CN111890350A - Robot, method of controlling the same, and computer-readable storage medium - Google Patents
Robot, method of controlling the same, and computer-readable storage medium Download PDFInfo
- Publication number
- CN111890350A CN111890350A CN202010536600.5A CN202010536600A CN111890350A CN 111890350 A CN111890350 A CN 111890350A CN 202010536600 A CN202010536600 A CN 202010536600A CN 111890350 A CN111890350 A CN 111890350A
- Authority
- CN
- China
- Prior art keywords
- track
- robot
- control model
- data
- output
- 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.)
- Pending
Links
Images
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
- B25J9/1656—Programme controls characterised by programming, planning systems for manipulators
- B25J9/1664—Programme controls characterised by programming, planning systems for manipulators characterised by motion, path, trajectory planning
-
- 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
- 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
- B25J9/163—Programme controls characterised by the control loop learning, adaptive, model based, rule based expert control
-
- 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
Abstract
The application discloses a robot, a control method thereof and a computer readable storage medium, wherein the method comprises the following steps: acquiring actual track data generated by the robot moving based on preset track data; determining a track offset error between the preset track data and the actual track data; inputting the track deviation error into a control model obtained by pre-training, and obtaining track correction data output by the control model; the control model is obtained by taking a track deviation error generated by the robot in the historical motion process and correspondingly determined track correction data as a training sample to train the control model; and controlling the robot by using the track correction data so as to correct the motion track of the robot. Through the mode, the complex track can be stably and accurately tracked, and the technical use difficulty can be reduced.
Description
Technical Field
The present application relates to the field of robots and intelligent control technologies, and in particular, to a robot, a control method thereof, and a computer-readable storage medium.
Background
At present, in the application of various robots, a track tracking technology is widely applied, and has important significance in stably and accurately tracking given reference tracks with different shapes. For example, in industrial dispensing, welding, and monitoring of movement of conveyor products, the robot is required to quickly converge to a given trajectory and track the trajectory in a given speed sequence. In the field of service robots, a mobile robot needs to follow a human at the same speed in real time or move at a given speed according to a given track, and the complexity of the track and the variability of the speed both increase the difficulty of the design of a tracking control algorithm.
In the tracking algorithm widely used at present, especially in industrial application, the motion of the robot is generally defined by the user's programming or a certain task environment is preset, and then the robot is repeatedly executed according to the plan. For complex track tracking in a service scene in daily life, the method needs to adjust a programming program in real time and is difficult to apply; in the industry, when the production requirements of small batches and various varieties are met, engineers are required to rewrite the program in time, heavy manual programming work is required, and the use is inconvenient. In addition, this method requires very specialized design knowledge for robot control in terms of tracking performance, depending on adjustment of programming parameters, and robot behavior is difficult to resemble to a human, which has limitations in service robot applications.
Disclosure of Invention
In order to solve the above problems, the present application provides a robot, a control method thereof, and a computer-readable storage medium, which can stably and accurately track a complex trajectory and can reduce the difficulty of technical use.
The technical scheme adopted by the application is to provide a control method of a robot, and the method comprises the following steps: acquiring actual track data generated by the robot moving based on preset track data; determining a track offset error between the preset track data and the actual track data; inputting the track deviation error into a control model obtained by pre-training, and obtaining track correction data output by the control model; the control model is obtained by taking a track deviation error generated by the robot in the historical motion process and correspondingly determined track correction data as a training sample to train the control model; and controlling the robot by using the track correction data so as to correct the motion track of the robot.
Wherein, the method also comprises: acquiring a track deviation error generated in the historical movement process of the robot and correspondingly determined track correction data; initializing the track deviation error in the historical movement process and the correspondingly determined track correction data to generate a training sample; inputting a training sample into a pre-established control model to train the control model; calculating to obtain the output weight of the control model; and when the output weight meets the preset requirement, finishing the training of the control model.
Wherein, calculate the output weight who obtains control model, include: calculating a hidden layer output matrix of the control model; and calculating to obtain the output weight according to the hidden layer output matrix.
Wherein, calculating according to the hidden layer output matrix to obtain the output weight, comprising: the output weight is calculated using the following formula:
wherein the content of the first and second substances,
representing trajectory modification data during historical motion; d represents a hidden layer output matrix; β represents the output weight.
Wherein, when the output weight satisfies the preset requirement, finish the training to the control model, include: and when the output weight and the input weight meet the first stability constraint condition and the hidden layer unit bias meets the second stability constraint condition, determining the output weight and finishing the training of the control model.
Wherein the first stability constraint is expressed using the following formula:
wherein, betaiRepresenting an output weight;
representing an input weight; the second stability constraint is expressed using the following formula: b i0; wherein, biIndicating a hidden layer cell bias.
Wherein, the computational formula of the control model is as follows:
wherein the content of the first and second substances,
representing trajectory correction data; beta is aiRepresenting output weights of the control model;
input weights representing a control model; e represents a trajectory offset error, and N represents the number of hidden layer neurons;
an activation function representing a control model; biThe hidden layer cell bias representing the control model.
Wherein, utilize the trajectory correction data control robot to revise the motion trajectory of robot, include: calculating the trajectory correction data to obtain a control instruction of the robot; and controlling the robot based on the control instruction to correct the motion trail of the robot.
Another technical solution adopted by the present application is to provide a robot, including a processor and a memory connected to the processor; the memory is for storing program data and the processor is for executing the program data to implement any of the methods provided in the above aspects.
Another technical solution adopted by the present application is to provide a computer-readable storage medium for storing program data, which when executed by a processor, is used for implementing any one of the methods provided in the above-mentioned solutions.
The beneficial effect of this application is: in contrast to the prior art, a robot control method of the present application includes: acquiring actual track data generated by the robot moving based on preset track data; determining a track offset error between the preset track data and the actual track data; inputting the track deviation error into a control model obtained by pre-training, and obtaining track correction data output by the control model; the control model is obtained by taking a track deviation error generated by the robot in the historical motion process and correspondingly determined track correction data as a training sample to train the control model; and controlling the robot by using the track correction data so as to correct the motion track of the robot. By the method, on one hand, the control model is trained through data generated in the historical movement process to correct the real-time track deviation error of the robot, so that the robot can complete the movement of a given track, and stable and accurate complex track tracking is realized; on the other hand, the control model can be better utilized to obtain the control habits of different users, so that the motion characteristic of the robot better accords with the control characteristic of the user during error correction, and the user can adjust and use the robot through historical data, thereby greatly reducing the technical use difficulty.
Drawings
In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings needed to be used in the description of the embodiments are briefly introduced below, and it is obvious that the drawings in the following description are only some embodiments of the present application, and it is obvious for those skilled in the art to obtain other drawings based on these drawings without creative efforts. Wherein:
fig. 1 is a schematic flowchart of a first embodiment of a control method of a robot provided in the present application;
FIG. 2 is a schematic diagram illustrating a comparison between a preset track and an actual track provided herein;
FIG. 3 is a schematic flow chart diagram illustrating the details of step 14 in FIG. 1 provided herein;
FIG. 4 is another schematic diagram comparing the preset track and the actual track provided in the present application;
fig. 5 is a schematic flowchart of a second embodiment of a control method of a robot provided by the present application;
FIG. 6 is a schematic diagram of an embodiment of a control model provided herein;
FIG. 7 is a schematic structural diagram of an embodiment of a robot provided herein;
FIG. 8 is a schematic structural diagram of an embodiment of a computer-readable storage medium provided in the present application.
Detailed Description
The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application. It is to be understood that the specific embodiments described herein are merely illustrative of the application and are not limiting of the application. It should be further noted that, for the convenience of description, only some of the structures related to the present application are shown in the drawings, not all of the structures. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present application.
Reference herein to "an embodiment" means that a particular feature, structure, or characteristic described in connection with the embodiment can be included in at least one embodiment of the application. The appearances of the phrase in various places in the specification are not necessarily all referring to the same embodiment, nor are separate or alternative embodiments mutually exclusive of other embodiments. It is explicitly and implicitly understood by one skilled in the art that the embodiments described herein can be combined with other embodiments.
In order to solve the problems, the control model is trained by using data generated by the robot in the historical motion process as training samples. And correcting the track deviation error generated in the actual track of the robot by using the trained control model so as to enable the robot to finish the motion of the given track and realize stable and accurate complex track tracking. The detailed description is given in the following examples.
Referring to fig. 1, fig. 1 is a schematic flowchart of a first embodiment of a control method of a robot provided by the present application, the method including:
step 11: and acquiring actual track data generated by the robot moving based on the preset track data.
In some embodiments, when the robot moves based on the preset trajectory data, the position information of each joint is collected by the sensor of the robot to obtain the position information of a plurality of continuous positions. The sensor may be a robot joint or a motor-side encoder to obtain positional information of the joint. Further, by processing the position information, the speed, direction, and the like of the robot at the current time can be acquired. The actual track data of the robot can be obtained according to the data collected by the sensors.
It can be understood that different robots have different data information obtained by the sensor, and reasonable data information is obtained according to the characteristics of the robots.
In some embodiments, step 11 may perform the acquisition of the actual trajectory at a set frequency, such as 60 seconds, 30 seconds, 20 seconds, 10 seconds, 5 seconds.
The description is made with reference to fig. 2: AB in fig. 2 represents the preset trajectory data and a 'B' represents the actual trajectory data, and it can be seen that there is a deviation between the two. When the robot performs relatively complex motion according to the preset track data, the robot cannot perform motion according to the preset track data well.
Step 12: and determining a track deviation error between the preset track data and the actual track data.
In some embodiments, determining the track offset error between the preset track data and the actual track data is performed by calculating current position information in the actual track data and target position information in the preset track data.
For example, the distance between the current position and the target position in the preset track data is calculated, and the offset displacement is calculated; calculating an included angle between the motion direction of the current position and the motion direction of the target position in the preset track data, and calculating an offset angle; and calculating the difference value between the speed of the current position and the speed of the target position in the preset track data, and calculating the speed amount of the deviation.
Step 13: and inputting the track deviation error into a control model obtained by pre-training, and obtaining track correction data output by the control model.
The control model is obtained by training the control model by taking a track deviation error generated by the robot in a historical motion process and correspondingly determined track correction data as a training sample. Specifically, the historical movement process may be a teaching process of the robot, or may be an operation process of the robot at other times. Both of these processes may produce a trajectory offset error and corresponding trajectory correction data.
In some embodiments, the control model may be established by using a gaussian mixture model, a hidden markov model, K-nearest neighbor, linear regression, a neural network, a support vector machine, or the like.
In some embodiments, the trajectory modification data may include a modification to speed, a modification to direction. Specifically, the velocity may include a linear velocity and an acceleration. The correction of the direction is mainly reflected in the pose change of the robot.
Step 14: and controlling the robot by using the track correction data so as to correct the motion track of the robot.
Specifically, referring to fig. 3, fig. 3 is a schematic flow chart of step 14.
Step 141: and calculating the track correction data to obtain a control instruction of the robot.
In some embodiments, the trajectory correction data may specifically be a speed correction value, and step 141 may be to calculate the speed correction value to obtain a corresponding speed control command of the robot. If the current track is an arc and the robot needs to decelerate, after the position error is detected, the position error is input into the control model to obtain a speed correction value, the corrected speed of the robot is obtained through calculation according to the speed correction value, and a control command is generated based on the speed.
Step 142: and controlling the robot based on the control instruction to correct the motion trail of the robot.
Referring to fig. 4, the present embodiment will be explained:
CD in fig. 4 represents preset track data, and C 'D' represents actual track data. When the robot obtains a motion instruction based on a preset track CD, the robot is located at the position C ', the track deviation error between the position C' and the position C can be calculated, the track deviation error is input into a control model obtained through pre-training, track correction data output by the control model are obtained, and track correction is carried out on the robot based on the track correction data, so that the robot moves according to the preset track CD. In the moving process of the robot, continuously or according to a time interval, acquiring a track deviation error between an actual track and a preset track of the robot, inputting the track deviation error into a control model obtained through pre-training, obtaining track correction data output by the control model, and performing track correction on the robot based on the track correction data, so that the actual track data of the robot is rapidly converged to the preset track data, the actual track data is gradually overlapped with the preset track data, and finally the target point D or D' is reached.
In contrast to the prior art, a robot control method of the present application includes: acquiring actual track data generated by the robot moving based on preset track data; determining a track offset error between the preset track data and the actual track data; inputting the track deviation error into a control model obtained by pre-training, and obtaining track correction data output by the control model; the control model is obtained by taking a track deviation error generated by the robot in the historical motion process and correspondingly determined track correction data as a training sample to train the control model; and controlling the robot by using the track correction data so as to correct the motion track of the robot. By the method, on one hand, the control model is trained through data generated in the historical movement process to correct the real-time track deviation error of the robot, so that the robot can complete the movement of a given track, and stable and accurate complex track tracking is realized; on the other hand can be better utilize the control model to obtain different users' control custom, make the motion characteristic of robot accord with the control characteristic that the user was when carrying out the error correction more to the user passes through historical data, like teaching data, just realizes using the regulation of robot, greatly reduced technique uses the degree of difficulty.
Referring to fig. 5, fig. 5 is a schematic flowchart of a second embodiment of a control method of a robot provided by the present application, the method further including:
step 51: and acquiring track deviation errors generated in the historical motion process of the robot and correspondingly determined track correction data.
It can be understood that the historical motion process of the robot may be a teaching process of the robot, or may be an operation process of the robot at other times. Among them, there are many teaching methods used in the teaching process of the robot, such as teleoperation teaching, drag teaching, and teaching in a virtual simulation environment.
In the historical motion process, the robot generates corresponding motion data, such as the moving speed, the angular speed, the position information of each joint, the position information of the center point of the tool, and the like of the robot. Taking a teaching process as an example, in the process of teaching the robot according to preset track data, a track deviation error exists inevitably, and when the track deviation error occurs, a teaching person who performs teaching operation on the robot sets corresponding track correction data according to the current track deviation error so as to correct the current motion track of the robot and enable the robot to return to the current position of the preset track data. Therefore, a large number of trajectory offset errors and corresponding trajectory correction data exist in the process of teaching the robot according to the preset trajectory data, and a large number of trajectory offset errors and corresponding trajectory correction data can be acquired.
In some embodiments, the robot collects position information of each joint through a sensor of the robot while being dragged for teaching to obtain position information of a plurality of consecutive positions. The sensor may be a robot joint or a motor-side encoder to obtain positional information of the joint.
Step 52: and initializing the track deviation error in the historical motion process and the corresponding determined track correction data to generate a training sample.
In some embodiments, some data is formatted, scaled by units, to meet the needs of the next training model.
In some embodiments, initializing the data information may be to correspondingly classify the trajectory deviation error and the trajectory modification data in the historical movement process to generate a plurality of training samples, form a training sample set, and further label training sample labels.
Step 53: and inputting the training samples into a pre-established control model so as to train the control model.
Step 54: and calculating to obtain the output weight of the control model.
In some embodiments, the control model includes an input layer, a hidden layer, an output layer. After a training sample is input into a pre-established control model, a track offset error passes through an input layer, then the track offset error is multiplied by corresponding weight and added with offset on a hidden layer, all node results are summed through a nonlinear function to obtain a hidden layer output matrix, and then the output weight is obtained through calculation according to the hidden layer output matrix.
Specifically, the output weight is calculated using the following formula:
wherein the content of the first and second substances,
representing trajectory modification data during historical motion;
a hidden layer output matrix is represented which,
representing input weights, biRepresenting the hidden layer element bias, g (-) is the activation function; β represents an output weight;Mindicates the number of e.
Step 55: and when the output weight meets the preset requirement, finishing the training of the control model.
In some embodiments, the parameters in the control model are constrained for system stability considerations. And when the output weight and the input weight meet the first stability constraint condition and the hidden layer unit bias meets the second stability constraint condition, determining the output weight and finishing the training of the control model. Specifically, the first stability constraint is expressed using the following formula:
wherein, betaiRepresenting an output weight; w is aiRepresenting the input weight.
The second stability constraint is expressed using the following formula:
bi＝0；
wherein, biIndicating a hidden layer cell bias.
In some embodiments, the trajectory offset error described above may be re-input into the trained control model to obtain trajectory correction data output by the control model. And comparing the track correction data with the track correction data corresponding to the track offset error during training, and further determining whether the control model meets the use requirement.
Steps 53-55 are explained with reference to FIG. 6:
fig. 6 is a schematic structural diagram of an embodiment of a control model provided in the present application, where an Extreme Learning Machine (ELM) is used to establish the control model in the present embodiment. In the training stage, the algorithm is not a gradient-based algorithm (back propagation) which is frequently tried in the traditional neural network, random input weight and hidden layer unit bias are adopted, and the output weight is obtained by calculation through a generalized inverse matrix theory. After the weights and biases on all network nodes are obtained, training of the Extreme Learning Machine (ELM) is complete. At this time, when the test data is input, the output of the control model can be calculated by using the trained control model so as to complete the prediction of the test data.
In fig. 6, the input layers are sequentially arranged from left to right, where the data input by the input layer is a track offset error, the hidden layer is arranged in the middle, the connection from the input layer to the hidden layer is full, and the output of the hidden layer is denoted as D, so the calculation formula of the output D of the hidden layer is as follows:
the output of the hidden layer is obtained by multiplying the input data (trajectory offset error) of the input layer by the corresponding weight plus the offset, then performing a nonlinear function, and summing the results of all nodes. D is the ELM nonlinear mapping (hidden layer output matrix). The output functions of the hidden layer nodes are not unique, and different output functions may be used for different hidden layer neurons. Wherein g (-) is an activation function, is a nonlinear piecewise continuous function meeting the ELM general approximation capability theorem, and commonly used functions include a Sigmoid function, a Gaussian function and the like. Here, the activation function employed in the present application needs to satisfy a continuously differentiable condition, so the activation function is as follows:
after the hidden layer calculation, the hidden layer output matrix enters the output layer, and then the output of the ELM control model for "generalized" is, according to the above formula:
where β is the output weight between the hidden layer and the output layer. The operation of the ELM control model from input to output is now the calculation process of the above formula. It is noted that the unknowns in the above formula are w, b, β to date. w, b, β are input weights, hidden layer unit bias, and output weights, respectively. The learning (or training) process of the neural network is known to adjust the weights and biases between neurons according to training data, and what is actually learned is included in the connection weights and biases. We next use the mechanism of ELM to solve for these three values (ELM training process).
Basically, the training of ELMs is divided into two main phases: (1) and (2) solving linear parameters of the random feature mapping.
In the first stage, hidden layer parameters are initialized randomly, and then input data is mapped to a new feature space (called ELM feature space) by using some nonlinear mapping functions as activation functions. Simply speaking, the weights and biases on the nodes of the ELM hidden layer are randomly generated. The stochastic feature mapping stage is different from many existing learning algorithms (e.g., SVM (Support Vector Machine) that performs feature mapping using kernel functions, RBM (Restricted Boltzmann Machine) that is used in deep neural networks, and auto-encoders/auto-decoders for feature learning). The non-linear mapping function in the ELM can be any non-linear piecewise continuous function. In ELM, hidden layer node parameters (w and b) are randomly generated (independent of training data) from an arbitrary continuous probability distribution, rather than being determined through training, thereby rendering a significant advantage in efficiency over conventional BP neural networks.
After the first stage of training, w, b have been randomly generated and determined, so that the hidden can be calculated according to the above formulaThe layer outputs a matrix D. In the second phase of ELM learning, we only need to solve the weights β of the output layers. To obtain β with good effect on the training sample set, it is necessary to ensure that its training error is minimal, and we can use the D β (D β is the output of the network) and the input training sample
The minimum squared error is evaluated as an evaluation training error (objective function), so that the solution with the minimum objective function is the optimal solution. Namely, solving the weight beta connecting the hidden layer and the output layer by a method of minimizing approximate square difference, wherein the objective function is as follows:
where D is the output matrix of the hidden layer,
for correcting data for traces in historical motion, training samples are input in the system
The optimal solution can be derived through the knowledge of linear algebra and matrix theory as follows:
wherein
Moore-Penrose generalized inverse matrix which is matrix D.
Further, to analyze the stability of the control system of the robot we designed, and to derive stability constraints for the relevant control parameters, we adopted the Lyapunov stability theory. According to the Lyapunov stability theory, a continuous differentiable Lyapunov function needs to be designed and meets the following conditions:
the global stability of the system can be ensured. Wherein c represents an input state quantity, c*Indicating a stable equilibrium point, i.e., an equilibrium state. We define the state input quantity as
Therefore, the final position and speed errors of the system can be converged to zero, and the Lyapunov function is designed as follows:
its derivative satisfies:
can find out through observation
Can be obtained by derivation of the control model formula that we designed, therefore, the above formula can be changed into by derivation after being brought into the control model formula:
observing the above formula, the derivative term is actually the first derivative of the activation function, i.e.
The above formula can be further simplified.
In conjunction with the median theorem, and g (0) ═ 0, the above formula can be eventually expanded to be written as:
wherein s isiTo representThe state of a certain intermediate point is obtained according to the median theorem and satisfies
Or
To this end, in order to make the derivative of the Lyapunov function negative, we obtain the stability constraint of the following control parameters, which should be satisfied for all hidden layer nodes in the ELM, i.e., i ═ 1,2, …, N:
by analyzing the global stability of the robot control system, stability constraint of control parameters of the designed control model based on the learning strategy is deduced, and then parameters in the design of the control model of the robot are set to be changed into an optimal problem for solving the problem to be constrained, namely:
when the above control model is satisfied during the training process
If so, ending the training of the control model. When the robot generates a track deviation error along the given track, the track deviation error is input into the trained control model, so that the control model outputs track correction data, and the robot is controlled to correct the actual motion track along the given complex track based on the track correction data, so that the robot moves along the given complex track, and the stable and accurate complex track tracking is realized.
In this way, the control model is trained based on the extreme learning machine algorithm, and the control model learns the track offset error in the historical motion process and the change rule of the track correction data determined correspondingly, so that the control characteristics of the robot in the historical motion process by a user are obtained. When the robot moves according to a given complex track, the control model can output track correction data by inputting real-time track deviation errors into the control model. And controlling the robot to correct track deviation errors, such as errors in speed and position, by using the track correction data so as to enable the robot to move along a given complex track. And the stability problem is considered in the control model training process, the system stability of the control model after the training is finished when the control model is applied to the robot is ensured, and the safety factor of the robot in use is improved. Meanwhile, the control model obtains generalization capability by learning the characteristics in the trajectory correction data in the historical movement process, when the control model is used for controlling the robot, no matter how the given trajectory changes, the trajectory correction data output by the control model is more in line with the control characteristics of a user during error correction, and the control characteristics of the user in the historical movement process can be reproduced when the robot performs trajectory correction according to the trajectory correction data. Meanwhile, the track deviation error is input into the control model, so that the control model outputs track correction data, the generalization characteristic of the control model is utilized, the robot has the tracking generalization capability on different complex tracks, the track correction is not completed by frequently adjusting control parameters by a user in other methods, and the use threshold is greatly reduced.
Referring to fig. 7, fig. 7 is a schematic structural diagram of an embodiment of the robot provided in the present application, the robot 70 includes a processor 71 and a memory 72 connected to the processor 71; wherein the memory 72 is used for storing program data and the processor 71 is used for executing the program data, for implementing the following method:
acquiring actual track data generated by the robot moving based on preset track data; determining a track offset error between the preset track data and the actual track data; inputting the track deviation error into a control model obtained by pre-training, and obtaining track correction data output by the control model; the control model is obtained by taking a track deviation error generated by the robot in the historical motion process and correspondingly determined track correction data as a training sample to train the control model; and controlling the robot by using the track correction data so as to correct the motion track of the robot.
It can be understood that, when the processor 71 is used for executing the program data, it is also used for implementing any method of the foregoing embodiments, and specific implementation steps thereof may refer to the foregoing embodiments, which are not described herein again.
Referring to fig. 8, fig. 8 is a schematic structural diagram of an embodiment of a computer-readable storage medium 80 provided in the present application, the computer-readable storage medium 80 is used for storing program data 81, and the program data 81, when being executed by a processor, is used for implementing the following method steps:
acquiring actual track data generated by the robot moving based on preset track data; determining a track offset error between the preset track data and the actual track data; inputting the track deviation error into a control model obtained by pre-training, and obtaining track correction data output by the control model; the control model is obtained by taking a track deviation error generated by the robot in the historical motion process and correspondingly determined track correction data as a training sample to train the control model; and controlling the robot by using the track correction data so as to correct the motion track of the robot.
It is understood that the program data 81, when executed by the processor, may be used to implement any method of the foregoing embodiments, and specific implementation steps thereof may refer to the foregoing embodiments, which are not described herein again.
In the several embodiments provided in the present application, it should be understood that the disclosed method and apparatus may be implemented in other manners. For example, the above-described device embodiments are merely illustrative, and for example, the division of the modules or units is only one logical division, and other divisions may be realized in practice, for example, a plurality of units or components may be combined or integrated into another system, or some features may be omitted, or not executed.
The units described as separate parts may or may not be physically separate, and parts displayed as units may or may not be physical units, may be located in one place, or may be distributed on a plurality of network units. Some or all of the units can be selected according to actual needs to achieve the purpose of the embodiment.
In addition, functional units in the embodiments of the present application may be integrated into one processing unit, or each unit may exist alone physically, or two or more units may be integrated into one unit. The integrated unit can be realized in a form of hardware, and can also be realized in a form of a software functional unit.
The integrated units in the other embodiments described above may be stored in a computer-readable storage medium if they are implemented in the form of software functional units and sold or used as separate products. Based on such understanding, the technical solution of the present application may be substantially implemented or contributed by the prior art, or all or part of the technical solution may be embodied in a software product, which is stored in a storage medium and includes instructions for causing a computer device (which may be a personal computer, a server, a network device, or the like) or a processor (processor) to execute all or part of the steps of the method according to the embodiments of the present application. And the aforementioned storage medium includes: various media capable of storing program codes, such as a usb disk, a removable hard disk, a Read-Only Memory (ROM), a Random Access Memory (RAM), a magnetic disk, or an optical disk.
The above description is only for the purpose of illustrating embodiments of the present application and is not intended to limit the scope of the present application, and all modifications of equivalent structures and equivalent processes, which are made by the contents of the specification and the drawings of the present application or are directly or indirectly applied to other related technical fields, are also included in the scope of the present application.
Claims (10)
1. A method of controlling a robot, the method comprising:
acquiring actual track data generated by the robot moving based on preset track data;
determining a track offset error between the preset track data and the actual track data;
inputting the track deviation error into a control model obtained by pre-training, and obtaining track correction data output by the control model; the control model is obtained by training the control model by taking a track deviation error generated by the robot in a historical motion process and correspondingly determined track correction data as a training sample;
and controlling the robot by using the track correction data so as to correct the motion track of the robot.
2. The method of claim 1,
the method further comprises the following steps:
acquiring a track deviation error generated in the historical movement process of the robot and correspondingly determined track correction data;
initializing the track deviation error in the historical movement process and the correspondingly determined track correction data to generate a training sample;
inputting the training sample into a pre-established control model to train the control model;
calculating to obtain the output weight of the control model;
and when the output weight meets a preset requirement, finishing the training of the control model.
3. The method of claim 2,
the calculating to obtain the output weight of the control model comprises:
calculating a hidden layer output matrix of the control model;
and calculating to obtain the output weight according to the hidden layer output matrix.
4. The method of claim 3,
the calculating the output weight according to the hidden layer output matrix includes:
calculating the output weight using the following formula:
min||DβT-S(uo)||；
wherein, S (u)o) Representing trajectory modification data during historical motion; d represents the hidden layer output matrix; β represents the output weight.
5. The method of claim 2,
when the output weight meets a preset requirement, ending the training of the control model, comprising:
and when the output weight and the input weight meet a first stability constraint condition and the bias of the hidden layer unit meets a second stability constraint condition, determining the output weight and finishing the training of the control model.
6. The method of claim 5,
the first stability constraint is expressed using the following formula:
wherein, betaiRepresenting the output weight; w is aiRepresenting the input weight;
the second stability constraint is expressed using the following formula:
bi＝0；
wherein, biRepresenting the hidden layer cell bias.
7. The method of claim 1,
the calculation formula of the control model is as follows:
wherein the content of the first and second substances,
representing the trajectory modification data; beta is aiAn output weight representing the control model; w is aiInput weights representing the control model; e represents the trajectory offset error, N represents the number of hidden layer neurons;
an activation function representing the control model; biA hidden layer cell bias representing the control model.
8. The method of claim 1,
the controlling the robot by using the trajectory correction data to correct the motion trajectory of the robot includes:
calculating the trajectory correction data to obtain a control instruction of the robot;
and controlling the robot based on the control instruction so as to correct the motion trail of the robot.
9. A robot comprising a processor and a memory coupled to the processor; the memory is for storing program data and the processor is for executing the program data to implement the method of any one of claims 1-8.
10. A computer-readable storage medium, characterized in that the computer-readable storage medium is used for storing program data, which, when being executed by a processor, is used for carrying out the method according to any one of claims 1-8.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010536600.5A CN111890350A (en) | 2020-06-12 | 2020-06-12 | Robot, method of controlling the same, and computer-readable storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010536600.5A CN111890350A (en) | 2020-06-12 | 2020-06-12 | Robot, method of controlling the same, and computer-readable storage medium |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN111890350A true CN111890350A (en) | 2020-11-06 |
Family
ID=73206323
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202010536600.5A Pending CN111890350A (en) | 2020-06-12 | 2020-06-12 | Robot, method of controlling the same, and computer-readable storage medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN111890350A (en) |
Cited By (13)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112258128A (en) * | 2020-11-11 | 2021-01-22 | 拉扎斯网络科技（上海）有限公司 | Target position estimation method, target position estimation device, electronic equipment and computer storage medium |
| CN112947264A (en) * | 2021-04-21 | 2021-06-11 | 苏州希盟科技股份有限公司 | Control method and device for dispenser, electronic equipment and medium |
| CN113070878A (en) * | 2021-03-26 | 2021-07-06 | 中国科学院深圳先进技术研究院 | Robot control method based on impulse neural network, robot and storage medium |
| CN113421470A (en) * | 2020-12-02 | 2021-09-21 | 中国人民解放军63920部队 | Teleoperation simulation training system and teleoperation simulation training method for space manipulator |
| CN113479405A (en) * | 2021-07-19 | 2021-10-08 | 合肥哈工龙延智能装备有限公司 | Control method for stably opening paper box by high-speed box packing machine |
| CN113554757A (en) * | 2021-07-01 | 2021-10-26 | 新疆大学 | Three-dimensional reconstruction method and system for workpiece track based on digital twinning |
| CN113925607A (en) * | 2021-11-12 | 2022-01-14 | 上海微创医疗机器人（集团）股份有限公司 | Operation training method, device, system, medium and equipment for surgical robot |
| CN114029961A (en) * | 2021-12-16 | 2022-02-11 | 滁州学院 | Robot control method and system for high-precision transmission of mechanical arm |
| CN114115242A (en) * | 2021-11-05 | 2022-03-01 | 江苏昱博自动化设备有限公司 | Self-learning positioning control method for storage transfer robot |
| CN115990891A (en) * | 2023-03-23 | 2023-04-21 | 湖南大学 | Robot reinforcement learning assembly method based on visual teaching and virtual-actual migration |
| WO2023082404A1 (en) * | 2021-11-11 | 2023-05-19 | 中国科学院深圳先进技术研究院 | Control method for robot, and robot, storage medium, and grabbing system |
| WO2023093594A1 (en) * | 2021-11-24 | 2023-06-01 | 泉州装备制造研究所 | Floor polishing robot trajectory deviation correction method and apparatus having error correction function |
| WO2023216581A1 (en) * | 2022-05-09 | 2023-11-16 | 深圳先进技术研究院 | Controller training method, trajectory tracking method, terminal device, and storage medium |
Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2017097807A (en) * | 2015-11-27 | 2017-06-01 | 富士通株式会社 | Learning method, learning program, and information processing device |
| CN108115681A (en) * | 2017-11-14 | 2018-06-05 | 深圳先进技术研究院 | Learning by imitation method, apparatus, robot and the storage medium of robot |
| CN108214485A (en) * | 2016-12-13 | 2018-06-29 | 发那科株式会社 | Robot controller and robot control method |
| CN108693829A (en) * | 2017-04-07 | 2018-10-23 | 发那科株式会社 | Control device |
| CN109605365A (en) * | 2018-10-30 | 2019-04-12 | 大连理工大学 | A kind of robot for space arrests control system, intensified learning method and dynamic modeling method |
| CN109773787A (en) * | 2018-12-29 | 2019-05-21 | 南京埃斯顿机器人工程有限公司 | The Schistosomiasis control method of mechanical equipment and its server system for having learning functionality |
| CN110977992A (en) * | 2020-01-02 | 2020-04-10 | 中山大学 | Non-kinematic model trajectory tracking method for mechanical arm and mechanical arm system |
-
2020
- 2020-06-12 CN CN202010536600.5A patent/CN111890350A/en active Pending
Patent Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2017097807A (en) * | 2015-11-27 | 2017-06-01 | 富士通株式会社 | Learning method, learning program, and information processing device |
| CN108214485A (en) * | 2016-12-13 | 2018-06-29 | 发那科株式会社 | Robot controller and robot control method |
| CN108693829A (en) * | 2017-04-07 | 2018-10-23 | 发那科株式会社 | Control device |
| CN108115681A (en) * | 2017-11-14 | 2018-06-05 | 深圳先进技术研究院 | Learning by imitation method, apparatus, robot and the storage medium of robot |
| CN109605365A (en) * | 2018-10-30 | 2019-04-12 | 大连理工大学 | A kind of robot for space arrests control system, intensified learning method and dynamic modeling method |
| CN109773787A (en) * | 2018-12-29 | 2019-05-21 | 南京埃斯顿机器人工程有限公司 | The Schistosomiasis control method of mechanical equipment and its server system for having learning functionality |
| CN110977992A (en) * | 2020-01-02 | 2020-04-10 | 中山大学 | Non-kinematic model trajectory tracking method for mechanical arm and mechanical arm system |
Non-Patent Citations (5)
| Title |
|---|
| R SUBASRI; A. M. NATARAJAN; S SURESH: "Discrete direct adaptive ELM controller for seismically excited non-linear base-isolated buildings", 《2013 IEEE SYMPOSIUM ON COMPUTATIONAL INTELLIGENCE IN CONTROL AND AUTOMATION (CICA)》 * |
| 张昕: "《控制工程基础与信号处理》", 30 September 2018, 北京理工大学出版社 * |
| 王枞: "《控制系统理论及应用》", 31 August 2009, 北京邮电大学出版社 * |
| 王跃新: "《创新思维应用学》", 31 August 2010, 吉林人民出版社 * |
| 魏延辉: "《UVMS系统控制技术》", 30 April 2017, 哈尔滨工程大学出版社 * |
Cited By (16)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112258128A (en) * | 2020-11-11 | 2021-01-22 | 拉扎斯网络科技（上海）有限公司 | Target position estimation method, target position estimation device, electronic equipment and computer storage medium |
| CN113421470A (en) * | 2020-12-02 | 2021-09-21 | 中国人民解放军63920部队 | Teleoperation simulation training system and teleoperation simulation training method for space manipulator |
| CN113070878A (en) * | 2021-03-26 | 2021-07-06 | 中国科学院深圳先进技术研究院 | Robot control method based on impulse neural network, robot and storage medium |
| WO2022199146A1 (en) * | 2021-03-26 | 2022-09-29 | 中国科学院深圳先进技术研究院 | Robot control method and robot based on spiking neural network, and storage medium |
| CN112947264A (en) * | 2021-04-21 | 2021-06-11 | 苏州希盟科技股份有限公司 | Control method and device for dispenser, electronic equipment and medium |
| CN113554757A (en) * | 2021-07-01 | 2021-10-26 | 新疆大学 | Three-dimensional reconstruction method and system for workpiece track based on digital twinning |
| CN113479405A (en) * | 2021-07-19 | 2021-10-08 | 合肥哈工龙延智能装备有限公司 | Control method for stably opening paper box by high-speed box packing machine |
| CN114115242A (en) * | 2021-11-05 | 2022-03-01 | 江苏昱博自动化设备有限公司 | Self-learning positioning control method for storage transfer robot |
| CN114115242B (en) * | 2021-11-05 | 2023-06-20 | 江苏昱博自动化设备有限公司 | Self-learning positioning control method of storage transfer robot |
| WO2023082404A1 (en) * | 2021-11-11 | 2023-05-19 | 中国科学院深圳先进技术研究院 | Control method for robot, and robot, storage medium, and grabbing system |
| CN113925607A (en) * | 2021-11-12 | 2022-01-14 | 上海微创医疗机器人（集团）股份有限公司 | Operation training method, device, system, medium and equipment for surgical robot |
| CN113925607B (en) * | 2021-11-12 | 2024-02-27 | 上海微创医疗机器人（集团）股份有限公司 | Operation robot operation training method, device, system, medium and equipment |
| WO2023093594A1 (en) * | 2021-11-24 | 2023-06-01 | 泉州装备制造研究所 | Floor polishing robot trajectory deviation correction method and apparatus having error correction function |
| CN114029961A (en) * | 2021-12-16 | 2022-02-11 | 滁州学院 | Robot control method and system for high-precision transmission of mechanical arm |
| WO2023216581A1 (en) * | 2022-05-09 | 2023-11-16 | 深圳先进技术研究院 | Controller training method, trajectory tracking method, terminal device, and storage medium |
| CN115990891A (en) * | 2023-03-23 | 2023-04-21 | 湖南大学 | Robot reinforcement learning assembly method based on visual teaching and virtual-actual migration |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111890350A (en) | Robot, method of controlling the same, and computer-readable storage medium | |
| CN111890351A (en) | Robot, method of controlling the same, and computer-readable storage medium | |
| CN109885883B (en) | Unmanned vehicle transverse motion control method based on GK clustering algorithm model prediction | |
| Cochran et al. | Source seeking for two nonholonomic models of fish locomotion | |
| Wang et al. | A new trajectory-planning beetle swarm optimization algorithm for trajectory planning of robot manipulators | |
| CN109702740B (en) | Robot compliance control method, device, equipment and storage medium | |
| Xu et al. | Robot trajectory tracking control using learning from demonstration method | |
| CN116460860B (en) | Model-based robot offline reinforcement learning control method | |
| CN113671962B (en) | Transverse control method of unmanned articulated sweeper | |
| Duan et al. | Sim-to-real learning of footstep-constrained bipedal dynamic walking | |
| Lu et al. | Human-robot collaboration using variable admittance control and human intention prediction | |
| Sun | Kinematics model identification and motion control of robot based on fast learning neural network | |
| Kimmel et al. | Belief-space planning using learned models with application to underactuated hands | |
| Hwang et al. | A fuzzy CMAC learning approach to image based visual servoing system | |
| Vallon et al. | Data-driven strategies for hierarchical predictive control in unknown environments | |
| Liu et al. | Modeling and control of robotic manipulators based on artificial neural networks: a review | |
| Jangid et al. | Kinematics-based end-effector path control of a mobile manipulator system on an uneven terrain using a two-stage Support Vector Machine | |
| Dong et al. | Friction modeling and compensation for haptic master manipulator based on deep Gaussian process | |
| Wu et al. | Learning robot anomaly recovery skills from multiple time-driven demonstrations | |
| Wang et al. | An online collision-free trajectory generation algorithm for human–robot collaboration | |
| Joukov et al. | Gaussian process based model predictive controller for imitation learning | |
| CN110977965B (en) | Robot, method of controlling the same, and computer storage medium | |
| CN115488881A (en) | Man-machine sharing autonomous teleoperation method and system based on multi-motor skill prior | |
| Hart et al. | Generalization and transfer in robot control | |
| CN115344047A (en) | Robot switching type predictive control trajectory tracking method based on neural network model |
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 |