CN111428317B - Joint friction torque compensation method based on 5G and cyclic neural network - Google Patents
Joint friction torque compensation method based on 5G and cyclic neural network Download PDFInfo
- Publication number
- CN111428317B CN111428317B CN202010262141.6A CN202010262141A CN111428317B CN 111428317 B CN111428317 B CN 111428317B CN 202010262141 A CN202010262141 A CN 202010262141A CN 111428317 B CN111428317 B CN 111428317B
- Authority
- CN
- China
- Prior art keywords
- moment
- friction
- joint
- friction torque
- time
- 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.)
- Active
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/044—Recurrent networks, e.g. Hopfield networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
- G06N3/084—Backpropagation, e.g. using gradient descent
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Health & Medical Sciences (AREA)
- Computing Systems (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- Life Sciences & Earth Sciences (AREA)
- Molecular Biology (AREA)
- Artificial Intelligence (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Health & Medical Sciences (AREA)
- Feedback Control In General (AREA)
- Manipulator (AREA)
Abstract
The invention relates to the technical field of robots, in particular to a joint friction torque compensation method based on a 5G and a cyclic neural network, which comprises the following steps of S1: the information collector is utilized to collect relevant information in real time in the movement process of the joint robot; step S2: after the information collector collects the information, predicting the friction moment of the next period through the friction moment estimator; step S3: the controller feeds back the calculated current friction torque compensation quantity to the driver, and the driver gives an action instruction to the joint robot and makes the joint robot obtain torque compensation in the next control period, so that the joint tracking precision of the joint robot can be improved; according to the invention, the friction model is fitted by adopting the cyclic neural network, the current friction moment compensation quantity is calculated directly through the model and is fed back to the robot control, so that the tracking precision of the system is improved.
Description
Technical Field
The invention relates to the technical field of robots, in particular to a joint friction torque compensation method based on a 5G and a cyclic neural network.
Background
The friction of the industrial robot joint is one of main factors causing the reduction of the tracking precision of the robot joint, especially in a low-speed state, the tracking precision is more obviously influenced by the joint friction, and the main reasons are that the friction is nonlinear change at a low speed and linear change at a high speed, so that the effective compensation of the friction moment of the robot joint is the key for improving the tracking precision, and the modeling of a friction model is the key for compensating the friction moment. Currently, friction models are classified into two types in the industry: (1) A static friction model, which regards friction as a function of speed, such as coulomb model, etc.; (2) The dynamic friction model is more established from a microscopic angle, and the found pre-slip phenomenon is described by adopting a differential equation mode, such as a Dahl model.
The current moment compensation method based on two friction models has obvious defects: (1) Dynamic friction models, which are difficult to succeed in practical applications due to the large increase in the number of parameters and the introduction of more unmeasurable quantities; (2) In the static friction model, as the parameters such as joint temperature, load, lubrication, abrasion and the like are continuously changed and the friction force is continuously changed along with the function of speed and the static friction is continuously changed, the real friction model cannot be well fitted, so that the feedforward compensation method based on the static friction model is invalid, and the joint tracking precision of the robot is reduced.
Disclosure of Invention
The invention designs a friction model modeling and friction moment compensation method based on a 5G and a cyclic neural network aiming at the problems of the background technology. In the compensation method, the current friction force of the system is not only related to the current angular velocity but also related to the previous angular velocity and friction force, so that factors such as temperature, load, lubrication, abrasion and dynamic and static friction switching of the robot are indirectly considered, friction moment can be better compensated compared with a static friction model, the joint tracking precision of the robot is improved, and a large number of unmeasurable quantities are avoided compared with a dynamic friction model.
The invention is realized by the following technical scheme:
a joint friction torque compensation method based on 5G and a cyclic neural network comprises the following steps:
step S1: the information collector is utilized to collect relevant information in real time in the movement process of the joint robot;
step S2: after the information collector collects information, predicting the next periodic friction moment through a friction moment estimator, wherein the friction moment estimator establishes two-way data communication with a cloud model training platform through a 5G network, and a corrected circulating neural network is formed among a signal output end of the friction moment estimator, a signal input end of a controller, a signal collector input end and a signal input end of the friction moment estimator in sequence;
step S3: the controller feeds back the calculated current friction torque compensation quantity to the driver, and the driver gives an action instruction to the joint robot and makes the joint robot obtain torque compensation in the next control period, so that the joint tracking precision of the joint robot can be improved.
As a further improvement of the above scheme, the related information in the step S1 includes a current angular velocity ω, a current moment M, and a current target velocity
Three sets of readily available information.
As a further improvement of the scheme, the current angular velocity omega and the current moment M in the related information can be directly read through a servo motor on the articulated robot, and the current target velocity
Can be obtained by a controller path planning module in the joint robot.
As a further improvement of the above solution, the signal output end of the information collector in step S2 may be electrically connected to the signal input end of the friction torque estimator, or may be electrically connected to the signal input end of the controller.
As a further improvement of the above solution, in the step S2, the friction torque estimator estimates the friction torque by using a combination of a 5G network and a modified cyclic neural network, and parameters of the cyclic neural network may be obtained by real-time training of the 5G network in a cloud model training platform, and the friction torque is obtained by real-time calculation in the friction torque estimator of the joint robot.
As a further improvement of the above solution, in the step S2, the friction torque may be modeled by using a modified recurrent neural network, and the specific procedure is as follows:
(a) Forward calculation:
z t ＝U 1 M t +U 2 ω t +Wh t-1 +b
h t ＝f(z t )
o t ＝Vh t +c
wherein U is 1 Is to input a current moment-state weight matrix, U 2 Is the input current angular velocity-state weight matrix, W is the state-state weight matrix, V is the state-output weight matrix, b and c are the network biases, h t The value of a hidden layer in the middle of a network at the moment t is represented, f (& gt) and sigma (& gt) represent nonlinear activation functions, in the process, f (& gt) adopts a Sigmoid function, sigma (& gt) adopts a Tanh function, and a model is output as a predicted friction torque
M t A friction torque vector omega representing time t-n to time t t An angular velocity vector representing the time t-n to the time t;
(b) Loss definition:
the controller compensates the current planned moment through the estimated friction moment and feeds back the current planned moment to the servo driver to complete the real-time updating of the friction moment, predict the angular velocity of the next period and realize the following operation
Defined as the predicted angular velocity at time t-1, this process is represented using a function Γ ():
wherein J t The moment calculated for the robot path planning module,
the angular velocity increasing value from the time t-1 to the time t is represented, I represents the moment of inertia, and the moment of inertia defaults to a constant value for a certain axis of the joint robot;
in the process, the angular velocity omega is currently measured at the moment t t And predicting angular velocity at time t
The regression Loss between the two is taken as Loss of Loss at t time, as follows: />
(c) And (5) reverse calculation:
first, calculate the Loss t For a pair of
Is the derivative of (2)
Next, calculate the Loss t Derivative of V and c
Finally, calculate the Loss t To U 1 、U 2 W and derivative of b
(d) And (5) weight updating:
wherein alpha represents learning rate, and alpha is 0.01 in the process.
Compared with the prior art, the invention has the beneficial effects that:
1. compared with the traditional fully-connected network, the invention not only considers the joint state information such as angular velocity, moment and the like at the current moment, but also considers the robot joint historical state information, and can save the dynamic quantity information which cannot be measured by self-learning, such as dynamic and static friction switching, lubrication, abrasion state and the like, in the model hidden layer by the difference between the friction moment estimated value and the friction moment measured value.
2. The cyclic neural network used in the invention is based on a model corrected by the traditional cyclic neural network, not only uses the angular velocity of the current measured value as an input variable, but also uses the friction torque of the current measured value as an input parameter, thereby effectively improving the accuracy of model prediction.
3. According to the invention, the joint friction torque compensation method based on 5G is adopted for the first time, so that real-time model training can be performed at the cloud, and compared with the traditional cloud training and robot side reasoning mode, the model timeliness can be improved. With continuous running of the robot, abrasion is gradually increased, and in a traditional deployment mode, the friction torque estimation model cannot be updated in real time, so that estimation deviation is larger and larger, and the problem is avoided based on real-time training of 5G.
Drawings
In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings that are needed for the description of the embodiments will be briefly described below, and it is obvious that the drawings in the following description are only some embodiments of the present invention, and that other drawings may be obtained according to these drawings without inventive effort for a person skilled in the art.
FIG. 1 is a control flow diagram of the present invention;
fig. 2 is a network structure diagram of the friction torque calculation in real time in the friction torque estimator in the present invention.
Detailed Description
In order to make the present application solution better understood by those skilled in the art, the following description will be made in detail and with reference to the accompanying drawings in the embodiments of the present application, it is apparent that the described embodiments are only some embodiments of the present application, not all embodiments. All other embodiments, which can be made by one of ordinary skill in the art based on the embodiments herein without making any inventive effort, shall fall within the scope of the present application.
The technical scheme of the invention is further described below with reference to the accompanying drawings.
A joint friction torque compensation method based on 5G and a cyclic neural network, as shown in figure 1, comprises the following steps:
step S1: the information acquisition device is utilized to acquire relevant information in real time in the movement process of the joint robot, wherein the relevant information comprises the current angular velocity omega, the current moment M and the current target velocity
Three groups of information which are easy to obtain, wherein the current angular speed omega and the current moment M can be directly read through a servo motor on the joint robot, and the current target speed +.>
The method can be obtained through a controller path planning module in the joint robot;
step S2: after the information collector collects information, the friction torque estimator predicts the next periodic friction torque, two-way data communication is established between the friction torque estimator and the cloud model training platform through a 5G network, a corrected cyclic neural network is formed among the signal output end of the friction torque estimator, the signal input end of the signal collector and the signal input end of the friction torque estimator in sequence, and the signal output end of the information collector can be electrically connected with the signal input end of the friction torque estimator and also can be electrically connected with the signal input end of the controller;
in the step S2, the friction torque estimator estimates the friction torque by adopting a mode of combining a 5G network and a modified cyclic neural network, and parameters of the cyclic neural network can be real-time in a cloud model training platform through the 5G networkTraining, real-time calculation of friction moment in friction moment estimator of joint robot, and network structure as shown in figure 2, wherein
Represents the predicted friction torque at time t, M t Representing the moment, ω, read from the servomotor at time t t Indicating the angular velocity value, h, read from the servo motor at time t t And representing the value of the hidden layer in the middle of the network at the moment t.
In the step S2, the friction moment can be modeled by using the corrected cyclic neural network, and the specific flow is as follows:
(a) Forward calculation:
z t ＝U 1 M t +U 2 ω t +Wh t-1 +b
h t ＝f(z t )
o t ＝Vh t +c
wherein U is 1 Is to input a current moment-state weight matrix, U 2 Is the input current angular velocity-state weight matrix, W is the state-state weight matrix, V is the state-output weight matrix, b and c are the network biases, h t The value of a hidden layer in the middle of a network at the moment t is represented, f (& gt) and sigma (& gt) represent nonlinear activation functions, in the process, f (& gt) adopts a Sigmoid function, sigma (& gt) adopts a Tanh function, and a model is output as a predicted friction torque
M t A friction torque vector omega representing time t-n to time t t An angular velocity vector representing the time t-n to the time t;
(b) Loss definition:
the controller compensates the current planned moment through the estimated friction moment and feeds back the current planned moment to the servo driver to complete the real-time friction momentUpdating, predicting the angular velocity of the next cycle, to be
Defined as the predicted angular velocity at time t-1, this process is represented using a function Γ ():
wherein J t The moment calculated for the robot path planning module,
the angular velocity increasing value from the time t-1 to the time t is represented, I represents the moment of inertia, and the moment of inertia defaults to a constant value for a certain axis of the joint robot;
in the process, the angular velocity omega is currently measured at the moment t t And predicting angular velocity at time t
The regression Loss between the two is taken as Loss of Loss at t time, as follows:
(c) And (5) reverse calculation:
first, calculate the Loss t For a pair of
Is the derivative of (2)
Next, calculate the Loss t Derivative of V and c
Finally, calculate the Loss t To U 1 、U 2 W and derivative of b
(d) And (5) weight updating:
wherein alpha represents learning rate, and alpha is 0.01 in the process.
Step S3: the controller feeds back the calculated current friction torque compensation quantity to the driver, and the driver gives an action instruction to the joint robot and makes the joint robot obtain torque compensation in the next control period, so that the joint tracking precision of the joint robot can be improved.
One specific application of the invention is:
a joint friction torque compensation method based on 5G and a cyclic neural network, as shown in figure 1, comprises the following steps:
step S1: the information acquisition device is utilized to acquire relevant information in real time in the movement process of the joint robot, wherein the relevant information comprises the current angular velocity omega, the current moment M and the current target velocity
Three groups of information which are easy to obtain, wherein the current angular speed omega and the current moment M can be directly read through a servo motor on the joint robot, and the current target speed +.>
The method can be obtained through a controller path planning module in the joint robot;
step S2: after the information collector collects the information, the friction torque estimator predicts the next periodic friction torque, and the friction torque estimator establishes two-way data communication with the cloud model training platform through a 5G network, so that the model can be trained in real time at the cloud by adopting a joint friction torque compensation method based on 5G for the first time, and compared with the traditional cloud training, the model timeliness can be improved by adopting a robot side reasoning mode; the signal output end of the friction torque estimator is also sequentially connected with the signal input end of the controller, the signal acquisition device input end and the signal input end of the friction torque estimator to form a corrected circulating neural network, the signal output end of the information acquisition device in the circulating neural network can be electrically connected with the signal input end of the friction torque estimator and also can be electrically connected with the signal input end of the controller, and the used circulating neural network not only uses the angular velocity of the current measured value as an input variable, but also uses the friction torque of the current measured value as an input parameter, thereby effectively improving the accuracy of model prediction;
in the step S2, the friction torque estimator adopts a mode of combining a 5G network and a modified cyclic neural network, parameters of the cyclic neural network can be obtained through real-time training of the 5G network in a cloud model training platform, the friction torque is obtained through real-time calculation in the friction torque estimator of the joint robot, and the calculation process of the friction torque is shown in figure 2; compared with the traditional fully-connected network, the method has the advantages that not only can the joint state information at the current moment such as angular speed, moment and the like be considered, but also the historical state information of the robot joint can be considered, and dynamic quantity information which cannot be measured by self-learning, such as dynamic and static friction switching, lubrication, abrasion state and the like, can be saved in a model hidden layer through the difference between the friction moment estimated value and the friction moment measured value.
In the step S2, the friction moment can be modeled by using the corrected cyclic neural network, and the specific flow is as follows:
(a) Forward calculation:
z t ＝U 1 M t +U 2 ω t +Wh t-1 +b
h t ＝f(z t )
o t ＝Vh t +c
wherein U is 1 Is to input a current moment-state weight matrix, U 2 Is the input current angular velocity-state weight matrix, W is the state-state weight matrix, V is the state-output weight matrix, b and c are the network biases, h t The value of a hidden layer in the middle of a network at the moment t is represented, f (& gt) and sigma (& gt) represent nonlinear activation functions, in the process, f (& gt) adopts a Sigmoid function, sigma (& gt) adopts a Tanh function, and a model is output as a predicted friction torque
M t A friction torque vector omega representing time t-n to time t t An angular velocity vector representing the time t-n to the time t;
(b) Loss definition:
the controller compensates the current planned moment through the estimated friction moment and feeds back the current planned moment to the servo driver to complete the real-time updating of the friction moment, predict the angular velocity of the next period and realize the following operation
Defined as the predicted angular velocity at time t-1, this process is represented using a function Γ ():
wherein J t The moment calculated for the robot path planning module,
the angular velocity increasing value from the time t-1 to the time t is represented, I represents the moment of inertia, and the moment of inertia defaults to a constant value for a certain axis of the joint robot;
in the process, the angular velocity omega is currently measured at the moment t t And predicting angular velocity at time t
Regression loss between as time tLoss of Loss is as follows:
(c) And (5) reverse calculation:
first, calculate the Loss t For a pair of
Is the derivative of (2)
Next, calculate the Loss t Derivative of V and c
Finally, calculate the Loss t To U 1 、U 2 W and derivative of b
(d) And (5) weight updating:
wherein alpha represents learning rate, and alpha is 0.01 in the process.
Step S3: the controller feeds back the calculated current friction torque compensation quantity to the driver, and the driver gives an action instruction to the joint robot and makes the joint robot obtain torque compensation in the next control period, so that the joint tracking precision of the joint robot can be improved.
In summary, the invention combines the 5G low-delay characteristic for the first time, carries out self-learning on the friction moment model of the joint robot in real time, predicts the friction moment according to the latest learned network parameters in real time, and compensates the moment; meanwhile, a circulating neural network is adopted to fit the friction model, the current friction moment compensation quantity is calculated directly through the model, and is fed back to the robot control, so that the tracking precision of the system is improved. In the compensation method, the current friction force of the system is not only related to the current angular velocity but also related to the previous angular velocity and friction force, so that factors such as temperature, load, lubrication, abrasion and dynamic and static friction switching of the robot are indirectly considered, friction moment can be better compensated compared with a static friction model, the joint tracking precision of the robot is improved, and a large number of unmeasurable quantities are avoided compared with a dynamic friction model.
The foregoing description of the preferred embodiments of the invention is not intended to be limiting, but rather is intended to cover all modifications, equivalents, and alternatives falling within the spirit and principles of the invention.
Claims (5)
1. The joint friction torque compensation method based on the 5G and the cyclic neural network is characterized by comprising the following steps of:
step S1: the information collector is utilized to collect relevant information in real time in the movement process of the joint robot;
step S2: after the information collector collects information, predicting the next periodic friction moment through a friction moment estimator, wherein the friction moment estimator establishes two-way data communication with a cloud model training platform through a 5G network, and a corrected circulating neural network is formed among a signal output end of the friction moment estimator, a signal input end of a controller, a signal collector input end and a signal input end of the friction moment estimator in sequence;
step S3: the controller feeds back the calculated current friction torque compensation quantity to the driver, and the driver gives an action instruction to the joint robot and makes the joint robot obtain torque compensation in the next control period, so that the joint tracking precision of the joint robot can be improved;
the friction moment can be modeled by using the corrected cyclic neural network in the step S2, and the specific flow is as follows:
(a) Forward calculation:
z t ＝U 1 M t +U 2 ω t +Wh t-1 +b
h t ＝f(z t )
o t ＝Vh t +c
wherein U is 1 Is to input a current moment-state weight matrix, U 2 Is the input current angular velocity-state weight matrix, W is the state-state weight matrix, V is the state-output weight matrix, b and c are the network biases, h t The value of a hidden layer in the middle of a network at the moment t is represented, f (& gt) and sigma (& gt) represent nonlinear activation functions, in the process, f (& gt) adopts a Sigmoid function, sigma (& gt) adopts a Tanh function, and a model is output as a predicted friction torque
M t A friction torque vector omega representing time t-n to time t t An angular velocity vector representing the time t-n to the time t;
(b) Loss definition:
the controller compensates the current planned moment through the estimated friction moment and feeds back the current planned moment to the servo driver to complete the real-time updating of the friction moment, predict the angular velocity of the next period and realize the following operation
Defined as the predicted angular velocity at time t-1, this process is represented using a function Γ ():
wherein J t The moment calculated for the robot path planning module,
the angular velocity increasing value from the time t-1 to the time t is represented, I represents the moment of inertia, and the moment of inertia defaults to a constant value for a certain axis of the joint robot;
in the process, the angular velocity omega is currently measured at the moment t t And predicting angular velocity at time t
The regression Loss between the two is taken as Loss of Loss at t time, as follows:
(c) And (5) reverse calculation:
first, calculate the Loss t For a pair of
Derivative of->
Next, calculate the Loss t Derivative of V and c
Finally, calculate the Loss t To U 1 、U 2 W and derivative of b
(d) And (5) weight updating:
wherein alpha represents learning rate, and alpha is 0.01 in the process.
2. The joint friction torque compensation method based on the 5G and the recurrent neural network according to claim 1, wherein the method comprises the following steps: the related information in the step S1 comprises a current angular velocity omega, a current moment M and a current target velocity
Three sets of readily available information.
3. The joint friction torque compensation method based on the 5G and the recurrent neural network according to claim 2, wherein the method comprises the following steps: the current angular velocity omega and the current moment M in the related information can be directly read through a servo motor on the joint robot, and the current target velocity
Can be obtained by a controller path planning module in the joint robot.
4. The joint friction torque compensation method based on the 5G and the recurrent neural network according to claim 1, wherein the method comprises the following steps: the signal output end of the information collector in the step S2 can be electrically connected with the signal input end of the friction torque estimator or the signal input end of the controller.
5. The joint friction torque compensation method based on the 5G and the recurrent neural network according to claim 1, wherein the method comprises the following steps: in the step S2, the friction torque estimator adopts a mode of combining a 5G network and a modified cyclic neural network, parameters of the cyclic neural network can be obtained by training the 5G network in a cloud model training platform in real time, and the friction torque is obtained by calculating in real time in the friction torque estimator of the joint robot.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010262141.6A CN111428317B (en) | 2020-04-06 | 2020-04-06 | Joint friction torque compensation method based on 5G and cyclic neural network |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010262141.6A CN111428317B (en) | 2020-04-06 | 2020-04-06 | Joint friction torque compensation method based on 5G and cyclic neural network |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN111428317A CN111428317A (en) | 2020-07-17 |
| CN111428317B true CN111428317B (en) | 2023-06-09 |
Family
ID=71557484
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202010262141.6A Active CN111428317B (en) | 2020-04-06 | 2020-04-06 | Joint friction torque compensation method based on 5G and cyclic neural network |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN111428317B (en) |
Families Citing this family (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112155552B (en) * | 2020-10-10 | 2022-05-31 | 杭州电子科技大学 | Joint wear condition analysis system based on lower limb joint moment estimation |
| CN112230542B (en) * | 2020-10-13 | 2022-10-28 | 上海傅利叶智能科技有限公司 | Method and device for compensating friction force or friction torque and rehabilitation robot |
| CN114594757B (en) * | 2020-12-07 | 2024-07-12 | 山东新松工业软件研究院股份有限公司 | Visual path planning method of cooperative robot |
| CN114619440B (en) * | 2020-12-10 | 2024-02-09 | 北京配天技术有限公司 | Method for correcting friction model, robot and computer readable storage medium |
| CN114074332B (en) * | 2022-01-19 | 2022-04-22 | 季华实验室 | Friction compensation method and device, electronic equipment and storage medium |
| CN114474078B (en) * | 2022-04-12 | 2022-06-17 | 季华实验室 | Friction force compensation method and device for mechanical arm, electronic equipment and storage medium |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| KR20140066544A (en) * | 2012-11-23 | 2014-06-02 | 삼성전자주식회사 | Robot and friction compensation method for the robot |
| CN109732605A (en) * | 2019-01-21 | 2019-05-10 | 厦门大学 | A kind of compensation method and system of joint of robot moment of friction |
| CN109976328A (en) * | 2017-12-28 | 2019-07-05 | 沈阳新松机器人自动化股份有限公司 | A kind of composite machine people |
| CN110346767A (en) * | 2019-05-31 | 2019-10-18 | 上海思致汽车工程技术有限公司 | A kind of test method and device for automobile lane change miscellaneous function |
| CN110705105A (en) * | 2019-10-08 | 2020-01-17 | 首都师范大学 | Modeling method and system for inverse dynamics model of robot |
-
2020
- 2020-04-06 CN CN202010262141.6A patent/CN111428317B/en active Active
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| KR20140066544A (en) * | 2012-11-23 | 2014-06-02 | 삼성전자주식회사 | Robot and friction compensation method for the robot |
| CN109976328A (en) * | 2017-12-28 | 2019-07-05 | 沈阳新松机器人自动化股份有限公司 | A kind of composite machine people |
| CN109732605A (en) * | 2019-01-21 | 2019-05-10 | 厦门大学 | A kind of compensation method and system of joint of robot moment of friction |
| CN110346767A (en) * | 2019-05-31 | 2019-10-18 | 上海思致汽车工程技术有限公司 | A kind of test method and device for automobile lane change miscellaneous function |
| CN110705105A (en) * | 2019-10-08 | 2020-01-17 | 首都师范大学 | Modeling method and system for inverse dynamics model of robot |
Non-Patent Citations (3)
| Title |
|---|
| Wu Yilei ,etc..Robust Recurrent Neural Network Control of Biped Robot.《J Intell Robot Syst》.2007,第151–155,168页. * |
| 党进 ; 倪风雷 ; 刘业超 ; 刘宏 ; .基于新型补偿控制策略的柔性关节控制器设计.机器人.2011,(第02期),第150-155页. * |
| 沈晓斌，等.机器人关节摩擦建模与自适应ＲＢＦ神经网络补偿计算力矩控制.《中国计量大学学报》.2020,第71-78页. * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN111428317A (en) | 2020-07-17 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111428317B (en) | Joint friction torque compensation method based on 5G and cyclic neural network | |
| JP2676397B2 (en) | Dynamic trajectory generation method for dynamic system | |
| CN110275436B (en) | RBF neural network self-adaptive control method of multi-single-arm manipulator | |
| WO2022121923A1 (en) | Smart modelling method and apparatus of complex industrial process digital twin system, device, and storage medium | |
| CN110877333A (en) | Flexible joint mechanical arm control method | |
| CN111608868B (en) | Maximum power tracking adaptive robust control system and method for wind power generation system | |
| CN109605377B (en) | Robot joint motion control method and system based on reinforcement learning | |
| CN111168682B (en) | Parallel robot and robust precise differentiator combined finite time convergence sliding mode control method | |
| CN116460860B (en) | Model-based robot offline reinforcement learning control method | |
| CN111273544B (en) | Radar pitching motion control method based on prediction RBF feedforward compensation type fuzzy PID | |
| CN115256401A (en) | Space manipulator shaft hole assembly variable impedance control method based on reinforcement learning | |
| Cheng et al. | Event-triggered-based adaptive command-filtered asymptotic tracking control for flexible robotic manipulators | |
| CN115416024A (en) | Moment-controlled mechanical arm autonomous trajectory planning method and system | |
| CN110039537B (en) | Online self-learning multi-joint motion planning method based on neural network | |
| CN113848905B (en) | Mobile robot track tracking method based on neural network and self-adaptive control | |
| Zeng et al. | DDPG-based continuous thickness and tension coupling control for the unsteady cold rolling process | |
| CN107450311A (en) | Inversion model modeling method and device and adaptive inverse control and device | |
| CN112947123B (en) | Exoskeleton robot tracking control method and system for inhibiting multi-source interference | |
| CN112462608B (en) | Discrete sliding mode track and speed tracking control method for high-speed train | |
| Al-Araji et al. | Design of a neural predictive controller for nonholonomic mobile robot based on posture identifier | |
| Qiao et al. | Application of reinforcement learning based on neural network to dynamic obstacle avoidance | |
| CN112947066B (en) | Manipulator improved finite time inversion control method | |
| CN115344047A (en) | Robot switching type predictive control trajectory tracking method based on neural network model | |
| CN114839878A (en) | Improved PPO algorithm-based biped robot walking stability optimization method | |
| Brown et al. | Design of a neural controller using reinforcement learning to control a rotational inverted pendulum |
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 |