CN109623819A - The acquisition methods and device of a kind of robot harmonic drive joint actual torque value - Google Patents
The acquisition methods and device of a kind of robot harmonic drive joint actual torque value Download PDFInfo
- Publication number
- CN109623819A CN109623819A CN201811587023.1A CN201811587023A CN109623819A CN 109623819 A CN109623819 A CN 109623819A CN 201811587023 A CN201811587023 A CN 201811587023A CN 109623819 A CN109623819 A CN 109623819A
- Authority
- CN
- China
- Prior art keywords
- harmonic drive
- drive joint
- machine learning
- learning model
- value
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Classifications
-
- B—PERFORMING OPERATIONS; TRANSPORTING
- B25—HAND TOOLS; PORTABLE POWER-DRIVEN TOOLS; MANIPULATORS
- B25J—MANIPULATORS; CHAMBERS PROVIDED WITH MANIPULATION DEVICES
- B25J9/00—Programme-controlled manipulators
- B25J9/16—Programme controls
-
- B—PERFORMING OPERATIONS; TRANSPORTING
- B25—HAND TOOLS; PORTABLE POWER-DRIVEN TOOLS; MANIPULATORS
- B25J—MANIPULATORS; CHAMBERS PROVIDED WITH MANIPULATION DEVICES
- B25J19/00—Accessories fitted to manipulators, e.g. for monitoring, for viewing; Safety devices combined with or specially adapted for use in connection with manipulators
-
- B—PERFORMING OPERATIONS; TRANSPORTING
- B25—HAND TOOLS; PORTABLE POWER-DRIVEN TOOLS; MANIPULATORS
- B25J—MANIPULATORS; CHAMBERS PROVIDED WITH MANIPULATION DEVICES
- B25J9/00—Programme-controlled manipulators
- B25J9/16—Programme controls
- B25J9/1628—Programme controls characterised by the control loop
-
- B—PERFORMING OPERATIONS; TRANSPORTING
- B25—HAND TOOLS; PORTABLE POWER-DRIVEN TOOLS; MANIPULATORS
- B25J—MANIPULATORS; CHAMBERS PROVIDED WITH MANIPULATION DEVICES
- B25J9/00—Programme-controlled manipulators
- B25J9/16—Programme controls
- B25J9/1628—Programme controls characterised by the control loop
- B25J9/1633—Programme controls characterised by the control loop compliant, force, torque control, e.g. combined with position control
Abstract
The invention discloses a kind of acquisition methods of robot harmonic drive joint actual torque value, 1) method includes:, utilizes the set of harmonic drive flexibility of joint error model acquisition harmonic drive joint output torque model value;2) sample data sets, are obtained, use sample data sets training machine learning model, then according to the position signal of the motor side in harmonic drive joint, the position signal of the connecting-rod head in harmonic drive joint and current of electric data, the predicted value of harmonic drive joint output torque is obtained;3), harmonic drive joint output torque model value and predicted value mean value are filtered, obtain the corresponding actual torque value of harmonic drive joint output torque model value.The invention discloses a kind of acquisition device of robot harmonic drive joint actual torque value.Using the embodiment of the present invention, the accuracy of the robot harmonic drive joint actual torque value of acquisition can be improved.
Description
Technical field
The present invention relates to a kind of acquisition methods of moment values and devices, are more particularly to a kind of robot harmonic drive joint
The acquisition methods and device of actual torque value.
Background technique
It can be measured based on flexibility error model using current of electric, motor side location information and connecting-rod head location information
Harmonic drive joint actual torque.
But when being fitted harmonic drive joint actual torque using motor side and the location error of connecting-rod head, to harmonic wave
Driving joint flexibility carries out modeling and needs accurate parameter.Parameter selection in model directly affects Flexible Model about Ecology precision.Harmonic wave
It is driven during flexible loading, uses fitting formula and empirical equation mostly, model is caused to generate constant error.
Therefore, the accuracy that the robot harmonic drive joint actual torque value of acquisition exists in the prior art not high is asked
Topic.
Summary of the invention
Technical problem to be solved by the present invention lies in provide a kind of robot harmonic drive joint actual torque value
Acquisition methods and device, to improve the accuracy of the robot harmonic drive joint actual torque value obtained.
The present invention is to solve above-mentioned technical problem by the following technical programs:
The embodiment of the invention provides a kind of acquisition methods of robot harmonic drive joint actual torque value, the methods
Include:
1), using harmonic drive flexibility of joint error model, according to the position signal of the motor side in harmonic drive joint, humorous
The position signal of the connecting-rod head of wave driving joint and current of electric data acquisition harmonic drive joint output torque model value
Set;
2) sample data sets, are obtained, using the sample data sets training machine learning model, then according to
The position signal and current of electric number of the position signal of the motor side in harmonic drive joint, the connecting-rod head in harmonic drive joint
According to obtaining the predicted value of harmonic drive joint output torque;
3), the harmonic drive joint output torque model value and the predicted value mean value are filtered, obtained
To the corresponding actual torque value of the harmonic drive joint output torque model value.
Optionally, the step 1), comprising:
Using formula,Calculate the harmonic wave in harmonic drive joint
It is driven deformation angle, wherein
ΔθfFor the harmonic drive deformation angle in harmonic drive joint;qdFor the position letter of the connecting-rod head in harmonic drive joint
Number;qmFor the position signal of motor side;Sgn () is sign function;KωFor the rigidity system of the wave producer in harmonic drive joint
Number;IiFor current of electric;cωFor the lag coefficient of the wave producer in harmonic drive joint;L is humorous in harmonic drive joint
Wave retarder reduction ratio;E is the nature truth of a matter;| | it is mod function;θerrFor the flexibility between the defeated end of motor and connecting rod output end
Error;I is the quantity of current of electric parameter;
Using formula,Calculate Flexible Transmission torsion angle, wherein
Δ θ is Flexible Transmission torsion angle;ΔθωFor the torsion angle of the input of wave producer;It exports and turns round for wave producer
Corner;The transmission ratio of l harmonic drive joint speed changer;
Utilize formula, τf=a1Δθ+a2Δθ2+a3Δθ3, harmonic drive flexibility of joint output torque is calculated, and then obtain
The set of harmonic drive joint output torque model value, wherein
τfFor harmonic drive flexibility of joint output torque;a1For preset first parameter;a2For preset second parameter;a3
For preset third parameter;
Using formula,Calculate harmonic drive joint output torque model value, wherein
τmodelFor harmonic drive joint output torque model value;Tan () is tangent function;cfFor preset first constant;
KfoFor preset second constant.
Optionally, the step 2) includes:
A: using the harmonic drive joint actual output torque value as target, the sample data sets are divided into multiple sons
Set, then each subclass is inputted in each local Gaussian process machine learning model respectively and is trained, wherein described
Sample data includes: the position signal of the motor side in harmonic drive joint, the position signal of the connecting-rod head in harmonic drive joint, electricity
The actual torque value in electromechanical flow data and harmonic drive joint;
B: by the position of the position signal of the motor side in the harmonic drive joint in step 1), the connecting-rod head in harmonic drive joint
Confidence number and current of electric data are input in the local Gaussian process machine learning model obtained after the training, calculate institute
State sample data to each local Gaussian process machine learning model distance, and obtain apart from the corresponding local Gaussian of maximum value
Process machine learning model;
C: obtaining the location factor of harmonic drive joint output torque, judges each distance in the location factor
Whether the maximum value in parameter is less than preset threshold;
D: if so, updating local Gaussian process machine learning model, and by updated local Gaussian process machine learning
The central point of model is as new match point;It returns and executes the step B, until the judging result of the step C is no;
E: if it is not, the harmonic drive joint output torque exported according to each local Gaussian process machine learning model
Predicted value weighted average, calculate the corresponding predicted value mean value of predicted value of harmonic drive joint output torque.
Optionally, the step A includes:
A1: the sample data sets are divided into multiple subclass, each subclass is then inputted into each part respectively
It is trained in Gaussian process machine learning model, wherein the subclass is according to the sample data in the subclass
What location factor determined at a distance from the central point of the local Gaussian process machine learning model;
A2: being directed to each local Gaussian process machine learning model, using formula,Meter
Calculate predicted value mean value, wherein
f(x*) it is predicted value mean value;Y is the predicted value of the local Gaussian process machine learning model of last iteration;α is
Predicted vector;x*For sample data;For the variance of Gaussian noise;I is the current matrix of training sample;()TFor transposed matrix;
K is the covariance matrix when previous iteration;
A3: utilizing formula,Calculate the initial prediction pair when previous iteration
The covariance answered, wherein
V(x*) it is the corresponding covariance of initial prediction when previous iteration;K () is that covariance calculates function;k*For institute
State covariance of the local Gaussian process machine learning model in last iteration;K is covariance matrix;x*For sample data;
For the variance of Gaussian noise;I is the current matrix of training sample;()TFor transposed matrix;
A4: judge the difference between the covariance and preset covariance, and prediction mean value and actual torque value
Within a preset range whether difference;
A5: if so, using the local Gaussian process machine learning model as the local Gaussian process engineering after training
Practise model;
A6: it if it is not, updating the parameter of local Gaussian process machine learning model, and returns and executes A2 step, until obtaining
Local Gaussian process machine learning model after training.
Optionally, the location factor of harmonic drive joint output torque is obtained, comprising:
Using formula,Calculate harmonic drive joint output torque position because
Son, wherein
wkPosition for harmonic drive joint output torque model value relative to k-th of local Gaussian process machine learning model
Set parameter;Exp () is the exponential function using the natural truth of a matter bottom of as;X is the subclass of input sample data;ckIt is high for each part
The location parameter of the central point of this process machine learning model;()TFor transposed matrix;W be and local Gaussian process machine learning
The identical diagonal matrix of model width;K is the number of local Gaussian process machine learning model;
Utilize formula, w (x, o)=[w1,...,wk], obtain the location factor of harmonic drive joint output torque, wherein
W (x, o) is the location factor of the corresponding harmonic drive joint data torque of input sample;X is input sample data
Subclass;O is the matrix that the central point of each local Gaussian process machine learning model is constituted;w1It is defeated for harmonic drive joint
Location parameter of the torque model value relative to the 1st local Gaussian process machine learning model out;wkIt is defeated for harmonic drive joint
Location parameter of the torque model value relative to k-th of local Gaussian process machine learning model out.
Optionally, the update local Gaussian process machine learning model, comprising:
Obtain new local Gaussian process machine learning model, and by the new local Gaussian process machine learning model
It is middle to be used as updated local Gaussian process machine learning model.
Optionally, the step E, comprising:
Using formula,Sample data is obtained in each local Gaussian process machine learning model
The probability of appearance, wherein
P (k | x) it is the probability that sample data x occurs in k-th of local Gaussian process machine learning model;X is sample
Data;K is the quantity of local Gaussian process machine learning model;M be obtain with the harmonic drive joint output torque mould
The number of the corresponding local Gaussian process machine learning model of offset;wkFor harmonic drive joint output torque model value relative to
The location parameter of k-th of local Gaussian process machine learning model;
Using formula,The corresponding predicted value mean value of harmonic drive joint output torque model value is calculated,
Wherein,
For the corresponding predicted value mean value of harmonic drive joint output torque model value;For k-th of local Gaussian process
The predicted value of machine learning model.
The embodiment of the invention provides a kind of acquisition device of robot harmonic drive joint actual torque value, described devices
Include:
Output module, for utilizing harmonic drive flexibility of joint error model, according to the motor side in harmonic drive joint
Position signal, harmonic drive joint connecting-rod head position signal and current of electric data acquisition harmonic drive joint power output
The set of square model value;
Module is obtained, for obtaining sample data sets, using the sample data sets training machine learning model, so
Afterwards according to the position signal of the motor side in the harmonic drive joint, the position signal and electricity of the connecting-rod head in harmonic drive joint
Electromechanical flow data obtains the predicted value of harmonic drive joint output torque;
Filter module, for being filtered to the harmonic drive joint output torque model value and the predicted value mean value
Wave processing, obtains the corresponding actual torque value of the harmonic drive joint output torque model value.
Optionally, the output module, is also used to:
Using formula,Calculate the harmonic wave in harmonic drive joint
It is driven deformation angle, wherein
ΔθfFor the harmonic drive deformation angle in harmonic drive joint;qdFor the position letter of the connecting-rod head in harmonic drive joint
Number;qmFor the position signal of motor side;Sgn () is sign function;KωFor the rigidity system of the wave producer in harmonic drive joint
Number;IiFor current of electric;cωFor the lag coefficient of the wave producer in harmonic drive joint;L is humorous in harmonic drive joint
Wave retarder reduction ratio;E is the nature truth of a matter;| | it is mod function;θerrFor the flexibility between the defeated end of motor and connecting rod output end
Error;I is the quantity of current of electric parameter;
Using formula,Calculate Flexible Transmission torsion angle, wherein
Δ θ is Flexible Transmission torsion angle;ΔθωFor the torsion angle of the input of wave producer;It exports and turns round for wave producer
Corner;The transmission ratio of l harmonic drive joint speed changer;
Utilize formula, τf=a1Δθ+a2Δθ2+a3Δθ3, harmonic drive flexibility of joint output torque is calculated, and then obtain
The set of harmonic drive joint output torque model value, wherein
τfFor harmonic drive flexibility of joint output torque;a1For preset first parameter;a2For preset second parameter;a3
For preset third parameter;
Using formula,Calculate harmonic drive joint output torque model value, wherein
τmodelFor harmonic drive joint output torque model value;Tan () is tangent function;cfFor preset first constant;
KfoFor preset second constant.
Optionally, the acquisition module, is also used to:
A: using the harmonic drive joint actual output torque value as target, the sample data sets are divided into multiple sons
Set, then each subclass is inputted in each local Gaussian process machine learning model respectively and is trained, wherein described
Sample data includes: the position signal of the motor side in harmonic drive joint, the position signal of the connecting-rod head in harmonic drive joint, electricity
The actual torque value in electromechanical flow data and harmonic drive joint;
B: by the position of the position signal of the motor side in the harmonic drive joint in step 1), the connecting-rod head in harmonic drive joint
Confidence number and current of electric data are input in the local Gaussian process machine learning model obtained after the training, calculate institute
State sample data to each local Gaussian process machine learning model distance, and obtain apart from the corresponding local Gaussian of maximum value
Process machine learning model;
C: obtaining the location factor of harmonic drive joint output torque, judges each distance in the location factor
Whether the maximum value in parameter is less than preset threshold;
D: if so, updating local Gaussian process machine learning model, and by updated local Gaussian process machine learning
The central point of model is as new match point;It returns and executes the step B, until the judging result of the step C is no;
E: if it is not, the harmonic drive joint output torque exported according to each local Gaussian process machine learning model
Predicted value weighted average, calculate the corresponding predicted value mean value of predicted value of harmonic drive joint output torque.
Optionally, the acquisition module, is also used to:
A1: the sample data sets are divided into multiple subclass, each subclass is then inputted into each part respectively
It is trained in Gaussian process machine learning model, wherein the subclass is according to the sample data in the subclass
What location factor determined at a distance from the central point of the local Gaussian process machine learning model;
A2: being directed to each local Gaussian process machine learning model, using formula,Meter
Calculate predicted value mean value, wherein
f(x*) it is predicted value mean value;Y is the predicted value of the local Gaussian process machine learning model of last iteration;α is
Predicted vector;x*For sample data;For the variance of Gaussian noise;I is the current matrix of training sample;()TFor transposed matrix;
K is the covariance matrix when previous iteration;
A3: utilizing formula,Calculate the initial prediction pair when previous iteration
The covariance answered, wherein
V(x*) it is the corresponding covariance of initial prediction when previous iteration;K () is that covariance calculates function;k*For institute
State covariance of the local Gaussian process machine learning model in last iteration;K is covariance matrix;x*For sample data;
For the variance of Gaussian noise;I is the current matrix of training sample;()TFor transposed matrix;
A4: judge the difference between the covariance and preset covariance, and prediction mean value and actual torque value
Within a preset range whether difference;
A5: if so, using the local Gaussian process machine learning model as the local Gaussian process engineering after training
Practise model;
A6: it if it is not, updating the parameter of local Gaussian process machine learning model, and returns and executes A2 step, until obtaining
Local Gaussian process machine learning model after training.
Optionally, the acquisition module, is also used to:
Using formula,Calculate harmonic drive joint output torque position because
Son, wherein
wkPosition for harmonic drive joint output torque model value relative to k-th of local Gaussian process machine learning model
Set parameter;Exp () is the exponential function using the natural truth of a matter bottom of as;X is the subclass of input sample data;ckIt is high for each part
The location parameter of the central point of this process machine learning model;()TFor transposed matrix;W be and local Gaussian process machine learning
The identical diagonal matrix of model width;K is the number of local Gaussian process machine learning model;
Utilize formula, w (x, o)=[w1,...,wk], obtain the location factor of harmonic drive joint output torque, wherein
W (x, o) is the location factor of the corresponding harmonic drive joint data torque of input sample;X is input sample data
Subclass;O is the matrix that the central point of each local Gaussian process machine learning model is constituted;w1It is defeated for harmonic drive joint
Location parameter of the torque model value relative to the 1st local Gaussian process machine learning model out;wkIt is defeated for harmonic drive joint
Location parameter of the torque model value relative to k-th of local Gaussian process machine learning model out.
Optionally, the acquisition module, is also used to:
Obtain new local Gaussian process machine learning model, and by the new local Gaussian process machine learning model
It is middle to be used as updated local Gaussian process machine learning model.
Optionally, the acquisition module, is also used to:
Using formula,Sample data is obtained in each local Gaussian process machine learning model
The probability of appearance, wherein
P (k | x) it is the probability that sample data x occurs in k-th of local Gaussian process machine learning model;X is sample
Data;K is the quantity of local Gaussian process machine learning model;M be obtain with the harmonic drive joint output torque mould
The number of the corresponding local Gaussian process machine learning model of offset;wkFor harmonic drive joint output torque model value relative to
The location parameter of k-th of local Gaussian process machine learning model;
Using formula,The corresponding predicted value mean value of harmonic drive joint output torque model value is calculated,
Wherein,
For the corresponding predicted value mean value of harmonic drive joint output torque model value;For k-th of local Gaussian process
The predicted value of machine learning model.
The present invention has the advantage that compared with prior art
Using the embodiment of the present invention, flexibility error model and machine learning method are merged, realizes that harmonic drive joint is practical
The identification and measurement of torque rely on the fitting parameter of flexibility error model compared with the existing technology, improve flexibility error model
Calculate the computational accuracy of flexible harmonic drive joint output torque.
Detailed description of the invention
Fig. 1 is a kind of stream of the acquisition methods of robot harmonic drive joint provided in an embodiment of the present invention actual torque value
Journey schematic diagram;
Fig. 2 is machine in a kind of acquisition methods of robot harmonic drive joint provided in an embodiment of the present invention actual torque value
The flow diagram of device study;
Fig. 3 is a kind of knot of the acquisition device of robot harmonic drive joint provided in an embodiment of the present invention actual torque value
Structure schematic diagram.
Specific embodiment
It elaborates below to the embodiment of the present invention, the present embodiment carries out under the premise of the technical scheme of the present invention
Implement, the detailed implementation method and specific operation process are given, but protection scope of the present invention is not limited to following implementation
Example.
To solve prior art problem, the embodiment of the invention provides a kind of robot harmonic drive joint actual torque values
Acquisition methods and device, first below just a kind of robot harmonic drive joint provided in an embodiment of the present invention actual torque value
Acquisition methods be introduced.
Fig. 1 is a kind of stream of the acquisition methods of robot harmonic drive joint provided in an embodiment of the present invention actual torque value
Journey schematic diagram;Fig. 2 is in a kind of acquisition methods of robot harmonic drive joint provided in an embodiment of the present invention actual torque value
The flow diagram of machine learning;As depicted in figs. 1 and 2, which comprises
S101: utilize harmonic drive flexibility of joint error model, according to the position signal of the motor side in harmonic drive joint,
The position signal and current of electric data acquisition harmonic drive joint output torque model value of the connecting-rod head in harmonic drive joint
Set.
Specifically, can use formula,Harmonic wave is calculated to pass
Diarthrodial harmonic drive deformation angle, wherein
ΔθfFor the harmonic drive deformation angle in harmonic drive joint;qdFor the position letter of the connecting-rod head in harmonic drive joint
Number;qmFor the position signal of motor side;Sgn () is sign function;KωFor the rigidity system of the wave producer in harmonic drive joint
Number;IiFor current of electric;cωFor the lag coefficient of the wave producer in harmonic drive joint;L is humorous in harmonic drive joint
Wave retarder reduction ratio;E is the nature truth of a matter;| | it is mod function;θerrFor the flexibility between the defeated end of motor and connecting rod output end
Error;I is the quantity of current of electric parameter;
Using formula,Calculate Flexible Transmission torsion angle, wherein
Δ θ is Flexible Transmission torsion angle;ΔθωFor the torsion angle of the input of wave producer;It exports and turns round for wave producer
Corner;The transmission ratio of l harmonic drive joint speed changer;
Utilize formula, τf=a1Δθ+a2Δθ2+a3Δθ3, harmonic drive flexibility of joint output torque is calculated, and then obtain
The set of harmonic drive joint output torque model value, wherein
τfFor harmonic drive flexibility of joint output torque;a1For preset first parameter;a2For preset second parameter;a3
For preset third parameter;
Using formula,Calculate harmonic drive joint output torque model value, wherein
τmodelFor harmonic drive joint output torque model value;Tan () is tangent function;cfFor preset first constant;
KfoFor preset second constant.
In practical applications, the system that flexible joint is made of retarder and motor, wherein retarder includes: flexible tooth
Wheel and wave producer, and the output axis connection of one end of retarder and joint of robot, the other end and robot of retarder close
The rotor of motor in section connects;The stator of motor and the pedestal of joint of robot connect.
S102: sample data sets are obtained, using the sample data sets training machine learning model, then according to institute
State the position signal and current of electric number of the position signal of the motor side in harmonic drive joint, the connecting-rod head in harmonic drive joint
According to obtaining the predicted value of harmonic drive joint output torque.
Specifically, the step S102 may include:
A: using the harmonic drive joint actual output torque value as target, the sample data sets are divided into multiple sons
Set, then each subclass is inputted in each local Gaussian process machine learning model respectively and is trained, wherein described
Sample data includes: the position signal of the motor side in harmonic drive joint, the position signal of the connecting-rod head in harmonic drive joint, electricity
The actual torque value in electromechanical flow data and harmonic drive joint.
In practical applications, step A may comprise steps of:
A1: the sample data sets are divided into multiple subclass, each subclass is then inputted into each part respectively
It is trained in Gaussian process machine learning model, wherein the subclass is according to the sample data in the subclass
What location factor determined at a distance from the central point of the local Gaussian process machine learning model;
In practical applications, local Gaussian process machine learning model can be with are as follows:
Wherein,
Y is the predicted value of harmonic drive joint output torque;H () is local Gaussian process machine learning model;K
(X, X) is western variance matrix;For the variance of Gaussian noise;I is the current matrix of training sample;X is in training sample
The position signal of motor side in harmonic drive joint, harmonic drive joint connecting-rod head position signal and current of electric number
According to the matrix of composition.
In addition, in order to improve the accuracy of trained model, the process of training part Gaussian process machine learning model
Three groups of sample datas can be used: being utilized respectively no-load test acquisition data, fixed load tests acquisition data, load change test is adopted
Collect data and carry out joint actual torque learning training, for example, no-load test acquisition data may include 140940 training points
Accordingly and 5560 measurement point datas;Fixed load test acquisition data may include 136220 trained point datas and 5500
Measure point data;It may include 135720 trained point datas and 5000 measurement point datas that load change test, which acquires data,.
Calculation amount when in order to reduce trained, can be corresponding by dividing the sample by input end of motor position signal in training sample
Input data section, i.e., according to the input end of motor position signal in the sample data in each control period to model center
Distance whether be greater than the set value value and divided, when the distance of input end of motor position signal to partial model center determine with
Afterwards, so that it may according to wkSample data is assigned to training in the partial model that threshold value is allowed.
A2: being directed to each local Gaussian process machine learning model, using formula,It calculates
Predicted value mean value, wherein
f(x*) it is predicted value mean value;Y is the predicted value of the local Gaussian process machine learning model of last iteration;α is
Predicted vector;X* is sample data;For the variance of Gaussian noise;I is the current matrix of training sample;()TFor transposition square
Battle array;K is the covariance matrix when previous iteration.
A3: utilizing formula,Calculate the initial prediction pair when previous iteration
The covariance answered, wherein
V(x*) it is the corresponding covariance of initial prediction when previous iteration;K () is that covariance calculates function;k*For institute
State covariance of the local Gaussian process machine learning model in last iteration;K is covariance matrix;x*For sample data;
For the variance of Gaussian noise;I is the current matrix of training sample;()TFor transposed matrix;
A4: judge the difference between the covariance and preset covariance, and prediction mean value and actual torque value
Within a preset range whether difference;
A5: if so, using the local Gaussian process machine learning model as the local Gaussian process engineering after training
Practise model;
A6: it if it is not, updating the parameter of local Gaussian process machine learning model, and returns and executes A2 step, until obtaining
Local Gaussian process machine learning model after training.
B: by the position signal of the motor side in the harmonic drive joint in step S102 step, the connecting rod in harmonic drive joint
The position signal and current of electric data at end are input in the local Gaussian process machine learning model obtained after the training,
Calculate the sample data to each local Gaussian process machine learning model distance, and obtain apart from the corresponding office of maximum value
Portion's Gaussian process machine learning model;
C: obtaining the location factor of harmonic drive joint output torque, judges each distance in the location factor
Whether the maximum value in parameter is less than preset threshold;
In practical applications, it can use formula,Harmonic drive is calculated to close
Save the location factor of output torque, wherein
wkPosition for harmonic drive joint output torque model value relative to k-th of local Gaussian process machine learning model
Set parameter;Exp () is the exponential function using the natural truth of a matter bottom of as;X is the subclass of input sample data;ckIt is high for each part
The location parameter of the central point of this process machine learning model;()TFor transposed matrix;W be and local Gaussian process machine learning
The identical diagonal matrix of model width;K is the number of local Gaussian process machine learning model;
Utilize formula, w (x, o)=[w1,...,wk], obtain the location factor of harmonic drive joint output torque, wherein
W (x, o) is the location factor of the corresponding harmonic drive joint data torque of input sample;X is input sample data
Subclass;O is the matrix that the central point of each local Gaussian process machine learning model is constituted;w1It is defeated for harmonic drive joint
Location parameter of the torque model value relative to the 1st local Gaussian process machine learning model out;wkIt is defeated for harmonic drive joint
Location parameter of the torque model value relative to k-th of local Gaussian process machine learning model out.
D: if so, updating local Gaussian process machine learning model, and by updated local Gaussian process machine learning
The central point of model is as new match point;It returns and executes the step B, until the judging result of the step C is no;
Specifically, available new local Gaussian process machine learning model, and by the new local Gaussian process
Updated local Gaussian process machine learning model is used as in machine learning model.In practical applications, new Gaussian process
Machine learning model refers to the local Gaussian process machine learning model obtained using new model parameter.
E: if it is not, the harmonic drive joint output torque exported according to each local Gaussian process machine learning model
Predicted value weighted average, calculate the corresponding predicted value mean value of predicted value of harmonic drive joint output torque.
In practical applications, it can use formula,Sample data is obtained in each local Gaussian
The probability occurred in process machine learning model, wherein
P (k | x) it is the probability that sample data x occurs in k-th of local Gaussian process machine learning model;X is sample
Data;K is the quantity of local Gaussian process machine learning model;M be obtain with the harmonic drive joint output torque mould
The number of the corresponding local Gaussian process machine learning model of offset;wkFor harmonic drive joint output torque model value relative to
The location parameter of k-th of local Gaussian process machine learning model;
Using formula,The corresponding predicted value mean value of harmonic drive joint output torque model value is calculated,
Wherein,
For the corresponding predicted value mean value of harmonic drive joint output torque model value;For k-th of local Gaussian process
The predicted value of machine learning model.
Using the above embodiment of the present invention, when the maximum value in each distance parameter in location factor is greater than preset threshold
When, local Gaussian process machine learning model can be automatically updated, sample data will enter new local Gaussian process engineering
It practises and being trained in model.This guarantees when sample data complexity, partial model number increases automatically, passes through this distribution side
Formula can also be such that covariance inverse matrix is updated.
S103: being filtered the harmonic drive joint output torque model value and the predicted value mean value,
Obtain the corresponding actual torque value of the harmonic drive joint output torque model value.
Specifically, the process of filtering processing can filter out for the moment values that discrete type is higher than setting value.
Using embodiment illustrated in fig. 1 of the present invention, flexibility error model and machine learning method are merged, realizes that harmonic drive is closed
The identification and measurement of actual torque are saved, relies on the fitting parameter of flexibility error model compared with the existing technology, improves flexible miss
Differential mode type calculates the computational accuracy of flexible harmonic drive joint output torque.
Corresponding with embodiment illustrated in fig. 1 of the present invention, the embodiment of the invention also provides a kind of robot harmonic drive passes
Save the acquisition device of actual torque value.
Fig. 3 is a kind of knot of the acquisition device of robot harmonic drive joint provided in an embodiment of the present invention actual torque value
Structure schematic diagram, as shown in figure 3, described device includes:
Output module 301, for utilizing harmonic drive flexibility of joint error model, according to the motor side in harmonic drive joint
Position signal, harmonic drive joint connecting-rod head position signal and current of electric data acquisition harmonic drive joint output
The set of moment model value;
Module 302 is obtained, for obtaining sample data sets, learns mould using the sample data sets training machine
Type, then according to the position signal of the motor side in the harmonic drive joint, the position signal of the connecting-rod head in harmonic drive joint
And current of electric data, obtain the predicted value of harmonic drive joint output torque;
Filter module 303, for the harmonic drive joint output torque model value and the predicted value mean value into
Row filtering processing, obtains the corresponding actual torque value of the harmonic drive joint output torque model value.
Using embodiment illustrated in fig. 3 of the present invention, flexibility error model and machine learning method are merged, realizes that harmonic drive is closed
The identification and measurement of actual torque are saved, relies on the fitting parameter of flexibility error model compared with the existing technology, improves flexible miss
Differential mode type calculates the computational accuracy of flexible harmonic drive joint output torque.
In a kind of specific embodiment of the embodiment of the present invention, the output module 301 is also used to:
Using formula,Calculate the harmonic wave in harmonic drive joint
It is driven deformation angle, wherein
ΔθfFor the harmonic drive deformation angle in harmonic drive joint;qdFor the position letter of the connecting-rod head in harmonic drive joint
Number;qmFor the position signal of motor side;Sgn () is sign function;KωFor the rigidity system of the wave producer in harmonic drive joint
Number;IiFor current of electric;cωFor the lag coefficient of the wave producer in harmonic drive joint;L is humorous in harmonic drive joint
Wave retarder reduction ratio;E is the nature truth of a matter;| | it is mod function;θerrFor the flexibility between the defeated end of motor and connecting rod output end
Error;I is the quantity of current of electric parameter;
Using formula,Calculate Flexible Transmission torsion angle, wherein
Δ θ is Flexible Transmission torsion angle;ΔθωFor the torsion angle of the input of wave producer;It exports and turns round for wave producer
Corner;The transmission ratio of l harmonic drive joint speed changer;
Utilize formula, τf=a1Δθ+a2Δθ2+a3Δθ3, harmonic drive flexibility of joint output torque is calculated, and then obtain
The set of harmonic drive joint output torque model value, wherein
τfFor harmonic drive flexibility of joint output torque;a1For preset first parameter;a2For preset second parameter;a3
For preset third parameter;
Using formula,Calculate harmonic drive joint output torque model value, wherein
τmodelFor harmonic drive joint output torque model value;Tan () is tangent function;cfFor preset first constant;
KfoFor preset second constant.
In a kind of specific embodiment of the embodiment of the present invention, the acquisition module 302 is also used to:
A: using the harmonic drive joint actual output torque value as target, the sample data sets are divided into multiple sons
Set, then each subclass is inputted in each local Gaussian process machine learning model respectively and is trained, wherein described
Sample data includes: the position signal of the motor side in harmonic drive joint, the position signal of the connecting-rod head in harmonic drive joint, electricity
The actual torque value in electromechanical flow data and harmonic drive joint;
B: by the position of the position signal of the motor side in the harmonic drive joint in step 1), the connecting-rod head in harmonic drive joint
Confidence number and current of electric data are input in the local Gaussian process machine learning model obtained after the training, calculate institute
State sample data to each local Gaussian process machine learning model distance, and obtain apart from the corresponding local Gaussian of maximum value
Process machine learning model;
C: obtaining the location factor of harmonic drive joint output torque, judges each distance in the location factor
Whether the maximum value in parameter is less than preset threshold;
D: if so, updating local Gaussian process machine learning model, and by updated local Gaussian process machine learning
The central point of model is as new match point;It returns and executes the step B, until the judging result of the step C is no;
E: if it is not, the harmonic drive joint output torque exported according to each local Gaussian process machine learning model
Predicted value weighted average, calculate the corresponding predicted value mean value of predicted value of harmonic drive joint output torque.
In a kind of specific embodiment of the embodiment of the present invention, the acquisition module 302 is also used to:
A1: the sample data sets are divided into multiple subclass, each subclass is then inputted into each part respectively
It is trained in Gaussian process machine learning model, wherein the subclass is according to the sample data in the subclass
What location factor determined at a distance from the central point of the local Gaussian process machine learning model;
A2: being directed to each local Gaussian process machine learning model, using formula,Meter
Calculate predicted value mean value, wherein
f(x*) it is predicted value mean value;Y is the predicted value of the local Gaussian process machine learning model of last iteration;α is
Predicted vector;x*For sample data;For the variance of Gaussian noise;I is the current matrix of training sample;()TFor transposed matrix;
K is the covariance matrix when previous iteration;
A3: utilizing formula,Calculate the initial prediction pair when previous iteration
The covariance answered, wherein
V (x*) is the corresponding covariance of initial prediction when previous iteration;K () is that covariance calculates function;K* is institute
State covariance of the local Gaussian process machine learning model in last iteration;K is covariance matrix;X* is sample data;
For the variance of Gaussian noise;I is the current matrix of training sample;()TFor transposed matrix;
A4: judge the difference between the covariance and preset covariance, and prediction mean value and actual torque value
Within a preset range whether difference;
A5: if so, using the local Gaussian process machine learning model as the local Gaussian process engineering after training
Practise model;
A6: it if it is not, updating the parameter of local Gaussian process machine learning model, and returns and executes A2 step, until obtaining
Local Gaussian process machine learning model after training.
In a kind of specific embodiment of the embodiment of the present invention, the acquisition module 302 is also used to:
Using formula,Calculate harmonic drive joint output torque position because
Son, wherein
wkPosition for harmonic drive joint output torque model value relative to k-th of local Gaussian process machine learning model
Set parameter;Exp () is the exponential function using the natural truth of a matter bottom of as;X is the subclass of input sample data;ckIt is high for each part
The location parameter of the central point of this process machine learning model;()TFor transposed matrix;W be and local Gaussian process machine learning
The identical diagonal matrix of model width;K is the number of local Gaussian process machine learning model;
Utilize formula, w (x, o)=[w1,...,wk], obtain the location factor of harmonic drive joint output torque, wherein
W (x, o) is the location factor of the corresponding harmonic drive joint data torque of input sample;X is input sample data
Subclass;O is the matrix that the central point of each local Gaussian process machine learning model is constituted;w1It is defeated for harmonic drive joint
Location parameter of the torque model value relative to the 1st local Gaussian process machine learning model out;wkIt is defeated for harmonic drive joint
Location parameter of the torque model value relative to k-th of local Gaussian process machine learning model out.
In a kind of specific embodiment of the embodiment of the present invention, the acquisition module 302 is also used to:
Obtain new local Gaussian process machine learning model, and by the new local Gaussian process machine learning model
It is middle to be used as updated local Gaussian process machine learning model.
In a kind of specific embodiment of the embodiment of the present invention, the acquisition module 302 is also used to:
Using formula,Sample data is obtained in each local Gaussian process machine learning model
The probability of appearance, wherein
P (k | x) it is the probability that sample data x occurs in k-th of local Gaussian process machine learning model;X is sample
Data;K is the quantity of local Gaussian process machine learning model;M be obtain with the harmonic drive joint output torque mould
The number of the corresponding local Gaussian process machine learning model of offset;wkFor harmonic drive joint output torque model value relative to
The location parameter of k-th of local Gaussian process machine learning model;
Using formula,The corresponding predicted value mean value of harmonic drive joint output torque model value is calculated,
Wherein,
For the corresponding predicted value mean value of harmonic drive joint output torque model value;For k-th of local Gaussian process
The predicted value of machine learning model.
The foregoing is merely illustrative of the preferred embodiments of the present invention, is not intended to limit the invention, all in essence of the invention
Made any modifications, equivalent replacements, and improvements etc., should all be included in the protection scope of the present invention within mind and principle.
Claims (14)
1. a kind of acquisition methods of robot harmonic drive joint actual torque value, which is characterized in that the described method includes:
1) it, using harmonic drive flexibility of joint error model, is passed according to the position signal of the motor side in harmonic drive joint, harmonic wave
The position signal of diarthrodial connecting-rod head and the set of current of electric data acquisition harmonic drive joint output torque model value;
2) sample data sets, are obtained, using the sample data sets training machine learning model, then according to the harmonic wave
The position signal and current of electric data of the position signal of the motor side of driving joint, the connecting-rod head in harmonic drive joint, are obtained
Take the predicted value of harmonic drive joint output torque;
3), the harmonic drive joint output torque model value and the predicted value mean value are filtered, obtain institute
State the corresponding actual torque value of harmonic drive joint output torque model value.
2. a kind of acquisition methods of robot harmonic drive joint according to claim 1 actual torque value, feature exist
In the step 1), comprising:
Using formula,Calculate the harmonic drive in harmonic drive joint
Deformation angle, wherein
ΔθfFor the harmonic drive deformation angle in harmonic drive joint;qdFor the position signal of the connecting-rod head in harmonic drive joint;qm
For the position signal of motor side;Sgn () is sign function;KωFor the stiffness coefficient of the wave producer in harmonic drive joint;Ii
For current of electric;cωFor the lag coefficient of the wave producer in harmonic drive joint;L is the harmonic reduction in harmonic drive joint
Device reduction ratio;E is the nature truth of a matter;| | it is mod function;θerrFor the flexibility error between the defeated end of motor and connecting rod output end;
I is the quantity of current of electric parameter;
Using formula,Calculate Flexible Transmission torsion angle, wherein
Δ θ is Flexible Transmission torsion angle;ΔθωFor the torsion angle of the input of wave producer;It exports and reverses for wave producer
Angle;The transmission ratio of l harmonic drive joint speed changer;
Utilize formula, τf=a1Δθ+a2Δθ2+a3Δθ3, harmonic drive flexibility of joint output torque is calculated, and then obtain harmonic wave
The set of driving joint output torque model value, wherein
τfFor harmonic drive flexibility of joint output torque;a1For preset first parameter;a2For preset second parameter;a3It is default
Third parameter;
Using formula,Calculate harmonic drive joint output torque model value, wherein
τmodelFor harmonic drive joint output torque model value;Tan () is tangent function;cfFor preset first constant;KfoFor
Preset second constant.
3. a kind of acquisition methods of robot harmonic drive joint according to claim 1 actual torque value, feature exist
In the step 2) includes:
A: using the harmonic drive joint actual output torque value as target, the sample data sets are divided into multiple subsets
It closes, then each subclass is inputted respectively in each local Gaussian process machine learning model and is trained, wherein the sample
Notebook data includes: the position signal of the motor side in harmonic drive joint, the position signal of the connecting-rod head in harmonic drive joint, motor
Current data and the actual torque value in harmonic drive joint;
B: the position of the position signal of the motor side in the harmonic drive joint in step 1), the connecting-rod head in harmonic drive joint is believed
Number and current of electric data be input in the local Gaussian process machine learning model obtained after the training, calculate the sample
Notebook data to each local Gaussian process machine learning model distance, and obtain apart from the corresponding local Gaussian process of maximum value
Machine learning model;
C: obtaining the location factor of harmonic drive joint output torque, judges each distance parameter in the location factor
In maximum value whether be less than preset threshold;
D: if so, updating local Gaussian process machine learning model, and by updated local Gaussian process machine learning model
Central point as new match point;It returns and executes the step B, until the judging result of the step C is no;
E: if it is not, according to the pre- of the harmonic drive joint output torque of each local Gaussian process machine learning model output
The weighted average of measured value calculates the corresponding predicted value mean value of predicted value of harmonic drive joint output torque.
4. a kind of acquisition methods of robot harmonic drive joint according to claim 3 actual torque value, feature exist
In the step A includes:
A1: the sample data sets are divided into multiple subclass, each subclass is then inputted into each local Gaussian respectively
It is trained in process machine learning model, wherein the subclass is the position according to the sample data in the subclass
What the factor determined at a distance from the central point of the local Gaussian process machine learning model;
A2: being directed to each local Gaussian process machine learning model, using formula,Calculate prediction
It is worth mean value, wherein
f(x*) it is predicted value mean value;Y is the predicted value of the local Gaussian process machine learning model of last iteration;α is prediction
Vector;x*For sample data;For the variance of Gaussian noise;I is the current matrix of training sample;()TFor transposed matrix;K is
Covariance matrix when previous iteration;
A3: utilizing formula,It is corresponding to calculate the initial prediction when previous iteration
Covariance, wherein
V(x*) it is the corresponding covariance of initial prediction when previous iteration;K () is that covariance calculates function;k*For the part
Covariance of the Gaussian process machine learning model in last iteration;K is covariance matrix;x*For sample data;For Gauss
The variance of noise;I is the current matrix of training sample;()TFor transposed matrix;
A4: judge the difference between the covariance and preset covariance, and the difference of prediction mean value and actual torque value
Whether within a preset range;
A5: if so, using the local Gaussian process machine learning model as the local Gaussian process machine learning mould after training
Type;
A6: it if it is not, updating the parameter of local Gaussian process machine learning model, and returns and executes A2 step, until being trained
Local Gaussian process machine learning model afterwards.
5. a kind of acquisition methods of robot harmonic drive joint according to claim 3 actual torque value, feature exist
In obtaining the location factor of harmonic drive joint output torque, comprising:
Using formula,The location factor of harmonic drive joint output torque is calculated,
Wherein,
wkPosition ginseng for harmonic drive joint output torque model value relative to k-th of local Gaussian process machine learning model
Number;Exp () is the exponential function using the natural truth of a matter bottom of as;X is the subclass of input sample data;ckFor each local Gaussian mistake
The location parameter of the central point of journey machine learning model;()TFor transposed matrix;W be and local Gaussian process machine learning model
Diagonal matrix of same size;K is the number of local Gaussian process machine learning model;
Utilize formula, w (x, o)=[w1,...,wk], obtain the location factor of harmonic drive joint output torque, wherein
W (x, o) is the location factor of the corresponding harmonic drive joint data torque of input sample;X is the son of input sample data
Set;O is the matrix that the central point of each local Gaussian process machine learning model is constituted;w1For harmonic drive joint power output
Location parameter of the square model value relative to the 1st local Gaussian process machine learning model;wkFor harmonic drive joint power output
Location parameter of the square model value relative to k-th of local Gaussian process machine learning model.
6. a kind of acquisition methods of robot harmonic drive joint according to claim 3 actual torque value, feature exist
In the update local Gaussian process machine learning model, comprising:
New local Gaussian process machine learning model is obtained, and will be made in the new local Gaussian process machine learning model
For updated local Gaussian process machine learning model.
7. a kind of acquisition methods of robot harmonic drive joint according to claim 3 actual torque value, feature exist
In the step E, comprising:
Using formula,Sample data is obtained to occur in each local Gaussian process machine learning model
Probability, wherein
P (k | x) it is the probability that sample data x occurs in k-th of local Gaussian process machine learning model;X is sample data;
K is the quantity of local Gaussian process machine learning model;M be obtain with the harmonic drive joint output torque model value pair
The number for the local Gaussian process machine learning model answered;wkIt is harmonic drive joint output torque model value relative to k-th
The location parameter of local Gaussian process machine learning model;
Using formula,The corresponding predicted value mean value of harmonic drive joint output torque model value is calculated,
In,
For the corresponding predicted value mean value of harmonic drive joint output torque model value;For k-th of local Gaussian process engineering
Practise the predicted value of model.
8. a kind of acquisition device of robot harmonic drive joint actual torque value, which is characterized in that described device includes:
Output module, for utilizing harmonic drive flexibility of joint error model, according to the position of the motor side in harmonic drive joint
Signal, harmonic drive joint connecting-rod head position signal and current of electric data acquisition harmonic drive joint output torque mould
The set of offset;
Module is obtained, for obtaining sample data sets, using the sample data sets training machine learning model, then root
According to the position signal of the motor side in the harmonic drive joint, the position signal of the connecting-rod head in harmonic drive joint and motor electricity
Flow data obtains the predicted value of harmonic drive joint output torque;
Filter module, for being filtered place to the harmonic drive joint output torque model value and the predicted value mean value
Reason, obtains the corresponding actual torque value of the harmonic drive joint output torque model value.
9. a kind of acquisition device of robot harmonic drive joint according to claim 8 actual torque value, feature exist
In the output module is also used to:
Using formula,Calculate the harmonic drive in harmonic drive joint
Deformation angle, wherein
ΔθfFor the harmonic drive deformation angle in harmonic drive joint;qdFor the position signal of the connecting-rod head in harmonic drive joint;qm
For the position signal of motor side;Sgn () is sign function;KωFor the stiffness coefficient of the wave producer in harmonic drive joint;Ii
For current of electric;cωFor the lag coefficient of the wave producer in harmonic drive joint;L is the harmonic reduction in harmonic drive joint
Device reduction ratio;E is the nature truth of a matter;| | it is mod function;θerrFor the flexibility error between the defeated end of motor and connecting rod output end;
I is the quantity of current of electric parameter;
Using formula,Calculate Flexible Transmission torsion angle, wherein
Δ θ is Flexible Transmission torsion angle;ΔθωFor the torsion angle of the input of wave producer;It exports and reverses for wave producer
Angle;The transmission ratio of l harmonic drive joint speed changer;
Utilize formula, τf=a1Δθ+a2Δθ2+a3Δθ3, harmonic drive flexibility of joint output torque is calculated, and then obtain harmonic wave
The set of driving joint output torque model value, wherein
τfFor harmonic drive flexibility of joint output torque;a1For preset first parameter;a2For preset second parameter;a3It is default
Third parameter;
Using formula,Calculate harmonic drive joint output torque model value, wherein
τmodelFor harmonic drive joint output torque model value;Tan () is tangent function;cfFor preset first constant;KfoFor
Preset second constant.
10. a kind of acquisition device of robot harmonic drive joint according to claim 8 actual torque value, feature exist
In the acquisition module is also used to:
A: using the harmonic drive joint actual output torque value as target, the sample data sets are divided into multiple subsets
It closes, then each subclass is inputted respectively in each local Gaussian process machine learning model and is trained, wherein the sample
Notebook data includes: the position signal of the motor side in harmonic drive joint, the position signal of the connecting-rod head in harmonic drive joint, motor
Current data and the actual torque value in harmonic drive joint;
B: the position of the position signal of the motor side in the harmonic drive joint in step 1), the connecting-rod head in harmonic drive joint is believed
Number and current of electric data be input in the local Gaussian process machine learning model obtained after the training, calculate the sample
Notebook data to each local Gaussian process machine learning model distance, and obtain apart from the corresponding local Gaussian process of maximum value
Machine learning model;
C: obtaining the location factor of harmonic drive joint output torque, judges each distance parameter in the location factor
In maximum value whether be less than preset threshold;
D: if so, updating local Gaussian process machine learning model, and by updated local Gaussian process machine learning model
Central point as new match point;It returns and executes the step B, until the judging result of the step C is no;
E: if it is not, according to the pre- of the harmonic drive joint output torque of each local Gaussian process machine learning model output
The weighted average of measured value calculates the corresponding predicted value mean value of predicted value of harmonic drive joint output torque.
11. a kind of acquisition device of robot harmonic drive joint according to claim 10 actual torque value, feature
It is, the acquisition module is also used to:
A1: the sample data sets are divided into multiple subclass, each subclass is then inputted into each local Gaussian respectively
It is trained in process machine learning model, wherein the subclass is the position according to the sample data in the subclass
What the factor determined at a distance from the central point of the local Gaussian process machine learning model;
A2: being directed to each local Gaussian process machine learning model, using formula,Calculate prediction
It is worth mean value, wherein
f(x*) it is predicted value mean value;Y is the predicted value of the local Gaussian process machine learning model of last iteration;α is prediction
Vector;x*For sample data;For the variance of Gaussian noise;I is the current matrix of training sample;()TFor transposed matrix;K is
Covariance matrix when previous iteration;
A3: utilizing formula,It is corresponding to calculate the initial prediction when previous iteration
Covariance, wherein
V(x*) it is the corresponding covariance of initial prediction when previous iteration;K () is that covariance calculates function;k*For the part
Covariance of the Gaussian process machine learning model in last iteration;K is covariance matrix;x*For sample data;For Gauss
The variance of noise;I is the current matrix of training sample;()TFor transposed matrix;
A4: judge the difference between the covariance and preset covariance, and the difference of prediction mean value and actual torque value
Whether within a preset range;
A5: if so, using the local Gaussian process machine learning model as the local Gaussian process machine learning mould after training
Type;
A6: it if it is not, updating the parameter of local Gaussian process machine learning model, and returns and executes A2 step, until being trained
Local Gaussian process machine learning model afterwards.
12. a kind of acquisition device of robot harmonic drive joint according to claim 10 actual torque value, feature
It is, the acquisition module is also used to:
Using formula,The location factor of harmonic drive joint output torque is calculated,
Wherein,
wkPosition ginseng for harmonic drive joint output torque model value relative to k-th of local Gaussian process machine learning model
Number;Exp () is the exponential function using the natural truth of a matter bottom of as;X is the subclass of input sample data;ckFor each local Gaussian mistake
The location parameter of the central point of journey machine learning model;()TFor transposed matrix;W be and local Gaussian process machine learning model
Diagonal matrix of same size;K is the number of local Gaussian process machine learning model;
Utilize formula, w (x, o)=[w1,...,wk], obtain the location factor of harmonic drive joint output torque, wherein
W (x, o) is the location factor of the corresponding harmonic drive joint data torque of input sample;X is the son of input sample data
Set;O is the matrix that the central point of each local Gaussian process machine learning model is constituted;w1For harmonic drive joint power output
Location parameter of the square model value relative to the 1st local Gaussian process machine learning model;wkFor harmonic drive joint power output
Location parameter of the square model value relative to k-th of local Gaussian process machine learning model.
13. a kind of acquisition device of robot harmonic drive joint according to claim 10 actual torque value, feature
It is, the acquisition module is also used to:
New local Gaussian process machine learning model is obtained, and will be made in the new local Gaussian process machine learning model
For updated local Gaussian process machine learning model.
14. a kind of acquisition of robot harmonic drive joint according to claim 10 actual torque value exists, feature exists
In the acquisition module is also used to:
Using formula,Sample data is obtained to occur in each local Gaussian process machine learning model
Probability, wherein
P (k | x) it is the probability that sample data x occurs in k-th of local Gaussian process machine learning model;X is sample data;
K is the quantity of local Gaussian process machine learning model;M be obtain with the harmonic drive joint output torque model value pair
The number for the local Gaussian process machine learning model answered;wkIt is harmonic drive joint output torque model value relative to k-th
The location parameter of local Gaussian process machine learning model;
Using formula,The corresponding predicted value mean value of harmonic drive joint output torque model value is calculated,
In,
For the corresponding predicted value mean value of harmonic drive joint output torque model value;For k-th of local Gaussian process engineering
Practise the predicted value of model.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201811587023.1A CN109623819B (en) | 2018-12-25 | 2018-12-25 | Method and device for acquiring actual torque value of harmonic drive joint of robot |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201811587023.1A CN109623819B (en) | 2018-12-25 | 2018-12-25 | Method and device for acquiring actual torque value of harmonic drive joint of robot |
Publications (2)
Publication Number | Publication Date |
---|---|
CN109623819A true CN109623819A (en) | 2019-04-16 |
CN109623819B CN109623819B (en) | 2020-08-21 |
Family
ID=66077238
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201811587023.1A Active CN109623819B (en) | 2018-12-25 | 2018-12-25 | Method and device for acquiring actual torque value of harmonic drive joint of robot |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN109623819B (en) |
Cited By (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN112276935A (en) * | 2019-07-22 | 2021-01-29 | 发那科株式会社 | Position and orientation determination device, position and orientation determination method, and position and orientation determination program |
CN112388628A (en) * | 2019-08-13 | 2021-02-23 | 罗伯特·博世有限公司 | Apparatus and method for training a gaussian process regression model |
CN112847365A (en) * | 2021-01-07 | 2021-05-28 | 西安电子科技大学 | Torque estimation method |
IT202100030044A1 (en) | 2021-11-26 | 2023-05-26 | Ergotech Srl | NEW HARMONIC REDUCER WITH COMPACT STRUCTURE MADE WITH PLASTIC MATERIAL |
Citations (8)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101122777A (en) * | 2007-09-18 | 2008-02-13 | 湖南大学 | Large condenser underwater operation environment two-joint robot control method |
CN102626930A (en) * | 2012-04-28 | 2012-08-08 | 哈尔滨工业大学 | Mechanical arm modular joint with power-off brake and multiple perceptive functions |
CN106873383A (en) * | 2017-04-17 | 2017-06-20 | 珞石(北京)科技有限公司 | A kind of On-Line Control Method for reducing industrial robot vibration |
US9694495B1 (en) * | 2013-06-24 | 2017-07-04 | Redwood Robotics Inc. | Virtual tools for programming a robot arm |
CN106994686A (en) * | 2016-12-01 | 2017-08-01 | 遨博(北京)智能科技有限公司 | The computational methods and device, robot of joint external force torque |
CN108284442A (en) * | 2017-01-24 | 2018-07-17 | 中国北方车辆研究所 | A kind of mechanical arm flexible joint control method based on fuzzy neural network |
CN108896304A (en) * | 2018-07-19 | 2018-11-27 | 中科新松有限公司 | Harmonic reducer of robot test device and system |
CN108994837A (en) * | 2018-08-20 | 2018-12-14 | 哈工大机器人(合肥)国际创新研究院 | A kind of mechanical arm zero-g balance control method of Dynamics Compensation |
-
2018
- 2018-12-25 CN CN201811587023.1A patent/CN109623819B/en active Active
Patent Citations (8)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101122777A (en) * | 2007-09-18 | 2008-02-13 | 湖南大学 | Large condenser underwater operation environment two-joint robot control method |
CN102626930A (en) * | 2012-04-28 | 2012-08-08 | 哈尔滨工业大学 | Mechanical arm modular joint with power-off brake and multiple perceptive functions |
US9694495B1 (en) * | 2013-06-24 | 2017-07-04 | Redwood Robotics Inc. | Virtual tools for programming a robot arm |
CN106994686A (en) * | 2016-12-01 | 2017-08-01 | 遨博(北京)智能科技有限公司 | The computational methods and device, robot of joint external force torque |
CN108284442A (en) * | 2017-01-24 | 2018-07-17 | 中国北方车辆研究所 | A kind of mechanical arm flexible joint control method based on fuzzy neural network |
CN106873383A (en) * | 2017-04-17 | 2017-06-20 | 珞石(北京)科技有限公司 | A kind of On-Line Control Method for reducing industrial robot vibration |
CN108896304A (en) * | 2018-07-19 | 2018-11-27 | 中科新松有限公司 | Harmonic reducer of robot test device and system |
CN108994837A (en) * | 2018-08-20 | 2018-12-14 | 哈工大机器人(合肥)国际创新研究院 | A kind of mechanical arm zero-g balance control method of Dynamics Compensation |
Cited By (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN112276935A (en) * | 2019-07-22 | 2021-01-29 | 发那科株式会社 | Position and orientation determination device, position and orientation determination method, and position and orientation determination program |
CN112388628A (en) * | 2019-08-13 | 2021-02-23 | 罗伯特·博世有限公司 | Apparatus and method for training a gaussian process regression model |
CN112847365A (en) * | 2021-01-07 | 2021-05-28 | 西安电子科技大学 | Torque estimation method |
CN112847365B (en) * | 2021-01-07 | 2022-08-02 | 西安电子科技大学 | Torque estimation method |
IT202100030044A1 (en) | 2021-11-26 | 2023-05-26 | Ergotech Srl | NEW HARMONIC REDUCER WITH COMPACT STRUCTURE MADE WITH PLASTIC MATERIAL |
Also Published As
Publication number | Publication date |
---|---|
CN109623819B (en) | 2020-08-21 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN109623819A (en) | The acquisition methods and device of a kind of robot harmonic drive joint actual torque value | |
CN110488202A (en) | Vehicular battery state-of-charge estimation method based on deep neural network | |
CN106385211B (en) | A kind of stepper motor load-toque estimate method | |
CN103684183B (en) | Rotational Speed of Asynchronous Motor method of estimation | |
CN107147342A (en) | A kind of induction motor parameter identification system and method | |
CN109873587B (en) | Automatic multi-parameter identification method for permanent magnet synchronous motor | |
CN109702745A (en) | A kind of modeling method of joint of robot fluctuation moment of friction | |
CN110296802A (en) | Shaft tower bolt looseness method of discrimination based on vibrating speed sensors waveform acquisition | |
CN112101457B (en) | PMSM demagnetizing fault diagnosis method based on torque signal fuzzy intelligent learning | |
CN101753073A (en) | Compensation system for cogging torque of motor and method thereof | |
CN108365787A (en) | A kind of Permanent-magnet Synchronous-motor Speed Servo System and its design method based on internal model control | |
CN106602952B (en) | A kind of PMSM permanent magnet flux linkage full rank discrimination method | |
CN112782594B (en) | Method for estimating SOC (state of charge) of lithium battery by data-driven algorithm considering internal resistance | |
CN105929340A (en) | Method for estimating battery SOC based on ARIMA | |
CN109131342A (en) | Acceleration transducer and the fusion speed-measuring method and device for taking turns shaft speed sensor | |
CN106939840A (en) | Method and apparatus for determining the gas mass flow in internal combustion engine | |
CN116659803A (en) | Method for acquiring aerodynamic load of continuous wind tunnel based on balance zero point on-line monitoring | |
CN116680569A (en) | Wind resource prediction method and device based on RWKV neural network | |
CN111781499A (en) | Electric tuning test method, device, electronic equipment, storage medium and system | |
CN111055277A (en) | Robot flexible parameter acquisition method and device, computer equipment and storage medium | |
Beering et al. | Estimation of rotational speeds based on gearbox vibrations via artificial neural networks | |
WO2013000654A2 (en) | Control device and method for controlling an electric machine | |
CN113114079A (en) | Electric excavator main drive motor control method and system and electric excavator | |
CN114070146A (en) | Fault detection method, device, equipment and storage medium | |
CN105129073B (en) | A kind of aircraft gust alleviation adaptive feedforward control method based on GOFIR |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant | ||
GR01 | Patent grant |