Mechanical arm compliance control method based on sensing monitoring

Abstract

The invention relates to the technical field of mechanical arm control, in particular to a mechanical arm compliance control method based on sensing monitoring. The method comprises the following steps: acquiring a relative load degree characteristic value of a monitoring position at a moment; the relative rigidity degree estimation coefficient of each monitoring position at each moment is obtained, and then the rigidity estimation credibility of each monitoring position at each moment is obtained by combining the rotating speed of the motor; obtaining a final decomposition weight of a monitoring position at each moment, and decomposing the relative rigidity degree estimation coefficient of the monitoring position at each moment based on an STL algorithm to obtain a trend term and a periodic term; predicting a predicted value of the relative rigidity degree estimation coefficient at the next moment; and based on the relative rigidity degree estimation coefficient predicted value of one monitoring position at the next moment, the relative rigidity degree estimation coefficient at each historical moment and the relative rigidity degree estimation coefficient predicted value, a predicted rigidity matrix adjustment coefficient is obtained, and the mechanical arm is controlled. The mechanical arm can be effectively controlled.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of mechanical arm control, and in particular to a mechanical arm compliant control method based on sensing monitoring. BACKGROUND

[0002] With the continuous improvement of the intelligence and automation level of the power system, the application demand of mechanical arms in power operation and maintenance is increasing. Especially in high-risk and unstructured operation scenes such as substations, mechanical arms can replace manual work to complete tasks such as detection, maintenance and switch operation, thereby effectively reducing labor costs and safety risks. However, under the conditions of a large number of device types, narrow space and sudden environmental changes, traditional mechanical arms often lack compliance and adaptive ability, which can easily lead to insufficient operation precision and even collision risks. In recent years, with the introduction of sensor technology and intelligent algorithms, the perception and control performance of mechanical arms has improved, but the efficiency of the compliant control strategy in the operation process is still limited, and it is difficult to balance the safety while ensuring the high precision and flexibility of the operation. Therefore, how to realize the compliant control of the mechanical arm in the operation and maintenance of the power equipment and improve its stability and accuracy in complex tasks plays an important role in the high-precision operation of the power equipment and the efficient completion of complex tasks.

[0003] In the compliant control of the mechanical arm, the existing variable impedance control method (VIC) often relies on static or preset parameters when adjusting the stiffness matrix, and the response to the change of environmental stiffness is insufficient. For example, when the mechanical arm contacts switches or cables of different hardness in the operation of the substation, the fixed stiffness setting can cause excessive impact on the hard switch, and the action on the soft cable is not powerful enough. The influence of sensor accuracy and mechanical arm dynamics interference leads to inaccurate or lagging stiffness adjustment, etc., thereby leading to poor adaptability of the mechanical arm in nonlinear or dynamic environmental characteristics, and it is difficult to ensure safe and efficient compliant operation of the mechanical arm in complex tasks. SUMMARY

[0004] In order to solve the above technical problems, the purpose of the present application is to provide a mechanical arm compliant control method based on sensing monitoring, and the technical solution adopted is as follows: One embodiment of the present application provides a mechanical arm compliant control method based on sensing monitoring, which comprises: Collecting the input current, loop current, input voltage, loop voltage and motor speed of the joint motor at each moment of each monitoring position of the mechanical arm; obtaining the relative load degree characteristic value at the moment according to the input current, loop current, input voltage and loop voltage of one monitoring position at the moment; According to the relative load degree characteristic value of one monitoring position at one time and the motor speed at the time and the previous time, the relative rigidity degree estimation coefficient of the time is obtained; according to the relative rigidity degree estimation coefficients of each adjacent time and the motor speed of one monitoring position at one time, the rigidity estimation credibility of the time is obtained; According to the rigidity estimation credibility of one monitoring position at each time and the initial decomposition weight of the STL algorithm, the final decomposition weight of each time is obtained; and the relative rigidity degree estimation coefficients of one monitoring position at each time are decomposed based on the STL algorithm to obtain the trend term and the periodic term by combining the final decomposition weight of one monitoring position at each time. Based on the trend term and the periodic term corresponding to one monitoring position, the relative rigidity degree estimation coefficient prediction value of the next time is predicted; based on the relative rigidity degree estimation coefficient prediction value of one monitoring position at the next time, the relative rigidity degree estimation coefficients and the relative rigidity degree estimation coefficient prediction values of each historical time, the rigidity matrix adjustment coefficient of the next time is obtained; and the mechanical arm is controlled based on the rigidity matrix adjustment coefficient of each monitoring position at the next time and the variable impedance control strategy.

[0005] Preferably, the relative load degree characteristic value of one monitoring position at one time is obtained according to the input current, the loop current, the input voltage and the loop voltage of the monitoring position at the time, and the relative rigidity degree estimation coefficient of the time is obtained according to the relative load degree characteristic value of one monitoring position at one time and the motor speed at the time and the previous time. The ratio of the loop current of one monitoring position at one time to the input current of the time is denoted as a first ratio; the ratio of the input voltage of the monitoring position at the time to the loop voltage of the time is denoted as a second ratio; and the relative load degree characteristic value of the monitoring position at the time is obtained by multiplying the first ratio and the second ratio.

[0006] Preferably, the relative rigidity degree estimation coefficient of one monitoring position at one time is obtained according to the relative load degree characteristic value of one monitoring position at one time and the motor speed at the time and the previous time. The difference between the relative load degree characteristic values of one monitoring position at one time and the previous time is negatively correlated by using an exponential function with a natural constant as the base to obtain a first mapping value of the time; the difference between the motor speeds of the monitoring position at the time and the previous time is negatively correlated by using an exponential function with a natural constant as the base to obtain a second mapping value; and the relative rigidity degree estimation coefficient of the monitoring position at the time is obtained by comparing the first mapping value and the second mapping value.

[0007] Preferably, the rigidity estimation credibility of one monitoring position at one time is obtained according to the relative rigidity degree estimation coefficients of each adjacent time and the motor speed of one monitoring position at one time. Taking a time point as a starting point, a preset number of time points before and after the time point are taken respectively, and the obtained time points are recorded as the neighborhood time points of the time point; the time interval between a neighborhood time point of a time point and the time point is negatively related mapped by an exponential function with a natural constant as a base to obtain a time distance weight of the neighborhood time point of the time point; for a monitoring position, the absolute value of the difference between the relative rigidity degree estimation coefficient of the time point and the relative rigidity degree estimation coefficients of the neighborhood time points of the time point is weighted and averaged by using the time distance weights of the neighborhood time points of the time point to obtain a first average difference; the absolute value of the difference between the motor speed of the time point and the motor speeds of the neighborhood time points of the time point is weighted and averaged by using the time distance weights of the neighborhood time points of the time point to obtain a second average difference; the ratio of the first average difference to the product of the second average difference and the motor speed of the time point is negatively related mapped by an exponential function with a natural constant as a base to obtain the rigidity estimation credibility of the monitoring position at the time point.

[0008] Preferably, the final decomposition weight of each time point is obtained according to the rigidity estimation credibility of each time point of a monitoring position and the initial decomposition weight of the STL algorithm, comprising: The rigidity estimation credibility of a time point of a monitoring position is multiplied by the initial decomposition weight when the relative rigidity degree estimation coefficient of the time point is locally weighted and decomposed by the STL algorithm to obtain a weight factor of the time point, and the sum of the rigidity estimation credibilities of all time points within the local range when the relative rigidity degree estimation coefficient of the time point is locally weighted and decomposed by the STL algorithm is compared with the weight factor of the time point to obtain the final decomposition weight of the monitoring position at the time point.

[0009] Preferably, the relative rigidity degree estimation coefficient prediction value of the next time point is predicted based on the trend item and the periodic item corresponding to a monitoring position, comprising: The trend prediction value and the periodic prediction value of the next time point are obtained by respectively predicting the trend item and the periodic item corresponding to a monitoring position by using the GRU algorithm; the relative rigidity degree estimation coefficient prediction value of the next time point is obtained by adding the trend prediction value and the periodic prediction value of the next time point.

[0010] Preferably, the rigidity matrix adjustment coefficient of the next time point is obtained based on the relative rigidity degree estimation coefficient prediction value of the next time point, the relative rigidity degree estimation coefficients of each historical time point and the relative rigidity degree estimation coefficient prediction value, comprising: if the relative rigidity degree estimation coefficient prediction value of the next moment of a monitoring position is greater than or equal to the relative rigidity degree estimation coefficient of the current moment, the time distance weight of a historical moment is obtained by using the exponential function with a natural constant as a base to negatively correlate the time interval between the historical moment and the current moment; the third average difference is obtained by weighting and averaging the absolute value of the difference between the relative rigidity degree estimation coefficient prediction value and the relative rigidity degree estimation coefficient of each historical moment based on the time distance weight of each historical moment; the third mapping value is obtained by using the exponential function with a natural constant as a base to negatively correlate the third average difference; the relative change degree is obtained by using the first preset value to subtract the ratio of the relative rigidity degree estimation coefficient prediction value of the next moment to the relative rigidity degree estimation coefficient of the current moment and taking the absolute value; the adjustment amplitude is obtained by multiplying the rigidity matrix adjustment coefficient of the current moment, the relative change degree and the third mapping value; and the rigidity matrix adjustment coefficient of the next moment is obtained by adding the rigidity matrix adjustment coefficient of the current moment and the adjustment amplitude. if the relative rigidity degree estimation coefficient prediction value of the next moment of a monitoring position is less than the relative rigidity degree estimation coefficient of the current moment, the rigidity matrix adjustment coefficient of the next moment is obtained by subtracting the adjustment amplitude from the rigidity matrix adjustment coefficient of the current moment.

[0011] The embodiment of the present application has at least the following beneficial effects: the input current, loop current, input voltage, loop voltage and motor speed of the joint motor of each monitoring position at each moment are collected by the sensor, and then the relative load degree characteristic value at the moment is obtained by analyzing the input current, loop current, input voltage and loop voltage of a monitoring position at a moment, and then the relative rigidity degree estimation coefficient at the moment is obtained by analyzing the relative load degree characteristic value of the adjacent previous moment and the motor speed, and then the rigidity estimation credibility at the moment is obtained according to the relative rigidity degree estimation coefficient of each adjacent moment and the motor speed of a monitoring position at a moment, and the final decomposition weight of each moment of the monitoring position is obtained; the trend item and the periodic item are obtained by using the final decomposition weight of each moment of the monitoring position to decompose the relative rigidity degree estimation coefficient of each moment of the monitoring position based on the STL algorithm, so as to achieve the purpose of predicting the relative rigidity degree estimation coefficient prediction value of the next moment, and finally the rigidity matrix adjustment coefficient of the next moment is obtained based on the relative rigidity degree estimation coefficient prediction value of the next moment of a monitoring position, the relative rigidity degree estimation coefficient of each historical moment and the relative rigidity degree estimation coefficient prediction value, and the mechanical arm is controlled based on the rigidity matrix adjustment coefficient of the next moment of each monitoring position and the variable impedance control strategy, so as to effectively control the mechanical arm, thereby facilitating the adjustment of the rigidity and flexibility of each monitoring position, and making the application more stable and reliable. BRIEF DESCRIPTION OF DRAWINGS

[0012] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, and the advantages thereof, below will briefly introduce the drawings required by the embodiments or prior art description. Obviously, the drawings in the following description are only some embodiments of the present application, and for those skilled in the art, other drawings can also be obtained from these drawings without creative effort.

[0013] Figure 1 A method flow chart of a mechanical arm compliant control method based on sensing monitoring provided by the embodiments of the present application. DETAILED DESCRIPTION

[0014] In order to further illustrate the technical means and effects adopted by the present application to achieve the predetermined invention purpose, the specific embodiments, structures, features and effects of a mechanical arm compliant control method based on sensing monitoring according to the present application will be described in detail below in combination with the drawings and preferred embodiments. In the following description, different "one embodiment" or "another embodiment" do not necessarily refer to the same embodiment. In addition, the specific features, structures or characteristics in one or more embodiments can be combined in any suitable form.

[0015] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which the present application belongs.

[0016] The specific scheme of the mechanical arm compliant control method based on sensing monitoring provided by the present application will be specifically described below in combination with the drawings.

[0017] Embodiment: The main application scenario of the present application is to collect the data of the monitoring position of the mechanical arm by using the sensor, and then analyze it, obtain the stiffness matrix adjustment coefficient at the future time according to the analysis result, and then perform compliant control on the mechanical arm.

[0018] Please refer to Figure 1 , which shows the method flow chart of the mechanical arm compliant control method based on sensing monitoring provided by the embodiments of the present application, which comprises the following steps: Step S1, collecting the input current, loop current, input voltage, loop voltage and motor speed of each joint motor of each monitoring position of the mechanical arm at each time; obtaining the relative load degree characteristic value at the time according to the input current, loop current, input voltage and loop voltage of one monitoring position at one time.

[0019] The present application mainly effectively controls the compliance of the mechanical arm in the application scenario to improve the stability of the operation. In order to achieve this goal, the joints and key positions of the mechanical arm need to be obtained as monitoring positions to reflect the motion change of the mechanical arm.

[0020] Further, sensors are arranged at each monitoring position to collect input current, loop current, input voltage, loop voltage and motor speed of the joint motor at each moment of each monitoring position in real time, and these data are standardized and de-dimensioned for subsequent analysis.

[0021] In the power equipment maintenance scene, the robot arm is mainly used for switch control and other operations, and the output of the joint motor is usually realized by variable impedance control (VIC) to achieve rigid and flexible adjustment. However, in different operating environments, operation and maintenance objects, and real-time running states of the robot arm, a fixed or preset stiffness coefficient matrix often cannot meet the requirements. If there is no adaptive adjustment mechanism, the robot arm may produce excessive impact in hard equipment operation or appear weak in flexible equipment operation. Therefore, it is necessary to adjust the stiffness coefficient matrix according to the dynamic characteristics of the working environment and the running state, so as to realize efficient, safe and compliant control of the robot arm in complex tasks.

[0022] Therefore, first, the relative load degree characteristic value at any monitoring position at any moment is obtained by using the collected data. Specifically, the ratio of the loop current at a moment of a monitoring position to the input current at the moment is denoted as a first ratio; the ratio of the input voltage at the moment of the monitoring position to the loop voltage at the moment is denoted as a second ratio; and the relative load degree characteristic value at the moment of the monitoring position is obtained by multiplying the first ratio and the second ratio.

[0023] The relative load degree characteristic value calculation model is specifically as follows: , wherein H represents the relative load degree characteristic value at a moment of a monitoring position; represents the measured value of the current in the motor loop at the moment of the monitoring position, that is, the loop current; represents the input value of the motor current at the moment of the monitoring position, that is, the input current; represents the input value of the motor voltage at the moment of the monitoring position; represents the measured value of the voltage in the motor loop at the moment of the monitoring position; is the first ratio, which represents the ratio of the measured value of the current in the motor loop at the moment of the monitoring position to the input value of the motor current at the moment, and the greater the value, the greater the load degree at the moment at the position, but it needs to be analyzed in combination with the input voltage at the moment; is the second ratio, which represents the ratio of the measured value of the voltage in the motor loop at the moment of the monitoring position to the input value of the motor voltage at the moment, so as to obtain the voltage drop ratio at the moment, and the greater the voltage drop degree at the moment, the greater the load degree of the motor at the moment.

[0024] By the above, the relative load degree eigenvalue at any monitoring position and at any moment can be obtained, and then at a monitoring position, a relative load degree eigenvalue sequence can be obtained in time sequence during the working process.

[0025] In step S2, the relative rigidity degree estimation coefficient at a monitoring position and at a moment is obtained according to the relative load degree eigenvalues at the moment and at the previous moment and the motor speed at the moment, and the rigidity estimation reliability at the moment is obtained according to the relative rigidity degree estimation coefficients at each adjacent moment and the motor speed at the moment.

[0026] After obtaining the relative load degree eigenvalue sequence at any moment by the above method, if the load eigenvalue at a monitoring position at adjacent moments increases significantly, and the motor speed at the position changes little, it indicates that the manipulator bears a large load at the position, but the speed has not responded significantly, reflecting that the rigidity of the manipulator is high. Based on this, the relative rigidity degree coefficient of each moment of each monitoring position is estimated, which can dynamically evaluate the rigidity characteristics of the manipulator under different working states, and provide reliable reference for the adjustment of the compliant control strategy and the working safety monitoring.

[0027] Thus, the relative rigidity degree estimation coefficient at a monitoring position and at a moment is obtained according to the relative load degree eigenvalues at the moment and at the previous moment and the motor speed at the moment.

[0028] Specifically, the difference between the relative load degree eigenvalues at a monitoring position and at a moment and at the previous moment is negatively correlated mapped by using an exponential function with a natural constant as the base to obtain a first mapping value at the moment; the difference between the motor speeds at the monitoring position at the moment and at the previous moment is negatively correlated mapped by using an exponential function with a natural constant as the base to obtain a second mapping value; and the relative rigidity degree estimation coefficient at the monitoring position at the moment is obtained by comparing the first mapping value and the second mapping value.

[0029] The specific calculation model of the relative rigidity degree estimation coefficient is: , Wherein, represents the relative rigidity degree estimation coefficient at a monitoring position t moment; e represents a natural constant, and the role of mapping with e is to prevent each part from being 0; and respectively represent the relative load degree eigenvalues at the monitoring position at moment t and at moment t-1; 、 respectively represent the motor speeds at the monitoring position at moment t and at moment t-1; For the first mapping value, it represents the increase factor of the relative load degree characteristic value at the monitoring position at time t and time t-1. The greater the value, the greater the load increase degree. It represents the increase factor of the motor speed at the monitoring position at time t and time t-1. The greater the value, the greater the motor speed increase. It represents the ratio of the increase factor of the relative load degree characteristic value and the increase factor of the motor speed at the monitoring position at time t and time t-1. The greater the value, the greater the load characteristic increase and the less the speed increase, indicating that the rigidity degree at this time is relatively greater.

[0030] Through the above model calculation, the relative rigidity degree estimation coefficient of the manipulator at different monitoring positions and different times can be obtained. However, due to the inertia effect in the movement of the manipulator and the dynamic coupling between the joints, the change of the single joint load may be affected by the movement of other joints, resulting in fluctuations or deviations in the relative rigidity degree estimation coefficient. To improve the stability of control, different decomposition weights can be assigned to the relative rigidity degree estimation coefficient estimation value at any monitoring position and any time, and the STL (Seasonal-Trend decomposition using Loess) method can be used to decompose the sequence of relative rigidity degree estimation coefficients corresponding to each time at a monitoring position, and extract more accurate trend items and periodic items. Subsequently, a prediction algorithm can be applied to the decomposed data to predict the rigidity degree in real time, and the rigidity coefficient matrix in the VIC (Variable Impedance Control) can be dynamically adjusted according to the prediction results, thereby ensuring the flexibility and control stability of the manipulator in complex working environments.

[0031] In a local range, if the relative rigidity degree estimation coefficient G at a certain time compared to the neighborhood has a significant mutation, while the speed and the neighborhood difference is small and the speed itself is low, it indicates that the response of the manipulator to the load change is lagging, and it may be affected by external disturbance or joint coupling, and the rigidity degree estimation value at this time may not truly reflect the actual rigidity change of the manipulator, and should be given a lower weight in decomposition and weight distribution to avoid misjudgment of rigidity change.

[0032] Thus, according to the relative rigidity degree estimation coefficient and the motor speed of each neighborhood at a monitoring position and a time, the rigidity estimation credibility at the time can be obtained, reflecting the possible degree of real rigidity change.

[0033] Specifically, a preset number of time points before and after a time point are taken as the neighborhood time points of the time point; a time interval between a neighborhood time point and the time point is negatively mapped by an exponential function with a natural constant as the base to obtain a time distance weight of the neighborhood time point of the time point; for a monitoring position, a first average difference is obtained by weighting and averaging the absolute value of the difference between the relative rigidity degree estimation coefficient of a time point and the relative rigidity degree estimation coefficients of the neighborhood time points of the time point, using the time distance weights of the neighborhood time points of the time point; a second average difference is obtained by weighting and averaging the absolute value of the difference between the motor speed of the time point and the motor speeds of the neighborhood time points of the time point, using the time distance weights of the neighborhood time points of the time point; a rigidity estimation credibility of the monitoring position at the time point is obtained by negatively mapping the ratio of the first average difference to the product of the second average difference and the motor speed of the time point, using the exponential function with the natural constant as the base.

[0034] The calculation model of the rigidity estimation credibility is specifically: , wherein, represents the rigidity estimation credibility of a monitoring position at a time point t, and represents the possibility that the relative rigidity degree estimation coefficient at any time point of any monitoring position is possibly true rigidity change; exp represents an exponential function with a natural constant e as the base; n represents the number of neighborhood time points of t, preferably, the reference value of the present application is 8, that is, a preset number of time points before t and a preset number of time points after t are taken, a total of 8, and the preset number is 4; when the neighborhood time points are obtained, if the number of time points before or after a time point is insufficient, all of them are obtained. represents the relative rigidity degree estimation coefficient of the monitoring position at the time point t, represents the relative rigidity degree estimation coefficient of the s-th neighborhood time point of t; represents the time interval between the s-th neighborhood time point of t and t, the smaller the value is, that is, the larger the value is, represents the time distance weight of the s-th neighborhood time point; , respectively represent the motor speed value of the monitoring position at the time point t and the motor speed value of the s-th neighborhood time point of the time point t; is the first average difference, representing the weighted mean of the absolute value sum of the difference between the relative rigidity degree estimation coefficient of the monitoring position at the time point t and the relative rigidity degree estimation coefficients of each neighborhood time point of the time point, that is, representing the difference feature of the relative rigidity degree estimation coefficient of the time point and the relative rigidity degree estimation coefficients of each neighborhood time point in the neighborhood of the time point; The second average difference represents the weighted average of the absolute values ​​of the differences between the motor speed at time t at the monitoring location and the motor speed at each neighboring time within that neighborhood. In other words, it represents the difference characteristic between the motor speed at time t and the motor speed at each neighboring time. The smaller this value, and the smaller the motor speed at that time, the better. The smaller the value, the greater the difference between the rigidity and the neighborhood, i.e. The larger the value, the greater the probability that it is caused by random disturbances, and the smaller the actual probability. Therefore, its weight will be smaller in the subsequent decomposition process, and vice versa.

[0035] Therefore, the reliability of the rigid estimate for each monitoring location at each time point can be obtained.

[0036] Step S3: Obtain the final decomposition weights for each moment based on the reliability of the rigidity estimate at each moment of a monitoring location and the initial decomposition weights of the STL algorithm; combine the final decomposition weights for each moment of a monitoring location with the relative rigidity estimation coefficients at each moment of the monitoring location based on the STL algorithm to obtain the trend term and the periodic term.

[0037] The above steps can yield the rigid estimate reliability for each monitoring location at each time point. Then, by combining the local weighted decomposition in STL, the weights assigned to the local data are corrected, resulting in a more effective final decomposition weight.

[0038] Therefore, the final decomposition weights at each time point are obtained based on the rigidity estimation reliability of a monitoring location at each time point and the initial decomposition weights of the STL algorithm.

[0039] Specifically, the reliability of the rigidity estimate at a monitoring location at a given time is compared with the sum of the reliability of the rigidity estimates at all times within the local range when the relative rigidity estimate coefficient at that time is used for local weighted decomposition using the STL algorithm. This sum is then multiplied by the initial decomposition weights when the relative rigidity estimate coefficient at that time is used for local weighted decomposition using the STL algorithm to obtain the weight factor at that time. Finally, the weight factor at a monitoring location at that time is compared with the sum of the weight factors at all times within the local range when the relative rigidity estimate coefficient at that time is used for local weighted decomposition using the STL algorithm to obtain the final decomposition weight at that monitoring location at that time.

[0040] The specific calculation model for the final decomposition weights at a monitoring location at a given time is as follows: , In the formula The relative stiffness estimation coefficient at time t at a monitoring location represents the final decomposition weight when the STL algorithm is used for local weighted decomposition. represents the initial weight of the local weighted decomposition of the rigidity degree estimation coefficient at time t at a monitoring position (i.e., the weight given by the STL algorithm); represents the rigidity estimation reliability at time t at a monitoring position, represents the number of time points contained in the local range in the decomposition of the relative rigidity degree estimation coefficient at time t by the STL algorithm, represents the rigidity estimation reliability of the a-th time point in the local range corresponding to time t, represents the initial decomposition weight of the a-th time point in the local range corresponding to time t; is the ratio of the rigidity estimation reliability at time t at the monitoring position to the sum of the rigidity estimation reliabilities of all time points in the local range corresponding to time t, represents the possible degree of authenticity of the relative rigidity degree estimation coefficient at time t at the monitoring position, and the larger the value, the higher the final decomposition weight. is the weight factor at time t at the monitoring position.

[0041] Through the above, the final decomposition weight of the relative rigidity degree estimation coefficient value at each time at a monitoring position in the local weighted decomposition can be obtained, and then the relative rigidity degree estimation coefficient of all time points at the monitoring position is decomposed using the STL algorithm combined with the final decomposition weight of each time point to obtain the trend item and the periodic item data.

[0042] Step S4, predicting the relative rigidity degree estimation coefficient prediction value at the next time based on the trend item and the periodic item corresponding to a monitoring position; obtaining the rigidity matrix adjustment coefficient at the next time based on the relative rigidity degree estimation coefficient prediction value at the next time at a monitoring position, the relative rigidity degree estimation coefficient at each historical time, and the relative rigidity degree estimation coefficient prediction value; and controlling the robot arm based on the rigidity matrix adjustment coefficient at the next time at each monitoring position combined with the variable impedance control strategy.

[0043] The trend item and the periodic item corresponding to a monitoring position are obtained above, and further analysis of the two items obtains the relative rigidity degree estimation coefficient prediction value at the next time. Specifically, the trend prediction value and the periodic prediction value at the next time are obtained by predicting the trend item and the periodic item corresponding to a monitoring position using the GRU algorithm; the relative rigidity degree estimation coefficient prediction value at the next time is obtained by adding the trend prediction value and the periodic prediction value at the next time. It should be noted that the relative rigidity degree estimation coefficient prediction value obtained by prediction and the relative rigidity degree estimation coefficient obtained by calculation are the same physical quantity.

[0044] Further, by comparing the difference between the relative rigidity degree estimation coefficient prediction value at the next time and the relative rigidity degree estimation coefficient at the current time, the adjustment coefficient of the rigidity matrix is adaptively adjusted, so as to realize the adaptive control of the rigidity and flexibility of the robot arm.

[0045] In the current application scenario of the mechanical arm, if the relative rigidity degree estimation coefficient of the next moment predicted by the feature sequence is increased compared with the current moment, it usually means that the environment or the operating object shows higher anti-deformation ability, that is, it becomes "hard". In order to ensure the stability and compliance of the robot end in the process of interacting with the environment, the rigidity matrix adjustment coefficient should be increased accordingly at this time, so that the robot outputs greater resistance to match the high rigidity of the environment, thereby avoiding position errors or control deviations caused by excessive flexibility. Conversely, if the predicted relative rigidity degree estimation coefficient is reduced, it means that the environment becomes "soft", and the rigidity matrix adjustment coefficient needs to be reduced to avoid the system being too rigid to produce shock or unnecessary force. For example, in the task of polishing the workpiece by the robot, when it is predicted that the surface material of the workpiece gradually becomes hard, the controller will increase the rigidity adjustment coefficient, so that the polishing head maintains a stable contact force; when the edge area of the workpiece is soft, the rigidity adjustment coefficient is reduced, so that the end can conform to the surface deformation without damaging the workpiece.

[0046] Therefore, the rigidity matrix adjustment coefficient of the next moment is predicted based on the relative rigidity degree estimation coefficient prediction value of the next moment of a monitoring position, the relative rigidity degree estimation coefficients of each historical moment and the relative rigidity degree estimation coefficient prediction value of the next moment.

[0047] Specifically, if the relative rigidity degree estimation coefficient prediction value of the next moment of a monitoring position is greater than or equal to the relative rigidity degree estimation coefficient of the current moment; the time interval between a historical moment and the current moment is negatively mapped to obtain the time distance weight of the historical moment by using the exponential function with the natural constant as the base; the absolute value of the difference between the relative rigidity degree estimation coefficient prediction value of each historical moment and the relative rigidity degree estimation coefficient of each historical moment is weighted and averaged to obtain a third average difference, and the third mapping value is obtained by negatively mapping the third average difference by using the exponential function with the natural constant as the base; the relative change degree is obtained by subtracting the ratio of the relative rigidity degree estimation coefficient prediction value of the next moment and the relative rigidity degree estimation coefficient of the current moment from the first preset value and taking the absolute value; the adjustment amplitude is obtained by multiplying the rigidity matrix adjustment coefficient of the current moment, the relative change degree and the third mapping value; the rigidity matrix adjustment coefficient of the next moment is obtained by adding the rigidity matrix adjustment coefficient of the current moment and the adjustment amplitude. If the relative rigidity degree estimation coefficient prediction value of the next moment of a monitoring position is less than the relative rigidity degree estimation coefficient of the current moment, the rigidity matrix adjustment coefficient of the next moment is obtained by subtracting the adjustment amplitude from the rigidity matrix adjustment coefficient of the current moment.

[0048] The calculation model of the rigidity matrix adjustment coefficient of the next moment is specifically: , , wherein, represents a stiffness matrix adjustment coefficient of the next time at the monitoring position, represents a stiffness matrix adjustment coefficient of the previous time at the next time at the monitoring position; exp() represents an exponential function with a natural constant as a base; represents the number of historical times before the next time, including the current time; , respectively represent a predicted value of the relative rigidity degree estimation coefficient of the next time at the current time at the monitoring position and a value of the relative rigidity degree estimation coefficient at the current time at the monitoring position (without residual, that is, the sum of the periodic term and the trend term); is a relative change degree, represents the absolute value of the difference between the ratio of the predicted value of the relative rigidity degree estimation coefficient of the next time at the current time at the monitoring position and the value of the relative rigidity degree estimation coefficient at the current time at the monitoring position and 1, indicating the relative change degree between the two times, the greater the value, the greater the adjustment degree of the stiffness matrix adjustment coefficient of the next time; , respectively represent a predicted value of the relative rigidity degree estimation coefficient of the n-s historical time at the monitoring position and a relative rigidity degree estimation coefficient; e represents a natural constant; represents the time interval between the n historical time (current time) and the n-s historical time, because there is a certain error between the predicted value and the actual value, by calculating the error between the historical data and the predicted value along with the change of time, the error of the predicted value at the next time at the current time is measured, which is used as a proportional coefficient of the rigidity degree adjustment size, to avoid the influence of the prediction error on the reckless or large adjustment of the stiffness matrix coefficient, so as to affect the stability of the mechanical arm work; is a third average difference, indicating the weighted mean of the current prediction error obtained by changing with time, if the value is smaller, then the current prediction error is smaller, the proportion of the stiffness matrix adjustment is higher, otherwise the proportion of the adjustment is smaller, is a third mapping value; It should be noted that the relative rigidity degree estimation coefficient of each historical moment, that is, the actual value of the relative rigidity degree estimation coefficient of each historical moment needs to be decomposed by STL, and then the trend item and the periodic item corresponding to each historical moment are added to obtain the relative rigidity degree estimation coefficient of each historical moment in the formula, because the relative rigidity degree estimation coefficient prediction value of the next moment is obtained by adding the trend prediction value and the periodic prediction value of the next moment, so as to ensure the uniformity of the comparison dimension, reflect the difference between the prediction and the actual value, discard the residual term when predicting, in order to avoid the volatility of adjustment and improve the smoothness and stability of the mechanical arm work.

[0049] After obtaining the stiffness matrix adjustment coefficient of the next moment of each monitoring position at the current moment, the coefficient can be applied to the variable impedance control (VIC) strategy of each joint of the mechanical arm in real time to dynamically adjust the joint stiffness, so that the mechanical arm can adaptively change the rigidity and flexibility when contacting different work objects or being disturbed by the external disturbance. At the same time, the sensor continuously monitors the current, voltage and joint speed signals, updates the rigidity degree estimation coefficient in real time, and combines the prediction model to continuously adjust the stiffness matrix of the next moment, so as to realize the closed-loop control of the rigidity and flexibility of the mechanical arm, and ensure that it has sufficient flexibility to protect the work object and maintains the operation precision and motion stability in the complex working environment.

[0050] It should be noted that the above-mentioned embodiment sequence of the present application is only for description, and does not represent the advantages and disadvantages of the embodiments. And the above describes a specific embodiment of the present application. In addition, the processes depicted in the drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multi-task processing and parallel processing are also possible or may be advantageous.

[0051] Each embodiment in the present application is described in a progressive manner, and the same or similar parts between each embodiment can be referred to each other, and each embodiment mainly describes the difference from other embodiments.

[0052] The above only describes the preferred embodiments of the present application, and does not limit the present application. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application shall be included in the protection scope of the present application.

Claims

1. A method for compliant control of a robotic arm based on sensory monitoring, the method comprising: The method comprises: Collecting input current, loop current, input voltage, loop voltage and motor speed of each joint motor of the mechanical arm at each monitoring position at each time; obtaining a relative load degree characteristic value at a time according to input current, loop current, input voltage and loop voltage at the time at one monitoring position; Obtaining a relative rigidity degree estimation coefficient at a time according to the relative load degree characteristic value at the time and the time preceding the time and the motor speed at the time and the time preceding the time at one monitoring position; obtaining rigidity estimation credibility at the time according to the relative rigidity degree estimation coefficients at each adjacent time and the motor speed at each adjacent time at one monitoring position; Obtaining a final decomposition weight at each time according to the rigidity estimation credibility at each time at one monitoring position and the initial decomposition weight of the STL algorithm; decomposing the relative rigidity degree estimation coefficient at each time at one monitoring position based on the STL algorithm to obtain a trend item and a periodic item according to the final decomposition weight at each time at one monitoring position; Predicting a relative rigidity degree estimation coefficient prediction value at the next time based on the trend item and the periodic item corresponding to one monitoring position; obtaining a rigidity matrix adjustment coefficient at the next time based on the relative rigidity degree estimation coefficient prediction value at the next time, the relative rigidity degree estimation coefficients at each historical time and the relative rigidity degree estimation coefficient prediction value at one monitoring position; controlling the mechanical arm based on the rigidity matrix adjustment coefficient at the next time at each monitoring position and the variable impedance control strategy.

2. The compliant control method for a robotic arm based on sensory monitoring according to claim 1, wherein, The relative load degree characteristic value at a time is obtained according to input current, loop current, input voltage and loop voltage at the time at one monitoring position, comprising: The ratio of the loop current at a time at one monitoring position to the input current at the time is denoted as a first ratio; the ratio of the input voltage at the time at one monitoring position to the loop voltage at the time is denoted as a second ratio; the relative load degree characteristic value at the time at one monitoring position is obtained by multiplying the first ratio and the second ratio.

3. The compliant control method for a robotic arm based on sensory monitoring according to claim 1, wherein, The relative rigidity degree estimation coefficient at a time is obtained according to the relative load degree characteristic value at the time and the time preceding the time and the motor speed at the time and the time preceding the time at one monitoring position, comprising: The difference between the relative load degree characteristic value at a time and the time preceding the time at one monitoring position is negatively correlated and mapped by using an exponential function with a natural constant as the base to obtain a first mapping value at the time; the difference between the motor speed at the time and the time preceding the time at one monitoring position is negatively correlated and mapped by using an exponential function with a natural constant as the base to obtain a second mapping value; the relative rigidity degree estimation coefficient at the time at one monitoring position is obtained by comparing the first mapping value and the second mapping value.

4. The compliant control method for a robotic arm based on sensor monitoring according to claim 1, characterized in that, The rigidity estimation credibility at a time is obtained according to the relative rigidity degree estimation coefficients at each adjacent time and the motor speed at each adjacent time at one monitoring position, comprising: Starting from a given moment, a predetermined number of moments are taken before and after that moment, and these moments are recorded as the neighborhood moments of that moment. A negative correlation mapping is performed on the time interval between a neighborhood moment and the given moment using an exponential function with a base of the natural constant to obtain the time distance weight of that neighborhood moment. For a monitoring location, the absolute value of the difference between the relative rigidity estimation coefficients of that moment and its neighboring moments is weighted and averaged using the time distance weights of the neighborhood moments of that monitoring location to obtain the first average difference. The absolute value of the difference between the motor speeds of that moment and its neighboring moments is weighted and averaged using the time distance weights of the neighborhood moments of that monitoring location to obtain the second average difference. A negative correlation mapping is performed on the ratio of the first average difference to the product of the second average difference and the motor speed at that moment using an exponential function with a base of the natural constant to obtain the rigidity estimation reliability of that monitoring location at that moment.

5. The compliant control method of a manipulator based on sensing monitoring according to claim 1, characterized in that, The step of obtaining the final decomposition weights for each time step based on the rigid estimation reliability of a monitoring location at each time step and the initial decomposition weights of the STL algorithm includes: The reliability of the rigidity estimate at a monitoring location at a given time is compared with the sum of the reliability of the rigidity estimates at all times within the local range when the relative rigidity estimate coefficient at that time is used for local weighted decomposition using the STL algorithm. This sum is then multiplied by the initial decomposition weights when the relative rigidity estimate coefficient at that time is used for local weighted decomposition using the STL algorithm to obtain the weight factor at that time. The final decomposition weight at that monitoring location at that time is then compared with the sum of the weight factors at all times within the local range when the relative rigidity estimate coefficient at that time is used for local weighted decomposition using the STL algorithm to obtain the final decomposition weight at that monitoring location at that time.

6. The compliant control method of a manipulator based on sensing monitoring according to claim 1, characterized in that, The predicted value of the relative stiffness estimation coefficient for the next moment based on the trend term and period term corresponding to a monitoring location includes: The GRU algorithm is used to predict the trend and periodic terms corresponding to a monitoring location to obtain the predicted trend and periodic values ​​for the next moment. The predicted trend and periodic values ​​for the next moment are added together to obtain the predicted relative stiffness estimation coefficient for the next moment.

7. The compliant control method of a manipulator based on sensing monitoring according to claim 1, characterized in that, The stiffness matrix adjustment coefficient for the next moment is obtained based on the predicted value of the relative stiffness estimation coefficient at a monitoring location at the next moment, the relative stiffness estimation coefficients at each historical moment, and the predicted value of the relative stiffness estimation coefficient. This includes: If the predicted value of the relative rigidity degree estimation coefficient of the next moment of a monitoring position is greater than or equal to the relative rigidity degree estimation coefficient of the current moment, a time distance weight of a historical moment is obtained by using an exponential function with a natural constant as a base to negatively correlate the time interval between the historical moment and the current moment; a third average difference is obtained by weighting and averaging the absolute value of the difference between the predicted value of the relative rigidity degree estimation coefficient of each historical moment and the relative rigidity degree estimation coefficient based on the time distance weight of each historical moment, and the third average difference is negatively correlated by using an exponential function with a natural constant as a base to obtain a third mapping value; a relative change degree is obtained by using a first preset value to subtract the ratio of the predicted value of the relative rigidity degree estimation coefficient of the next moment to the relative rigidity degree estimation coefficient of the current moment and taking the absolute value; an adjustment amplitude is obtained by multiplying the rigidity matrix adjustment coefficient of the current moment, the relative change degree and the third mapping value; and the rigidity matrix adjustment coefficient of the next moment is obtained by adding the rigidity matrix adjustment coefficient of the current moment and the adjustment amplitude. If the predicted value of the relative rigidity degree estimation coefficient of the next moment of a monitoring position is less than the relative rigidity degree estimation coefficient of the current moment, the rigidity matrix adjustment coefficient of the next moment is obtained by subtracting the adjustment amplitude from the rigidity matrix adjustment coefficient of the current moment.

Images