Self-correction method for sensing model of electrothermal driving type artificial muscle

Abstract

The invention discloses a sensing model self-correction method for electrothermal driving type artificial muscle, which comprises the following steps of: acquiring inductance values of the artificial muscle under different lengths by using an inductance measuring device, fitting, and constructing a fitting formula capable of realizing accurate association of the inductance and the length of the artificial muscle, the model is also a perceptual model of artificial muscles; the inductance measuring device is used for continuously recording the inductance value of the artificial muscle under cyclic driving, and then transmitting the acquired inductance value to the computer through the digital-to-analog conversion chip; the computer judges whether the artificial muscle generates overload and overdrive circulation or not by judging the maximum value of the inductance in each driving process to enable the structure to generate irreversible distortion, if the structure is distorted, the maximum value of the inductance can be reduced, and if the maximum value of the inductance is kept unchanged all the time, the sensing model is not modified. According to the method, the perception model of the artificial muscle can still be used under the condition that the structure is changed.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of flexible sensors, and particularly to a self-correction method for a sensing model of an electro-thermal driving artificial muscle. BACKGROUND

[0002] Soft robots are a class of robots made of flexible materials that mimic the structure or movement of living organisms. Compared with traditional rigid robots, soft robots have higher adaptability, flexibility and safety, and exhibit unique advantages in medical, exploration, human-computer interaction and other fields. Soft actuators are the core components of soft robots, responsible for converting energy into mechanical movement. Unlike traditional rigid actuators (such as motors, hydraulic cylinders), soft actuators achieve movement through the deformation of flexible materials, with continuous deformation, high adaptability and biomimetic characteristics.

[0003] The commonly used soft actuator materials currently include shape memory alloys, dielectric elastomers, ionomer metal composite materials, and fluid-driven materials, but the above materials face two major problems of high cost and complex preparation when applied to soft robots. With the rapid development of the field of soft robots, there is a demand for lightweight, high energy density, and biocompatible driving methods, which has driven the research of twisted artificial muscles. Researchers have found that the twisted structure of high polymer fibers (such as nylon, polyethylene) can achieve a contraction-dilation behavior similar to that of a muscle, i.e., an artificial muscle.

[0004] In recent years, a mature commercial nylon filament with a silver layer has emerged. The twisted high polymer artificial muscle made from this nylon filament through twisting, winding, and annealing processes has both self-sensing and driving functions, and has more convenient application conditions. In order to solve the problem of difficult displacement measurement of artificial muscles, the existing resistance-length model and temperature-length model are used, but the measurement values of these two models are affected by many factors and cannot reflect the single corresponding relationship with length. Therefore, the present application proposes an inductance-length model design method using the inductance characteristics, and a self-correction method according to the actual measurement value of inductance is designed to improve the applicability of the inductance-length model, in view of the problem of permanent deformation of the structure caused by excessive load or exceeding the maximum number of stretches, which leads to a decrease in model accuracy. SUMMARY

[0005] The present application aims to overcome the shortcomings and deficiencies of the prior art, and provides a self-correction method for a sensing model of an electro-thermal driving artificial muscle. The self-correction method according to the actual measurement value of inductance is designed to improve the applicability of the sensing model, in view of the problem that the overstretching of the artificial muscle will cause the structure to distort and change the geometric parameters of the artificial muscle, resulting in the failure of the original sensing model. The self-correction method ensures that the sensing model of the artificial muscle can still be used even if the structure changes.

[0006] To achieve the above object, the technical scheme provided by the application is a self-modification method for a sensing model of an electrothermal driving type artificial muscle, comprising the following steps:

[0007] S1, using an inductance measuring device to collect inductance values under different lengths of the artificial muscle and performing fitting to construct a fitting formula capable of realizing accurate correlation between inductance and length of the artificial muscle, the fitting formula being a polynomial formula and also a sensing model of the artificial muscle;

[0008] S2, using the inductance measuring device to continuously record inductance values of the artificial muscle under cyclic driving, and then transmitting the collected inductance values to a computer through a digital-analog conversion chip, the computer judging a maximum value of inductance in each driving process to judge whether the artificial muscle is subjected to overloading and overdrive cycle to cause irreversible distortion of the structure, that is, a part of the artificial muscle cannot contract when it should contract, if the structure is distorted, the maximum value of inductance will decrease, if the maximum value of inductance remains unchanged, the sensing model is not modified; wherein, if it is determined that the artificial muscle is subjected to structural distortion, the parameters of the correction term in the sensing model are re-fitted, after the fitting is completed, fitting degree analysis is performed, the formula is brought into the recorded inductance value of the current inductance peak value to calculate the square deviation, if the deviation is within the preset range, it indicates that the correction is completed, and a new polynomial formula is output as the final sensing model, if the deviation exceeds the preset range, the peak value of inductance is continuously selected for correction until the deviation of the formula is within the preset range, and a new polynomial formula is output as the final sensing model.

[0009] Further, the step S1 comprises the following steps:

[0010] S11, using a digital bridge instrument to collect inductance data under different lengths of the artificial muscle and performing recording, and then inputting the data into MATLAB mathematical software;

[0011] S12, constructing a formula framework with inverse proportion and non-linearity to represent the correlation between inductance and length of the artificial muscle, as shown below:

[0012]

[0013] In the formula, L represents inductance, a and c are fitting parameters based on the curve change trend, and l is the length of the artificial muscle;

[0014] S13, using MATLAB to add a supplementary term capable of dynamically changing after the formula (1) through the least squares method, so that a correction term bl is added to the formula (1) m :

[0015]

[0016] Wherein, b, m are correction term constants.

[0017] Further, the step S2 comprises the following steps:

[0018] S21, after the initial inductance data has been collected and the inductance-length fitting formula has been obtained by initial fitting, the output data line of the digital bridge instrument is connected to the ADC input channel of the digital-analog conversion chip, and the inductance value, a continuously changing physical quantity, is converted into a digital quantity recognizable by the single-chip microcomputer through the digital-analog conversion chip; the single-chip microcomputer, as a microcomputer, is a data processing center;

[0019] S22, the timer inside the single-chip microcomputer is started; firstly, the digital-analog conversion chip is started by configuring the pin value of the digital-analog conversion chip; the working mode and the count value of the timer are set by configuring the corresponding timer control register to match the required rate of data acquisition; then, the timer interrupt is started, so that the single-chip microcomputer can execute the subsequent sampling task at the set time interval;

[0020] S24, after the analog-digital conversion is completed, the single-chip microcomputer stores the received digital signal into a designated data buffer for subsequent processing and analysis; for the sampled data, two error processing mechanisms, i.e., timeout detection and signal amplitude overrun, are set to ensure the reliability of the sampling process and the effectiveness of the data;

[0021] S25, after the sampling is completed, the data is filtered and smoothed, and the statistical characteristics thereof are calculated;

[0022] S26, for the peak value of the inductance value in each driving process after the statistics, if the peak value decreases, it is determined that the artificial muscle has overloading and overdriving cycle to cause irreversible distortion of the structure, a correction instruction is issued, the polynomial formula is corrected through the peak value of the new inductance value, and if the peak value does not change, it is determined that the artificial muscle structure is normal and does not need to be corrected;

[0023] S27, a correction algorithm of the perception model is designed; after the single-chip microcomputer issues the correction instruction, the single-chip microcomputer automatically selects the updated peak value of the inductance value as the inductance initial value, and defines the inductance initial value as a characteristic inductance value; at this time, the length of the artificial muscle is taken as a characteristic length corresponding to the characteristic inductance value, to form a characteristic data pair (l feature , L feature ); the characteristic data pair is brought into the fitting formula to update the correction term constants b and m, to obtain updated parameters b' and m':

[0024]

[0025] Wherein, l feature represents the characteristic length, and L featureThe representative characteristic inductance; through the updated parameters b', m', a new fitting formula can be obtained, that is, the self-correction process is completed;

[0026] S28, verifying the calculation accuracy of the corrected formula, selecting the inductance L of the record period of the inductance peak meas The calculation result L of the adaptive fitting formula pred Comparison, using root mean square error RMSE:

[0027]

[0028] In the formula, n represents the number of inductance data in the record period, i represents the bit number of inductance data, L pred,i represents the predicted value of the inductance at the i-th bit, L meas,i represents the measured value of the inductance at the i-th bit.

[0029] If the prediction accuracy deviation is greater than the preset range In the preset range, wherein L meas is the measured value of the inductance, the corrected fitting formula is output, and then the next inductance collection cycle is entered; if the prediction accuracy deviation is not in the preset range, return to data screening, re-establish the fitting parameters, calculate the parameters b and m, and cycle accordingly.

[0030] Compared with the prior art, the present application has the following advantages and beneficial effects:

[0031] The existing research provides various implementation methods for length self-sensing of the polymer twisted type artificial muscle, including temperature self-sensing method and resistance self-sensing method. The temperature self-sensing method detects the artificial surface temperature by detecting the change of the resistance value of the thermistor generated with temperature, and then performs data fitting between the temperature and the length to realize the self-sensing of the muscle length. The resistance self-sensing method directly monitors the resistance value (input voltage / current) of the silver plating layer on the muscle surface. Since the silver plating layer has metal characteristics, its resistivity will change with temperature, and the temperature change will cause the length change. Therefore, the muscle length change can be reflected by the resistance value, and the self-sensing of the artificial muscle length is realized.

[0032] Although the existing temperature self-sensing method and resistance self-sensing method realize length sensing by associating the muscle length with other parameters, the above methods have certain limitations. The temperature self-sensing method and the resistance self-sensing method establish an artificial muscle length model by means of data fitting, but the length of the artificial muscle is related to the driving temperature and the applied load. When the load changes, the displacement produced by applying the same driving voltage (temperature) is different, resulting in poor adaptability of the model. In addition, none of the three models has a correction function. If the muscle is stretched too much during the stretching process, structural creep or even local plastic deformation will occur, which will cause the original fitting curve to be unable to adapt to the deformed model.

[0033] The present application fully considers the problem of insufficient adaptability of the existing prediction model, combines the fitting correction term parameters in the fitting curve of the sensing model, and proposes a method based on the correction of the parameters, which can more accurately predict the length of the artificial muscle during the driving process, fills the gap in the existing research, realizes high-precision prediction of the length of the artificial muscle, and promotes the application of the polymer twisted artificial muscle in the field of soft robots. BRIEF DESCRIPTION OF DRAWINGS

[0034] Figure 1 It is a sensing model schematic diagram. In the diagram, X is the axial length, AX is the change amount of the axial length, F is the pulling force for changing the total length, B is the magnetic flux, under the action of F, the axial length increases, the interval of the circular ring increases, the leakage magnetic flux increases, B decreases, and then the inductance decreases. This diagram is to illustrate that the inductance of the artificial muscle and the length are inversely proportional.

[0035] Figure 2 It is a self-correction method schematic diagram. DETAILED DESCRIPTION

[0036] The present application will be further described in detail below in combination with the embodiments and the drawings, but the embodiments of the present application are not limited thereto.

[0037] As shown in Figure 1 and Figure 2 , the present embodiment discloses a self-correction method of a sensing model of an electrothermal driving type artificial muscle, and the specific circumstances are as follows:

[0038] S1, using an inductance measuring device to collect inductance values of the artificial muscle at different lengths and performing fitting to construct a fitting formula capable of accurately correlating the inductance-length of the artificial muscle, the fitting formula is a polynomial formula, which is also a sensing model of the artificial muscle, including the following steps:

[0039] S11, using a digital bridge instrument to collect inductance data of the artificial muscle at different lengths and recording, and then inputting the data into MATLAB mathematical software;

[0040] S12, construct a formula framework with inverse proportion, non-linear to represent the relationship between artificial muscle inductance and length, as follows:

[0041]

[0042] In the formula, L represents inductance, a and c are fitting parameters based on the trend of curve change, and l is the length of artificial muscle;

[0043] S13, use MATLAB to add a dynamic supplementary term to formula (1) by least square method, so as to add a correction term bl to formula (1) m :

[0044]

[0045] In the formula, b and m are correction term constants. After experimental verification, the corrected formula shows high consistency between calculated inductance value and measured inductance value, and has high precision.

[0046] S2, use inductance measuring device to continuously record the inductance value of artificial muscle under cyclic driving, and then transmit the collected inductance value to computer through digital-analog conversion chip. The computer judges the maximum value of inductance in each driving process to determine whether the artificial muscle is overloaded and overdriven cycle to cause irreversible distortion of the structure, that is, the artificial muscle cannot contract at a certain place where it should contract. If the structure is distorted, the maximum value of inductance will decrease. If the maximum value of inductance remains unchanged, the perception model will not be modified. If it is determined that the artificial muscle has structural distortion, the correction term parameters in the perception model are re-fitted. After fitting is completed, the fitting degree is analyzed, and the formula is brought into the recorded inductance value of the current inductance peak of the computer to calculate the square deviation. If the deviation is within the preset range, it means that the correction is completed, and the new polynomial formula is output as the final perception model. If the deviation exceeds the preset range, the peak value of inductance is selected for correction until the deviation of the formula is within the preset range, and the new polynomial formula is output as the final perception model. The method comprises the following steps:

[0047] S21, after collecting initial inductance data and performing initial fitting to obtain inductance-length fitting formula, connect the output data line of digital bridge instrument to the ADC input channel of digital-analog conversion chip, and convert the continuously changing physical quantity such as inductance value into digital quantity that can be recognized by single-chip microcomputer through digital-analog conversion chip. The single-chip microcomputer is a microcomputer and is the data processing center.

[0048] S22. Start the microcontroller's internal timer. First, start the digital-to-analog converter chip by configuring the pin values ​​of the digital-to-analog converter chip. Configure the corresponding timer control register to set the timer's working mode and count value to match the data acquisition rate required. Then, enable the timer interrupt so that the microcontroller can execute subsequent sampling tasks according to the set time interval.

[0049] S24. After the analog-to-digital conversion is completed, the microcontroller stores the received digital signal into the specified data buffer for subsequent processing and analysis. For the sampled data, two error handling mechanisms are set up: timeout detection and signal amplitude exceeding the limit, to ensure the reliability of the sampling process and the validity of the data.

[0050] S25. After sampling, the data is filtered and smoothed, and its statistical characteristics are calculated.

[0051] S26. For the peak value of the inductance value in each driving process after statistics, if the peak value decreases, it is determined that the artificial muscle has been overloaded and overdriven, causing irreversible distortion of the structure. A correction instruction is issued, and the polynomial formula is corrected by the new peak value of the inductance value. If the peak value does not change, it is determined that the artificial muscle structure is normal and no correction is needed.

[0052] S27. Design a correction algorithm for the perception model. After the microcontroller issues a correction command, the microcontroller automatically selects the peak value of the updated inductance value as the initial inductance value. This initial inductance value is defined as the feature inductance value. At this time, the length of the artificial muscle is taken as the corresponding feature length under the feature inductance value, forming a feature data pair (l feature ,L feature Substitute the feature data pairs into the fitting formula to update the correction term constants b and m, obtaining the updated parameters b' and m':

[0053]

[0054] In the formula, l feature L represents the feature length. feature The characteristic inductance is represented; a new fitting formula can be obtained through the updated parameters b' and m', thus completing the self-correction process;

[0055] S28. Verify the accuracy of the corrected formula calculation by selecting the inductance L during the recording period of the inductance peak. meas The result L calculated by the adaptive fitting formula pred For comparison, the root mean square error (RMSE) is used:

[0056]

[0057] In the formula, n represents the number of inductance data points in the recording period, i represents the number of bits in the inductance data, and L...pred,i represents the predicted value of the inductance at the i-th bit, L meas,i represents the measured value of the inductance at the i-th bit;

[0058] If the prediction accuracy deviation (which can be simply referred to as the prediction deviation) is less than or equal to 1.5%, wherein L meas is the measured value of the inductance, then a revised fitting formula is outputted, and then the next inductance collection cycle is entered; if the prediction accuracy deviation is greater than 1.5%, then data screening is returned to, the fitting parameters are re-established, the parameters b and m are calculated, and the cycle is repeated.

[0059] The above embodiments are preferred embodiments of the present application, but the embodiments of the present application are not limited by the above embodiments, and any changes, modifications, substitutions, combinations, simplifications made without departing from the spirit and principles of the present application are equivalent replacement methods and are included in the protection scope of the present application.

Claims

1. A method for self-correction of a perception model of an electrothermal drive type artificial muscle, characterized by, Comprise the following steps: S1, using inductance measuring device to collect artificial muscle under different length inductance value and fitting, construct can realize accurate correlation artificial muscle inductance-length fitting formula, the fitting formula is polynomial formula, also is the perception model of artificial muscle; S2, using inductance measuring device continuously records the inductance value of artificial muscle under cyclic driving, then the collected inductance value is transmitted to computer through digital analog conversion chip, the computer judges the maximum value of inductance in each driving process to judge whether artificial muscle occurs overload and overdrive cycle makes the structure produce irreversible distortion, that is, artificial muscle cannot contract at a certain place, if the structure distortion will lead to the maximum value of inductance decreases, if the maximum value of inductance remains unchanged, the perception model is not modified;Among them, if it is determined that the artificial muscle has structural distortion, the modified term parameters in the perception model are re-fitted, and after the fitting is completed, the fitting degree is analyzed, the formula is brought into the recorded inductance value of the current inductance peak of the computer, and the square deviation is calculated, if the deviation is within the preset range, the correction is completed, and the new polynomial formula is output as the final perception model, if the deviation exceeds the preset range, the peak of inductance value is selected for correction until the deviation of the formula is within the preset range, and the new polynomial formula is output as the final perception model.

2. The self-correcting method of a perception model of an electro-thermal driving artificial muscle according to claim 1, characterized in that, The step S1 comprises the following steps: S11, using digital bridge instrument to collect inductance data under different lengths of artificial muscle and record, then inputting the data into MATLAB mathematical software; S12, a formula framework with inverse proportion, nonlinearity is constructed to represent the correlation between artificial muscle inductance and length, as follows: In the formula, L represents inductance, a and c are fitting parameters based on the curve change trend, and l is the length of artificial muscle; S13, using MATLAB to add a dynamic supplementary term to the back of formula (1) by the least square method, so as to add a correction term bl to formula (1) m : In the formula, b and m are correction term constants.

3. The self-correcting method of a perception model of an electro-thermal drive artificial muscle according to claim 2, characterized in that, The step S2 comprises the following steps: S21, after collecting the initial inductance data and performing initial fitting to obtain the inductance-length fitting formula, the output data line of the digital bridge instrument is connected to the ADC input channel of the digital analog conversion chip, the inductance value of the continuous changing physical quantity is converted into digital quantity that can be recognized by the single-chip microcomputer through the digital analog conversion chip, and the single-chip microcomputer is the data processing center as a microcomputer; S22, start the timer in the single-chip microcomputer, first start the digital analog conversion chip by configuring the pin value of the digital analog conversion chip, configure the timer control register to set the working mode and count value of the timer to match the required rate of data acquisition, then start the timer interrupt to enable the single-chip microcomputer to perform subsequent sampling tasks at the set time interval; S24, after the analog-digital conversion is completed, the single-chip microcomputer stores the received digital signal in the specified data buffer for subsequent processing and analysis, sets the timeout detection and signal amplitude overrun error processing mechanisms to ensure the reliability of the sampling process and the effectiveness of the data; S25, after sampling, filter and smooth the data, and calculate its statistical characteristics; S26, for each drive process after the statistics inductance value peak, if the peak is reduced, it is determined that artificial muscle occurs overload and overdrive cycle makes the structure produces irreversible distortion, issues a correction instruction, through the new inductance value peak to modify the polynomial formula, if the peak does not change, it is determined that the artificial muscle structure is normal, and no correction is needed; S27, design the correction algorithm of the perception model, after the single-chip microcomputer issues a correction instruction, the single-chip microcomputer automatically selects the peak value of the updated inductance value as the inductance initial value, defines the inductance initial value as the characteristic inductance value, at this time the length of the artificial muscle is taken as the characteristic length corresponding to the characteristic inductance value, to form a characteristic data pair (l feature ,L feature ), and the characteristic data pair is brought into the fitting formula to update the correction term constants b and m, to obtain the updated parameters b' and m': In the formula, l feature L represents a characteristic length feature L represents a characteristic inductance; a new fitting formula can be obtained through the updated parameters b' and m', that is, the self-correction process is completed; S28, check the accuracy of the modified formula, select the record period of inductance L meas The calculation result L of the fitted formula after adaptation pred Comparison, using root mean square error RMSE: where n represents the number of inductance data in a recording period, i represents the number of bits of inductance data, L pred,i represents the predicted value of inductance at the i-th bit, L meas,i represents the measured value of inductance at the i-th bit; If the prediction accuracy deviation is within the preset range Within a preset range, wherein L meas is the measured value of inductance, the corrected fitting formula is output, and then the next inductance collection cycle is entered; if the prediction accuracy deviation is not within the preset range, the data screening is returned, the fitting parameters are re-established, the parameters b and m are calculated, and the cycle is repeated.

Images