0100] This example demonstrates the accuracy verification of a time-varying impedance parameter identification method for the ankle joint during continuous changes in ankle joint posture while in a seated position. The specific steps include:

[0101] S1: Define the random perturbation angle θ(t) in the sagittal plane of the ankle joint that varies with time t; calculate the first and second derivatives of the perturbation angle sequentially, i.e., obtain the angular velocity during the perturbation process. and angular acceleration

[0102] S2: During the continuous change of ankle joint position, the ankle joint interaction torque is not zero. Set the ankle joint interaction torque M. v (t) = 50sin(0.1πt) N·m, as Figure 2 As shown in the first row;

[0103] S3: Set the preset values ​​of impedance parameters in the sagittal plane of the ankle joint, including the preset value of time-varying stiffness K(t), the preset value of time-varying damping B(t), and the preset value of time-invariant inertia I.

[0104] First, based on the linear relationship between joint stiffness and the square of the interaction moment amplitude, a preset stiffness value is set.

[0105]

[0106] Where K(t) is the preset value of the time-varying stiffness parameter; M v (t) is the set value of the interaction torque, and the positive and negative signs of the subscripts represent the two opposite directions of the interaction torque; α0, α1 and α2 are a series of linear parameters that map the square of the interaction torque amplitude to each element of the stiffness matrix. These parameters can be given randomly, but the preset value of the time-varying stiffness parameter K(t) should be within the normal range of the stiffness of the human ankle joint.

[0107] Secondly, based on the linear relationship between the absolute values ​​of joint damping and the interaction torque amplitude, the preset damping value is set.

[0108]

[0109] Where B(t) is the preset value of the time-varying damping parameter; M v (t) represents the set interaction torque; the positive or negative sign of the subscript indicates the positive or negative direction of the interaction torque, for example... Represents the interaction torque Mv (t) is the torque component in the dorsiflexion direction of the ankle joint; β0, β1 and β2 are a series of linear parameters that map the absolute values ​​of the interaction torque amplitudes to the elements of the damping matrix. These parameters can be given randomly, but the preset value of the time-varying damping parameter B(t) should be within the normal range of human ankle joint damping.

[0110] Finally, set the time-invariant inertia preset value I. This parameter can be given randomly, but it must still be ensured that it is within the normal range of human ankle joint inertia.

[0111] S4: Obtain the preset value of ankle joint recovery torque;

[0112] Based on the above preset values ​​for stiffness, damping, and inertia, the preset value M of the ankle joint restoring torque is generated through a dynamic model. r (t):

[0113]

[0114]

[0115] M K (t)=K(t)θ(t)

[0116] M r (t)=M I (t)+M B (t)+M K (t)

[0117] Among them, M I (t), M B (t), M K (t) represent the torque components related to inertia, damping, and stiffness in the ankle joint restoring torque; I, B(t), and K(t) are the preset values ​​of the time-varying inertia parameter, time-varying damping parameter, and time-varying stiffness parameter, respectively; the preset value of the ankle joint restoring torque M is... r (t) is M I (t), M B (t), M K The sum of (t), as Figure 2 As shown in the second row;

[0118] S5: Obtain the preset value of ankle joint measurement torque without noise;

[0119] Noise-free ankle joint measurement torque preset value M o1 (t) is determined by the preset value of the restoring torque M r (t) and the preset value of the interaction torque M v (t) consists of two parts:

[0120] M o1 (t)=M r(t)+M v (t)

[0121] S6: Obtain the preset value of the measured torque of the ankle joint;

[0122] The preset value M for measuring torque at the ankle joint in the absence of noise. o1 Add white noise signal M to (t) n (t) Obtain the preset value M of the measured torque of the ankle joint. o2 (t):

[0123] M o2 (t)=M o1 (t)+M n (t)

[0124] Among them, the white noise signal M n (t) is used to represent various interference signals during the analog signal measurement process; preset value M of the measured torque of the ankle joint. o2 (t) Measured values ​​of the forces between the extremities and the environment during a simulated actual experiment, such as... Figure 2 As shown in the third row.

[0125] S7: Verification of the accuracy of the method for identifying time-varying impedance parameters of the human ankle joint;

[0126] Because the interaction forces in this task are time-varying, numerical analysis is used to obtain the measured torque from the preset value M. o2 Extract the restoring torque M′ used for impedance identification from (t). r (t); Set the preset joint angle value θ(t) and the restoring torque M′ r (t) are used as the input and output for identifying the time-varying impedance parameters of the human ankle joint, respectively. The impedance parameter identification method to be verified is used to identify the ankle joint inertia, damping, and stiffness. The identified impedance parameter values ​​are compared with the preset impedance parameter values ​​in S3 to verify the accuracy of the proposed impedance parameter identification method. The results are as follows: Figure 3 As shown; further, the preset value of the impedance parameter is set to y. i Impedance parameter identification value set to The accuracy of the human ankle joint time-varying impedance parameter identification method is quantified by evaluating the variance contribution rate (VAF) and root mean square error (RMSE). The calculation formula is as follows:

[0127]

[0128]

[0129] Among them, y i , These represent the preset and identified impedance parameters, respectively; n is the number of data samples in the entire identification process; var represents the variance calculation.

[0130] Table 1 shows the identification statistics of VAF and RMSE for each impedance parameter. Since the preset and identified values ​​of inertia are constants, only the RMSE index of the inertia parameter is shown in the table. It can be seen that the identification method of time-varying impedance parameters of the human ankle joint in this embodiment has good identification accuracy; furthermore, this verification system is suitable for verifying the accuracy of the identification method of time-varying impedance parameters of the human ankle joint.

[0131] Table 1

[0132]

[0133] The unit for identifying stiffness RMSE is Nm / rad, and the unit for identifying damping RMSE is Nm·s / rad.

[0134] In summary, compared with the prior art, the present invention has the following advantages:

[0135] Human mechanical impedance cannot be directly measured using physical sensors; instead, it must be measured indirectly through parameter identification experiments and methods. Due to the lack of true measurement values, it is impossible to determine the accuracy of the human impedance identification results. For these reasons, this invention provides a method and system for verifying the accuracy of human mechanical impedance identification methods. Specifically, based on the time-invariant or time-varying characteristics of interaction forces / torques, muscle activation, and limb posture during human-environment interaction, the accuracy of time-invariant or time-varying impedance parameter identification methods can be verified accordingly. For tasks involving interaction at limb extremities or joints, the accuracy of human limb extremity impedance or joint impedance identification can be verified. For different impedance parameter identification methods, a comparative analysis of identification accuracy can be conducted to determine their superiority or inferiority. Accurate human impedance parameter identification results can be used to analyze the dynamic interaction characteristics between the human body and the external environment, and can also be used to guide the anthropomorphic compliant design or anthropomorphic compliant interactive control of robots. Verification shows that this invention effectively achieves the above objectives.

[0136] Those skilled in the art will readily understand that the above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for verifying the accuracy of human body mechanical impedance identification methods, characterized in that, Includes the following steps: Step 1: Set the disturbance displacement / angle information, and calculate the first and second derivatives of the disturbance displacement / angle information in sequence to obtain the velocity / angular velocity information and acceleration / angular acceleration information during the disturbance process. Step 2: Set the interactive force / torque level to be time-varying or time-invariant based on the characteristics of the interactive task; Step 3: Preset the stiffness based on the square of the interaction force / torque and its linear mapping relationship with stiffness, preset the damping based on the linear mapping relationship between the interaction force / torque and damping, and randomly preset the inertia; among them, the linear parameters reflecting the linear mapping relationship are all randomly set. Step 4: Establish an impedance dynamics model and obtain preset values ​​for restoring force / torque. Specifically, input displacement / angle into the stiffness dynamics model, velocity / angular velocity into the damping dynamics model, and acceleration / angular acceleration into the inertial dynamics model. Combine these with preset stiffness, damping, and inertia to obtain the stiffness component, damping component, and inertial component of the restoring force / torque. Sum these three components to obtain the preset value for the restoring force / torque. Step 5: Superimpose the preset restoring force / torque with the set interaction force / torque to obtain a noise-free measured force / torque; add random white noise to the noise-free measured force / torque output signal to obtain simulated measured force / torque information; Step Six: Based on the different characteristics of interactive forces / torques in different interactive tasks, extract simulated restoring forces / torques from simulated measured forces / torques using analytical or numerical methods, simulating the process of extracting restoring forces / torques from force / torque data measured by sensors; using perturbation displacement / angle, velocity / angular velocity, and acceleration / angular acceleration as input signals, and simulated restoring forces / torques as output signals, use the human mechanical impedance identification method to be verified to identify the stiffness, damping, and inertia identification values ​​of human limb extremities / joints; Step 7: Compare the identified values ​​of human body stiffness, damping, and inertia with the preset stiffness, damping, and inertia respectively, and verify the accuracy of the human body mechanical impedance identification method through the variance contribution rate and root mean square error evaluation index.

2. The method for verifying the accuracy of human body mechanical impedance identification according to claim 1, characterized in that, In step two, the interaction force / torque f is set according to the type of interaction task. v The magnitude of (t); In step three, the inertia I in the mechanical impedance is randomly preset, and the stiffness K(t) and damping B(t) in the mechanical impedance are preset; Each component of stiffness has a linear relationship with the square of the interaction force / moment amplitude: Each component of the damping has a linear relationship with the absolute value of the interaction force / torque amplitude: Wherein, stiffness K(t) is a matrix containing i*j elements; K ij Let α be any element in the preset stiffness matrix; α0…α6 are a series of linear parameters that map the squares of the interaction force / moment amplitudes to the elements of the stiffness matrix; the damping B(t) is a matrix containing i*j elements, B ij (t) represents any element in the preset damping matrix; i = 1, 2, 3, j = 1, 2, 3; β0…β6 are a series of linear parameters that map the absolute values ​​of the interaction force / torque amplitudes to the elements of the damping matrix; f is the setpoint for the interaction force / torque. v A component of (t) has subscripts x, y and z that represent the interaction force / torque components along the X, Y and Z axes, respectively; + and - represent the two opposite directions of the interaction force / torque.

3. The method for verifying the accuracy of human body mechanical impedance identification according to claim 1 or 2, characterized in that, In step four, an impedance dynamics model is established, with the restoring force / torque preset to f. r (t)=f I (t)+f B (t)+f K (t); Among them, the inertial dynamics model Describe the inertia component f in the restoring force / torque I (t) with inertia I and acceleration / angular acceleration The dynamic relationship between them; Damped dynamic model Describe the damping component f in the restoring force / torque B (t) and time-varying damping value B(t) and velocity / angular velocity The dynamic relationship between them; Stiffness dynamics model f K (t)=K(t)q(t) describes the stiffness component f in the restoring force / torque. K The dynamic relationship between (t) and time-varying stiffness K(t) and displacement / angle q(t); In step five, set the white noise f. n (t), through the preset value f of the interaction force / torque v (t), preset value of restoring force / torque f r (t) and white noise setpoint f n By superimposing (t), the simulated measured force / torque f is obtained. o2 (t)=f v (t)+f r (t)+f n (t); Where q(t) is the random perturbation displacement / angle set in step one, its first and second derivatives are calculated to obtain the perturbation velocity / angular velocity, respectively. With acceleration / angular acceleration 4. The method for verifying the accuracy of human body mechanical impedance identification according to claim 1 or 3, characterized in that, The method for extracting the simulated restoring force / torque from the simulated measured force / torque in step six is ​​as follows: depending on the type of interaction force / torque, analytical or numerical methods are used to extract the simulated measured force / torque f. o2 Extract the simulated restoring force / torque f′ used to identify impedance parameters from (t). r (t); where, when the interaction force / torque f v When (t) is 0, the simulated restoring force / torque f′ r (t) equals the simulated measured force / torque f o2 (t); when the interaction force / torque f v When (t) is a constant value that is not equal to 0, the simulated measured force / torque f o2 (t) Perform a mean-removal operation to obtain the simulated restoring force / torque f′ r (t); when the interaction force / torque f v (t) When it has time-varying characteristics, the simulated measured force / torque f is analyzed using numerical methods. o2 The restoring force / torque and the interaction force / torque in (t) are decoupled to obtain the simulated restoring force / torque f′. r (t).

5. The method for verifying the accuracy of human body mechanical impedance identification according to claim 1, 3, or 4, characterized in that, The disturbance displacement / angle q(t) and its velocity / angular velocity are... Acceleration / Angular Acceleration As input to the human body impedance identification model, the simulated restoring force / torque f′ is used. r (t) is the output of the human body impedance identification model. Using the human body impedance identification model and method to be verified, the inertia, damping, and stiffness of the human limb extremities / joints are identified to obtain the inertia identification value I′, damping identification value B′(t), and stiffness identification value K′(t); the following relationships are satisfied: The accuracy of the human body impedance identification method is verified by comparing the identified impedance parameter values ​​I′, B′(t), and K′(t) with the preset impedance parameter values ​​I, B(t), and K(t) using the variance contribution rate and root mean square error evaluation index.

6. A system for verifying the accuracy of human body mechanical impedance identification methods, characterized in that, include: The simulation input module is used to set the disturbance displacement / angle information, and sequentially calculate the first and second derivatives of the disturbance displacement / angle information to obtain the velocity / angular velocity information and acceleration / angular acceleration information during the disturbance duration; and to set the time-varying or time-invariant interactive force / torque level according to the characteristics of the interactive task. The impedance parameter preset module is used to preset the stiffness based on the square of the interaction force / torque and its linear mapping relationship with stiffness, preset the damping based on the linear mapping relationship between the interaction force / torque and damping, and randomly preset the inertia; among them, the linear parameters reflecting the linear mapping relationship are all randomly set. The restoring force / torque preset module is used to establish an impedance dynamics model and obtain preset values ​​for the restoring force / torque. Specifically, displacement / angle information is input into the stiffness dynamics model, velocity / angular velocity information is input into the damping dynamics model, and acceleration / angular acceleration information is input into the inertial dynamics model. Combined with preset stiffness, damping, and inertia, the stiffness component, damping component, and inertial component of the restoring force / torque are obtained. The three components are then summed to obtain the preset value of the restoring force / torque. The output module is used to superimpose the preset restoring force / torque with the set interaction force / torque to obtain a noise-free measured force / torque; and to add random white noise to the noise-free measured force / torque output signal to obtain simulated measured force / torque information. The human impedance identification module is used to extract simulated restoring force / torque from simulated measured force / torque using analytical or numerical methods, based on the different characteristics of interactive forces / torques in different interactive tasks. It simulates the process of extracting restoring force / torque from force / torque data measured by sensors. It uses perturbation displacement / angle, velocity / angular velocity, and acceleration / angular acceleration as input signals and simulated restoring force / torque as output signals. It uses the human impedance identification method to be verified to identify the stiffness, damping, and inertia of human limb extremities / joints. The impedance parameter identification accuracy verification module is used to compare the identified values ​​of human body stiffness, damping, and inertia with preset stiffness, damping, and inertia, respectively, and to verify the accuracy of the human body mechanical impedance identification method through variance contribution rate and root mean square error evaluation indicators.

7. The system for verifying the accuracy of the human body mechanical impedance identification method according to claim 6, characterized in that, The impedance parameter preset module sets the interaction force / torque f according to the interaction task type. v The magnitude of (t); the inertia I in the randomly preset mechanical impedance, and the stiffness K(t) and damping B(t) in the preset mechanical impedance; Each component of stiffness has a linear relationship with the square of the interaction force / moment amplitude: Each component of the damping has a linear relationship with the absolute value of the interaction force / torque amplitude: Wherein, stiffness K(t) is a matrix containing i*j elements; K ij Let α be any element in the preset stiffness matrix; α0…α6 are a series of linear parameters that map the squares of the interaction force / moment amplitudes to the elements of the stiffness matrix; the damping B(t) is a matrix containing i*j elements, B ij (t) represents any element in the preset damping matrix; β0…β6 are a series of linear parameters that map the absolute values ​​of the interaction force / torque amplitudes to the elements of the damping matrix; i = 1, 2, 3, j = 1, 2, 3; f is the setpoint for the interaction force / torque. v A component of (t) has subscripts x, y and z that represent the interaction force / torque components along the X, Y and Z axes, respectively; + and - represent the two opposite directions of the interaction force / torque.

8. The system for verifying the accuracy of the human body mechanical impedance identification method according to claim 6 or 7, characterized in that, An impedance dynamics model is established in the restoring force / torque preset module, with the restoring force / torque preset value being f. r (t)=f I (t)+f B (t)+f K (t); Among them, the inertial dynamics model Describe the inertia component f in the restoring force / torque I (t) with inertia I and acceleration / angular acceleration The dynamic relationship between them; Damped dynamic model Describe the damping component f in the restoring force / torque B (t) and time-varying damping value B(t) and velocity / angular velocity The dynamic relationship between them; Stiffness dynamics model f K (t)=K(t)q(t) describes the stiffness component f in the restoring force / torque. K The dynamic relationship between (t) and time-varying stiffness K(t) and displacement / angle q(t); Where q(t) is the set random perturbation displacement / angle, its first and second derivatives are obtained to get the perturbation velocity / angular velocity, respectively. With acceleration / angular acceleration 9. The system for verifying the accuracy of the human body mechanical impedance identification method according to claim 6, characterized in that, The method by which the human body impedance identification module extracts the simulated restoring force / torque from the simulated measured force / torque is as follows: Output module setting white noise f n (t), through the preset value f of the interaction force / torque v (t), preset value of restoring force / torque f r (t) and white noise setpoint f n By superimposing (t), the simulated measured force / torque f is obtained. o2 (t)=f v (t)+f r (t)+f n (t); Depending on the type of interactive force / torque, analytical or numerical methods are used to obtain the simulated and measured force / torque f. o2 Extract the simulated restoring force / torque f′ used to identify impedance parameters from (t). r (t); where, when the interaction force / torque f v When (t) is 0, the simulated restoring force / torque f′ r (t) equals the simulated measured force / torque f o2 (t); when the interaction force / torque f v When (t) is a constant value that is not equal to 0, the simulated measured force / torque f o2 (t) Perform a mean-removal operation to obtain the simulated restoring force / torque f′ r (t); when the interaction force / torque f v (t) When it has time-varying characteristics, the simulated measured force / torque f is analyzed using numerical methods. o2 The restoring force / torque and the interaction force / torque in (t) are decoupled to obtain the simulated restoring force / torque f′. r (t).

10. The system for verifying the accuracy of the human body mechanical impedance identification method according to claim 6 or 9, characterized in that, The human body impedance identification module will include the disturbance displacement / angle q(t) and its velocity / angular velocity. Acceleration / Angular Acceleration As input, simulate the restoring force / torque f′ r (t) is used as the output. The human body impedance identification method to be verified is used to identify the inertia, damping, and stiffness of the human limb extremities / joints, and to obtain the inertia identification value I′, the damping identification value B′(t), and the stiffness identification value K′(t); the following relationship is satisfied: The accuracy of the human body impedance identification method is verified by comparing the identified impedance parameter values ​​I′, B′(t), and K′(t) with the preset impedance parameter values ​​I, B(t), and K(t) using the variance contribution rate and root mean square error evaluation index.

Images