CN114905515B - Robot control method and system based on flexible perception neural network - Google Patents
Robot control method and system based on flexible perception neural network Download PDFInfo
- Publication number
- CN114905515B CN114905515B CN202210588502.5A CN202210588502A CN114905515B CN 114905515 B CN114905515 B CN 114905515B CN 202210588502 A CN202210588502 A CN 202210588502A CN 114905515 B CN114905515 B CN 114905515B
- Authority
- CN
- China
- Prior art keywords
- workpiece
- polishing
- processed
- information
- analysis
- 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
-
- 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
- B25J9/1679—Programme controls characterised by the tasks executed
-
- B—PERFORMING OPERATIONS; TRANSPORTING
- B25—HAND TOOLS; PORTABLE POWER-DRIVEN TOOLS; MANIPULATORS
- B25J—MANIPULATORS; CHAMBERS PROVIDED WITH MANIPULATION DEVICES
- B25J11/00—Manipulators not otherwise provided for
- B25J11/005—Manipulators for mechanical processing tasks
- B25J11/0065—Polishing or grinding
-
- 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
- B25J9/1633—Programme controls characterised by the control loop compliant, force, torque control, e.g. combined with position control
-
- 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
- B25J9/1694—Programme controls characterised by use of sensors other than normal servo-feedback from position, speed or acceleration sensors, perception control, multi-sensor controlled systems, sensor fusion
- B25J9/1697—Vision controlled systems
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02P—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN THE PRODUCTION OR PROCESSING OF GOODS
- Y02P90/00—Enabling technologies with a potential contribution to greenhouse gas [GHG] emissions mitigation
- Y02P90/02—Total factory control, e.g. smart factories, flexible manufacturing systems [FMS] or integrated manufacturing systems [IMS]
Landscapes
- Engineering & Computer Science (AREA)
- Robotics (AREA)
- Mechanical Engineering (AREA)
- Manipulator (AREA)
- Finish Polishing, Edge Sharpening, And Grinding By Specific Grinding Devices (AREA)
- Constituent Portions Of Griding Lathes, Driving, Sensing And Control (AREA)
Abstract
The invention discloses a robot control method and a system based on a flexible perception neural network, and relates to the technical field of artificial intelligence, wherein the method comprises the following steps: obtaining image information of the workpiece to be processed through the visual information acquisition module; inputting the target polishing standard information and the image information of the workpiece to be processed into a workpiece polishing analysis model to obtain a workpiece polishing scheme to be processed; controlling a flexible polishing robot to polish the workpiece according to a polishing scheme of the workpiece to be processed through a motion trail control module; according to the workpiece polishing force information, motion force compensation information is obtained; and correcting the polishing scheme of the workpiece to be processed through analysis of the motion force compensation information and the polishing pose space information by a flexible sensing network. The polishing force compensation is carried out on the robot in any posture through real-time interaction with the flexible polishing robot, so that the polishing force compensation of the robot in any posture is realized, the polishing force of a workpiece is ensured, the polishing precision of the workpiece is improved, and the technical effect of polishing the quality of the workpiece is further ensured.
Description
Technical Field
The invention relates to the field of artificial intelligence, in particular to a robot control method and system based on a flexible perception neural network.
Background
Workpiece polishing is a processing method for changing physical properties of a material surface by friction, and is widely applied to the fields of machinery manufacturing industry, processing industry, mould industry, wood industry, leather industry and the like. The polishing work aims at obtaining the specific surface roughness of the workpiece, removing burrs on the surface of the workpiece of the product, enabling the workpiece to be smooth, and enabling the workpiece to be easy to continue to process or to reach a finished product, and under the situation that the application of robots is becoming popular, the application of the robots in the field of workpiece polishing is becoming widespread, and the robot has the advantages of being high in polishing efficiency and high in intelligent degree.
However, the polishing control precision of the polishing robot in the prior art in the operation process is not high, so that the technical problem that the quality of polished workpieces does not reach the standard is caused.
Disclosure of Invention
According to the robot control method and system based on the flexible perception neural network, the technical problem that the quality of a polished workpiece does not reach the standard due to low polishing control precision of a polishing robot in the operation process in the prior art is solved, real-time interaction between the machine vision acquisition information and the polishing force acquisition information and the flexible polishing robot is achieved, polishing force compensation is carried out on the robot in any posture, polishing force of the workpiece is ensured, polishing precision of the workpiece is improved, and further technical effect of polishing the quality of the workpiece is guaranteed.
In view of the above problems, the invention provides a robot control method and system based on a flexible perception neural network.
In a first aspect, the present application provides a robot control method based on a flexible sensory neural network, the method comprising: image acquisition is carried out on the workpiece to be processed through a visual information acquisition module, so that image information of the workpiece to be processed is obtained; uploading the image information of the workpiece to be processed to a polishing robot control system based on a data interaction module for interaction analysis; the method comprises the steps that target polishing standard information is obtained, the polishing robot control system inputs the target polishing standard information and the image information of a workpiece to be processed into a workpiece polishing analysis model suitable for the workpiece polishing analysis, and a model output result is obtained, wherein the model output result comprises a workpiece polishing scheme to be processed; the workpiece polishing scheme to be processed is sent to the flexible polishing robot through the data interaction module, and the flexible polishing robot is controlled to polish the workpiece according to the workpiece polishing scheme to be processed through the motion trail control module; acquiring workpiece polishing force information in the workpiece polishing process through a force axis sensor module, and performing force compensation analysis according to the workpiece polishing force information to acquire action force compensation information; and capturing the gesture according to the visual information acquisition module to obtain polishing gesture space information, analyzing the motion force compensation information and the polishing gesture space information through a flexible sensing network, and correcting the polishing track of the polishing scheme of the workpiece to be processed.
In another aspect, the present application further provides a robot control system based on a flexible sensory neural network, the system including: the robot forming module is used for the flexible polishing robot and comprises a visual information acquisition module, a force axis sensor module, a data interaction module and a motion trail control module; the visual information acquisition module is used for acquiring images of the workpiece to be processed through the visual information acquisition module to obtain image information of the workpiece to be processed; the data interaction module is used for uploading the image information of the workpiece to be processed to a polishing robot control system for interaction analysis based on the data interaction module; the model output module is used for acquiring target polishing standard information, the polishing robot control system inputs the target polishing standard information and the image information of the workpiece to be processed into an applicable workpiece polishing analysis model to obtain a model output result, and the model output result comprises a workpiece polishing scheme to be processed; the motion trail control module is used for sending the workpiece polishing scheme to be processed to the flexible polishing robot through the data interaction module, and controlling the flexible polishing robot to polish the workpiece according to the workpiece polishing scheme to be processed through the motion trail control module; the force axis sensor module is used for acquiring workpiece polishing force information in the workpiece polishing process through the force axis sensor module, and performing force compensation analysis according to the workpiece polishing force information to acquire action force compensation information; and the polishing track correction module is used for capturing the gesture according to the visual information acquisition module to obtain polishing gesture space information, and analyzing the motion force compensation information and the polishing gesture space information through a flexible sensing network to correct the polishing track of the workpiece to be processed in a polishing scheme.
One or more technical solutions provided in the present application have at least the following technical effects or advantages:
the visual information acquisition module is used for acquiring images of the workpiece to be processed, the image information of the workpiece to be processed is uploaded to the polishing robot control system for interactive analysis based on the data interaction module, and target polishing standard information and the image information of the workpiece to be processed are input into the applicable workpiece polishing analysis model to obtain a model output result, namely a workpiece polishing scheme to be processed; the flexible polishing robot is controlled by the motion track control module to polish the workpiece according to the polishing scheme of the workpiece to be processed, meanwhile, the force axis sensor module is used for acquiring polishing force information of the workpiece in the polishing process of the workpiece, force compensation analysis is carried out, action force compensation information is obtained, meanwhile, gesture capturing is carried out according to the visual information acquisition module, polishing pose space information is obtained, and the action force compensation information and the polishing pose space information are analyzed through the flexible sensing network to correct the polishing track of the polishing scheme of the workpiece to be processed. And then reach and gather information and the dynamics of polishing and gather information and carry out real-time interaction with flexible polishing robot through the machine vision, realize carrying out the dynamics of polishing to the robot of arbitrary gesture and compensate, ensure the dynamics of polishing of work piece, improve the work piece precision of polishing, and then guarantee the technological effect of polishing work piece quality.
Drawings
Fig. 1 is a schematic flow chart of a robot control method based on a flexible sensing neural network;
FIG. 2 is a schematic flow chart of obtaining standard image information of a workpiece to be processed in a robot control method based on a flexible perception neural network;
FIG. 3 is a schematic flow chart of a robot control method based on a flexible sensory neural network for outputting a polishing scheme of a workpiece to be processed;
FIG. 4 is a schematic structural diagram of a robot control system based on a flexible sensory neural network according to the present application;
reference numerals illustrate: the robot comprises a robot constitution module 11, a visual information acquisition module 12, a data interaction module 13, a model output module 14, a motion track control module 15, a force axis sensor module 16 and a polishing track correction module 17.
Detailed Description
The present application is described below with reference to the drawings in the present application.
Example 1
As shown in fig. 1, the present application provides a robot control method based on a flexible sensory neural network, the method being applied to a polishing robot control system, the system being communicatively connected to a flexible polishing robot, the method comprising:
step S100: the flexible polishing robot comprises a visual information acquisition module, a force axis sensor module, a data interaction module and a motion trail control module;
specifically, workpiece polishing is a processing method of surface modification technology, which changes physical properties of a material surface by friction, and is widely used in fields of machine manufacturing industry, processing industry, mold industry, wood industry, leather industry and the like. The polishing work aims at obtaining the specific surface roughness of the workpiece, removing burrs on the surface of the workpiece of the product, enabling the workpiece to be smooth, and enabling the workpiece to be easy to continue to process or to reach a finished product, and under the situation that the application of robots is becoming popular, the application of the robots in the field of workpiece polishing is becoming widespread, and the robot has the advantages of being high in polishing efficiency and high in intelligent degree.
Work piece is polished and is realized through flexible polishing robot, and flexible polishing robot body end axle fuses with flexible perception neural network, and wrist position has unique shrink design, improves the rigidity, effectively avoids the shake that the developments polished in-process appears, has the advantage of polishing extremely, including visual information acquisition module, force axis sensor module, data interaction module and motion trail control module, through the mutual cooperation control between each module, accomplishes the meticulous polishing to the work piece.
Step S200: the visual information acquisition module is used for acquiring images of the workpiece to be processed to obtain image information of the workpiece to be processed;
specifically, the visual information acquisition module is used for acquiring images of the workpiece to be processed, namely the optical imaging module, and the visual information acquisition module can be used for identifying and acquiring images of the workpiece to be processed by adopting a camera or a camera, and comprises information such as the size of the workpiece, the structure of the workpiece, the color of the workpiece, burrs of the workpiece and the like.
Step S300: uploading the image information of the workpiece to be processed to the polishing robot control system based on the data interaction module for interaction analysis;
specifically, the image information of the workpiece to be processed is uploaded to the polishing robot control system for interactive analysis based on the data interaction module, the data interaction module is used for carrying out data transmission interaction between the flexible polishing robot and the polishing robot control system, and the polishing robot control system is used for carrying out polishing scheme analysis and formulation based on the visual information of the workpiece.
Step S400: the method comprises the steps that target polishing standard information is obtained, the polishing robot control system inputs the target polishing standard information and the image information of a workpiece to be processed into a workpiece polishing analysis model suitable for the workpiece polishing analysis, and a model output result is obtained, wherein the model output result comprises a workpiece polishing scheme to be processed;
as shown in fig. 2, further, before the target polishing standard information and the workpiece image information to be processed are input into the applicable workpiece polishing analysis model, step S400 of the present application further includes:
step S410: denoising and filtering the image information of the workpiece to be processed based on an image filtering algorithm to obtain denoising image information of the workpiece to be processed;
step S420: carrying out gray level equalization processing on the denoising workpiece image information to be processed to obtain average workpiece image information to be processed;
step S430: and carrying out image enhancement on the image information of the workpiece to be processed with the average value based on the generated countermeasure network, and then carrying out image grid segmentation on the enhanced image to obtain the image information of the standard workpiece to be processed.
Specifically, target polishing standard information of the workpiece to be processed is obtained, wherein the target polishing standard information is the surface smoothness and surface curvature requirement for polishing the workpiece. The polishing robot control system inputs the target polishing standard information and the workpiece image information to be processed into an applicable workpiece polishing analysis model, and the applicable workpiece polishing analysis model is used for making a workpiece polishing scheme. Before the target polishing standard information and the workpiece image information to be processed are input into a suitable workpiece polishing analysis model, denoising and filtering are carried out on the workpiece image information to be processed based on an image filtering algorithm, namely, noise of a target image is suppressed under the condition that the detail characteristics of the image are reserved as much as possible, the method is an indispensable operation in image preprocessing, and the effectiveness and reliability of subsequent image processing and analysis are directly affected by the processing effect. The common image filtering algorithm comprises nonlinear filtering, median filtering, wavelet filtering, bilateral filtering and the like, and the image information of the workpiece to be processed after denoising is obtained.
And then carrying out gray level equalization treatment on the denoising workpiece image information, namely image gray level histogram equalization, which is a simple and effective image enhancement technology, and changing the gray level of each pixel in the image by changing the histogram of the image, wherein the method is mainly used for enhancing the contrast of the image with smaller dynamic range and obtaining the image information of the workpiece to be processed after the equalization. Image enhancement is carried out on the image information of the workpiece to be processed with the mean value based on the generation countermeasure network, namely useful information in the image is enhanced, the visual effect of the image is improved, and the whole or partial characteristics of the image are purposefully emphasized for the application occasion of the given image.
And then, carrying out image grid segmentation on the enhanced image, wherein the finer the grid segmentation is, the more detailed the image detail characteristics are, the better the effect is, and the preprocessed standard workpiece image information to be processed is obtained. Inputting standard workpiece image information to be processed and the target polishing standard information into a workpiece polishing analysis model to obtain a model output result, wherein the model output result comprises a workpiece polishing scheme to be processed, and the workpiece polishing scheme to be processed comprises workpiece polishing motion track, polishing speed, polishing force and other information. The image application effect is improved by denoising, equalizing and image enhancement preprocessing on the workpiece collected image, so that the image is standardized, and the accuracy of a polishing scheme of the workpiece to be processed is further improved.
Step S500: the workpiece polishing scheme to be processed is sent to the flexible polishing robot through the data interaction module, and the flexible polishing robot is controlled to polish the workpiece according to the workpiece polishing scheme to be processed through the motion trail control module;
step S600: acquiring workpiece polishing force information in the workpiece polishing process through the force axis sensor module, and performing force compensation analysis according to the workpiece polishing force information to acquire action force compensation information;
specifically, the workpiece polishing scheme to be processed is sent to the flexible polishing robot through the data interaction module, and the flexible polishing robot is controlled to polish the workpiece according to the workpiece polishing scheme to be processed through the motion trail control module. And in the polishing process, polishing force information of the workpiece in each force axis direction in the workpiece polishing process is acquired through the force axis sensor module. Because the dead weight of the robot and the gesture influence in the polishing process possibly cause insufficient polishing force, force compensation analysis is carried out according to the workpiece polishing force information, and action force compensation information needing to be subjected to operation force compensation is obtained.
Step S700: and capturing the gesture according to the visual information acquisition module to obtain polishing gesture space information, analyzing the motion force compensation information and the polishing gesture space information through a flexible sensing network, and correcting the polishing track of the polishing scheme of the workpiece to be processed.
Specifically, gesture capturing is performed according to the visual information acquisition module, and polishing pose space information, namely current position and pose information of the flexible polishing robot, is obtained. Analyzing the motion force compensation information and the polishing pose space information through a flexible sensing network, wherein the flexible sensing network interacts with a tail end shaft of a flexible robot. And correcting the polishing track of the polishing scheme of the workpiece to be processed through the analysis result, improving the force feedback response speed, realizing polishing force compensation for the robot with any gesture, ensuring the polishing force of the workpiece, improving the polishing precision of the workpiece, and further ensuring the quality of the polished workpiece.
As shown in fig. 3, further, step S430 of the present application further includes:
step S431: classifying the basic information of the workpiece to be processed through a polished workpiece feature decision tree to obtain polished workpiece classification feature information;
step S432: calibrating the workpiece to be processed according to the classifying characteristic information of the polished workpiece to determine the calibrating parameters of the polished workpiece;
step S433: calling the applicable workpiece polishing analysis model from a workpiece polishing analysis model library based on the polishing workpiece calibration parameters;
step S434: and inputting the target polishing standard information and the standard workpiece image information to be processed into the applicable workpiece polishing analysis model, and outputting the workpiece polishing scheme to be processed.
Specifically, basic information of the workpiece to be processed is classified through a polished workpiece feature decision tree, the polished workpiece feature decision tree is constructed through workpiece feature information, polished workpiece classification feature information is obtained, and the polished workpiece classification feature information is a workpiece feature classification result and comprises workpiece dimension specifications, workpiece types, workpiece materials and the like. The polishing scheme analysis modes are different for workpieces with different characteristic categories, so that the polishing workpiece calibration parameters are determined according to the polishing workpiece classification characteristic information, and the polishing workpiece calibration parameters are used for dividing model selection when polishing analysis is carried out on the workpieces.
And calling the applicable workpiece polishing analysis model from a workpiece polishing analysis model library based on the polishing workpiece calibration parameters, wherein the applicable workpiece polishing analysis model is a polishing scheme analysis model applicable to the workpiece. And inputting the target polishing standard information and the standard workpiece image information to be processed into the applicable workpiece polishing analysis model, and outputting the workpiece polishing scheme to be processed. By classifying and identifying the polished workpieces, the personalized call is applicable to the polished analysis model of the workpieces, and the accuracy and rationality of the output result of the model are improved.
Further, step S434 of the present application further includes:
step S4341: the applicable workpiece polishing analysis model comprises an input layer, a convolution image network layer, a characteristic analysis logic layer and an output layer;
step S4342: the target polishing standard information and the standard workpiece image information to be processed are used as input layers and input into the convolution image network layer, and feature information of the workpiece to be processed is obtained;
step S4343: inputting the target polishing standard information and the workpiece characteristic information to be processed into the characteristic analysis logic layer to obtain a polishing scheme of the workpiece to be processed;
step S4344: and outputting the polishing scheme of the workpiece to be processed as an output result through the output layer.
Specifically, the applicable workpiece polishing analysis model comprises an input layer, a convolution image network layer, a feature analysis logic layer and an output layer, wherein the target polishing standard information and the standard workpiece image information to be processed are used as the input layer and input into the convolution image network layer. The convolution image network layer is used for extracting features of the workpiece image to obtain feature information of the workpiece to be processed, wherein the feature information of the workpiece to be processed comprises workpiece flatness features, curvature features, hole defect features and the like.
And inputting the target polishing standard information and the workpiece characteristic information to be processed into the characteristic analysis logic layer, wherein the characteristic analysis logic layer is used for carrying out matching analysis on the target polishing standard information and the workpiece characteristic information to be processed, and obtaining the workpiece polishing scheme to be processed output by the logic layer. And outputting the workpiece polishing scheme to be processed through the output layer as an output result, and analyzing and formulating the workpiece polishing scheme to be processed by constructing a multi-level applicable workpiece polishing analysis model, thereby improving scheme formulation accuracy and formulation efficiency.
Further, the step S4342 of obtaining the feature information of the workpiece to be processed further includes:
step S43421: obtaining a preset convolution feature set according to the target polishing standard information, wherein the preset convolution feature set comprises workpiece flatness features, curvature features and hole defect features;
step S43422: inputting the standard workpiece image information to be processed into the convolution image network layer as input information to perform feature extraction;
step S43423: and obtaining output information of the convolution image network layer, wherein the output information comprises the workpiece feature information to be processed conforming to the preset convolution feature set.
Specifically, a preset convolution feature set is obtained according to the target polishing standard information, wherein the preset convolution feature set is a preset feature standard meeting workpiece polishing requirements and comprises workpiece flatness features, curvature features and hole defect features. And taking the required workpiece flatness characteristics, curvature characteristics and hole defect characteristics as a preset convolution characteristic set, which is equivalent to extracting image characteristic values. The standard workpiece image information to be processed is used as input information to be input into the convolution image network layer for feature extraction, the convolution image network layer is a deep feedforward neural network with the characteristics of local connection, weight sharing and the like, and remarkable effects are achieved on various visual tasks in the fields of image and video analysis, such as image classification, target detection, image segmentation and the like.
The convolutional image network layer comprises two parts, convolution and neural network, wherein the convolution is a feature extractor, and the neural network can be regarded as a classifier. And obtaining output information of the convolution image network layer, wherein the output information comprises the workpiece feature information to be processed conforming to the preset convolution feature set. The technical effects of carrying out matching analysis on standard product production image information in a convolution extraction mode, enabling the feature matching result of the workpiece to be processed to be more accurate and further improving the making accuracy of the polishing scheme of the workpiece to be processed are achieved.
Further, the steps of the present application further include:
step S810: performing analysis effect verification on the applicable workpiece polishing analysis model to obtain model analysis accuracy;
step S820: if the model analysis accuracy does not reach the preset analysis accuracy, obtaining model analysis deviation based on the difference value between the model analysis accuracy and the preset analysis accuracy;
step S830: and based on a PSO algorithm and the model analysis deviation degree, iteratively updating the applicable workpiece polishing analysis model to obtain the applicable workpiece polishing optimization analysis model.
And the model analysis accuracy is obtained by verifying the analysis effect of the applicable workpiece polishing analysis model, namely verifying the model polishing analysis accuracy, wherein the model analysis accuracy indicates the model analysis accuracy. If the model analysis accuracy does not reach the preset analysis accuracy, namely the training output accuracy of the workpiece polishing analysis model does not reach the standard, taking the difference value between the model analysis accuracy and the preset analysis accuracy as the model analysis deviation degree, namely the model analysis accuracy which needs to be optimized, wherein the greater the deviation degree, the lower the making accuracy of the workpiece polishing analysis scheme.
Because the fitting degree of the applicable workpiece polishing analysis model is low, the fitting degree cannot be suitable for analysis of the current workpiece polishing scheme, and the applicable workpiece polishing analysis model is iteratively updated based on a PSO algorithm and the model analysis deviation degree. The PSO algorithm, namely the particle swarm optimization algorithm, is a random optimization algorithm based on a population, can simulate and iterate continuously until the state is balanced or optimal, and stores the state in balance or optimal state to obtain a workpiece polishing optimization analysis model after the PSO algorithm is updated and optimized. The model is optimized through the PSO algorithm, so that the output deviation of the model is reduced, the accuracy and efficiency of the output result of the model are improved, and the accuracy of workpiece polishing analysis is further improved.
Further, the step S830 of the present application further includes:
step S831: constructing a particle optimization space according to the model training parameters of the applicable workpiece polishing analysis model;
step S832: initializing the particle optimization space to obtain particle swarm constraint parameters, and iteratively calculating a particle swarm fitness function according to the model analysis deviation degree and the particle swarm constraint parameters;
step S833: when a preset termination condition is reached, a first output result of the particle swarm fitness function is obtained, wherein the first output result comprises optimal result particles;
step S834: and mapping the optimal result particles to the applicable workpiece polishing analysis model for optimization training to obtain the applicable workpiece polishing optimization analysis model.
Specifically, the model training parameters of the applicable workpiece polishing and optimizing analysis model are model parameter training dimensions, the model training dimensions comprise information such as model training weights, the number of hidden layers, the number of hidden layer nodes and the like, the particle optimization space is a virtual space for optimizing the applicable workpiece polishing and optimizing analysis model, and the particle optimization space is a multidimensional virtual space, and the space dimensions are the same as the model training parameter dimensions of the applicable workpiece polishing and optimizing analysis model. Initializing the particle optimization space, wherein the particle swarm constraint parameters are model analysis accuracy ranges, iteratively calculating a particle swarm fitness function according to the model analysis deviation degree and the particle swarm constraint parameters, further updating the positions and the speeds of particles in the particle swarm, inputting all the particles into a model for training, evaluating the quality of the particles by calculating the fitness function of the particle swarm, and adjusting the positions and the speeds of each particle by the fitness function.
And when a preset termination condition is reached, obtaining a first output result of the particle swarm fitness function, wherein the first output result comprises optimal result particles. Further, the PSO algorithm stops including two possibilities, one is that the particles are in an equilibrium or optimal state, the other is that the operation limit is exceeded, no specific analysis is carried out on the condition that the operation limit is exceeded, and the optimal result particles are the optimal states of the particles; and mapping the optimal result particles to the applicable workpiece polishing analysis model for optimization training. The output accuracy of the optimized and trained applicable workpiece polishing analysis model is improved, the optimized and trained applicable workpiece polishing analysis model is subjected to particle swarm optimization algorithm, so that the output deviation of the model is reduced, the accuracy and efficiency of the output result of the model are improved, and the accuracy of workpiece polishing analysis is further improved.
In summary, the robot control method and system based on the flexible perception neural network provided by the application have the following technical effects:
the visual information acquisition module is used for acquiring images of the workpiece to be processed, the image information of the workpiece to be processed is uploaded to the polishing robot control system for interactive analysis based on the data interaction module, and target polishing standard information and the image information of the workpiece to be processed are input into the applicable workpiece polishing analysis model to obtain a model output result, namely a workpiece polishing scheme to be processed; the flexible polishing robot is controlled by the motion track control module to polish the workpiece according to the polishing scheme of the workpiece to be processed, meanwhile, the force axis sensor module is used for acquiring polishing force information of the workpiece in the polishing process of the workpiece, force compensation analysis is carried out, action force compensation information is obtained, meanwhile, gesture capturing is carried out according to the visual information acquisition module, polishing pose space information is obtained, and the action force compensation information and the polishing pose space information are analyzed through the flexible sensing network to correct the polishing track of the polishing scheme of the workpiece to be processed. And then reach and gather information and the dynamics of polishing and gather information and carry out real-time interaction with flexible polishing robot through the machine vision, realize carrying out the dynamics of polishing to the robot of arbitrary gesture and compensate, ensure the dynamics of polishing of work piece, improve the work piece precision of polishing, and then guarantee the technological effect of polishing work piece quality.
Example two
Based on the same inventive concept as the robot control method based on the flexible sensing neural network in the foregoing embodiment, the present invention further provides a robot control system based on the flexible sensing neural network, as shown in fig. 4, where the system includes:
the robot forming module 11 is used for the flexible polishing robot and comprises a visual information acquisition module, a force axis sensor module, a data interaction module and a motion trail control module;
the visual information acquisition module 12 is used for acquiring images of the workpiece to be processed through the visual information acquisition module to obtain image information of the workpiece to be processed;
the data interaction module 13 is used for uploading the image information of the workpiece to be processed to a polishing robot control system for interaction analysis based on the data interaction module;
the model output module 14 is used for acquiring target polishing standard information, the polishing robot control system inputs the target polishing standard information and the image information of the workpiece to be processed into an applicable workpiece polishing analysis model to obtain a model output result, and the model output result comprises a workpiece polishing scheme to be processed;
the motion trail control module 15 is used for sending the workpiece polishing scheme to be processed to the flexible polishing robot through the data interaction module, and controlling the flexible polishing robot to polish the workpiece according to the workpiece polishing scheme to be processed through the motion trail control module;
the force axis sensor module 16 is used for acquiring workpiece polishing force information in the workpiece polishing process through the force axis sensor module, and performing force compensation analysis according to the workpiece polishing force information to acquire action force compensation information;
and the polishing track correction module 17 is used for capturing the gesture according to the visual information acquisition module to obtain polishing gesture space information, and analyzing the motion force compensation information and the polishing gesture space information through a flexible sensing network to correct the polishing track of the workpiece to be processed in a polishing scheme.
Further, the model output module further includes:
the denoising filter unit is used for denoising and filtering the image information of the workpiece to be processed based on an image filtering algorithm to obtain denoising image information of the workpiece to be processed;
the equalization processing unit is used for carrying out gray level equalization processing on the denoising workpiece image information to be processed to obtain average workpiece image information to be processed;
the image enhancement unit is used for carrying out image enhancement on the image information of the workpiece to be processed with the average value based on the generation countermeasure network, and then carrying out image grid segmentation on the enhanced image to obtain the image information of the standard workpiece to be processed.
Further, the system further comprises:
the feature classification unit is used for classifying the basic information of the workpiece to be processed through a polished workpiece feature decision tree to obtain polished workpiece classification feature information;
the parameter calibration unit is used for calibrating the workpiece to be processed according to the classification characteristic information of the polished workpiece to determine the calibration parameters of the polished workpiece;
the model calling unit is used for calling the applicable workpiece polishing analysis model from a workpiece polishing analysis model library based on the polishing workpiece calibration parameters;
the model output unit is used for inputting the target polishing standard information and the standard workpiece image information to be processed into the applicable workpiece polishing analysis model and outputting the workpiece polishing scheme to be processed.
Further, the model output unit further includes:
the model forming unit is used for the applicable workpiece polishing analysis model and comprises an input layer, a convolution image network layer, a characteristic analysis logic layer and an output layer;
the convolution network layer unit is used for taking the target polishing standard information and the standard workpiece image information to be processed as input layers, inputting the target polishing standard information and the standard workpiece image information to be processed into the convolution image network layer, and obtaining workpiece characteristic information to be processed;
the characteristic analysis logic layer unit is used for inputting the target polishing standard information and the workpiece characteristic information to be processed into the characteristic analysis logic layer to obtain a workpiece polishing scheme to be processed;
and the output layer unit is used for outputting the polishing scheme of the workpiece to be processed as an output result through the output layer.
Further, the convolutional network layer unit further includes:
the convolution characteristic determining unit is used for obtaining a preset convolution characteristic set according to the target polishing standard information, wherein the preset convolution characteristic set comprises workpiece flatness characteristics, curvature characteristics and hole defect characteristics;
the convolution feature extraction unit is used for inputting the standard workpiece image information to be processed into the convolution image network layer as input information to perform feature extraction;
the convolution network output unit is used for obtaining output information of the convolution image network layer, and the output information comprises the workpiece feature information to be processed which accords with the preset convolution feature set.
Further, the system further comprises:
the model verification unit is used for verifying the analysis effect of the applicable workpiece polishing analysis model to obtain model analysis accuracy;
a model deviation degree analysis unit, configured to obtain a model analysis deviation degree based on a difference between the model analysis accuracy and the preset analysis accuracy if the model analysis accuracy does not reach the preset analysis accuracy;
and the model updating unit is used for carrying out iterative updating on the applicable workpiece polishing analysis model based on a PSO algorithm and the model analysis deviation degree to obtain the applicable workpiece polishing optimization analysis model.
Further, the model updating unit further includes:
the optimization space construction unit is used for constructing a particle optimization space according to the model training parameters of the applicable workpiece polishing analysis model;
the fitness function calculation unit is used for initializing the particle optimization space to obtain particle swarm constraint parameters, and iteratively calculating a particle swarm fitness function according to the model analysis deviation degree and the particle swarm constraint parameters;
the fitness function output unit is used for obtaining a first output result of the particle swarm fitness function when a preset termination condition is reached, wherein the first output result comprises optimal result particles;
and the model optimization unit is used for mapping the optimal result particles to the applicable workpiece polishing analysis model for optimization training to obtain the applicable workpiece polishing optimization analysis model.
The application provides a robot control method based on a flexible perception neural network, which comprises the following steps: image acquisition is carried out on the workpiece to be processed through a visual information acquisition module, so that image information of the workpiece to be processed is obtained; uploading the image information of the workpiece to be processed to a polishing robot control system based on a data interaction module for interaction analysis; the method comprises the steps that target polishing standard information is obtained, the polishing robot control system inputs the target polishing standard information and the image information of a workpiece to be processed into a workpiece polishing analysis model suitable for the workpiece polishing analysis, and a model output result is obtained, wherein the model output result comprises a workpiece polishing scheme to be processed; the workpiece polishing scheme to be processed is sent to the flexible polishing robot through the data interaction module, and the flexible polishing robot is controlled to polish the workpiece according to the workpiece polishing scheme to be processed through the motion trail control module; acquiring workpiece polishing force information in the workpiece polishing process through a force axis sensor module, and performing force compensation analysis according to the workpiece polishing force information to acquire action force compensation information; and capturing the gesture according to the visual information acquisition module to obtain polishing gesture space information, analyzing the motion force compensation information and the polishing gesture space information through a flexible sensing network, and correcting the polishing track of the polishing scheme of the workpiece to be processed. The technical problem that the quality of a polished workpiece does not reach the standard due to low polishing control precision of the conventional polishing robot in the operation process is solved. The robot has the advantages that real-time interaction between the machine vision acquisition information and the polishing force acquisition information and the flexible polishing robot is achieved, polishing force compensation is conducted on the robot in any posture, workpiece polishing force is guaranteed, workpiece polishing precision is improved, and further technical effect of workpiece polishing quality is guaranteed.
The specification and drawings are merely exemplary illustrations of the present application, and the present invention is intended to cover such modifications and variations if they fall within the scope of the invention and its equivalents.
Claims (6)
1. A robot control method based on a flexible sensory neural network, wherein the method is applied to a polishing robot control system communicatively connected to a flexible polishing robot, the method comprising:
the flexible polishing robot comprises a visual information acquisition module, a force axis sensor module, a data interaction module and a motion trail control module;
the visual information acquisition module is used for acquiring images of the workpiece to be processed to obtain image information of the workpiece to be processed;
uploading the image information of the workpiece to be processed to the polishing robot control system based on the data interaction module for interaction analysis;
acquiring target polishing standard information, inputting the target polishing standard information and the image information of the workpiece to be processed into a suitable workpiece polishing analysis model by using a polishing robot control system to obtain a model output result, wherein the model output result comprises a workpiece polishing scheme to be processed, and performing denoising and filtering on the image information of the workpiece to be processed based on an image filtering algorithm to obtain denoising workpiece image information to be processed; carrying out gray level equalization processing on the denoising workpiece image information to be processed to obtain average workpiece image information to be processed; performing image enhancement on the image information of the workpiece to be processed with the average value based on a generated countermeasure network, and performing image grid segmentation on the enhanced image to obtain standard image information of the workpiece to be processed; the method comprises the following steps: classifying the basic information of the workpiece to be processed through a polished workpiece feature decision tree to obtain polished workpiece classification feature information; calibrating the workpiece to be processed according to the classifying characteristic information of the polished workpiece to determine the calibrating parameters of the polished workpiece; calling the applicable workpiece polishing analysis model from a workpiece polishing analysis model library based on the polishing workpiece calibration parameters; inputting the target polishing standard information and the standard workpiece image information to be processed into the applicable workpiece polishing analysis model, and outputting the workpiece polishing scheme to be processed;
the workpiece polishing scheme to be processed is sent to the flexible polishing robot through the data interaction module, and the flexible polishing robot is controlled to polish the workpiece according to the workpiece polishing scheme to be processed through the motion trail control module;
acquiring workpiece polishing force information in the workpiece polishing process through the force axis sensor module, and performing force compensation analysis according to the workpiece polishing force information to acquire action force compensation information;
and capturing the gesture according to the visual information acquisition module to obtain polishing gesture space information, analyzing the motion force compensation information and the polishing gesture space information through a flexible sensing network, and correcting the polishing track of the polishing scheme of the workpiece to be processed.
2. The method of claim 1, wherein the method comprises:
the applicable workpiece polishing analysis model comprises an input layer, a convolution image network layer, a characteristic analysis logic layer and an output layer;
the target polishing standard information and the standard workpiece image information to be processed are used as input layers and input into the convolution image network layer, and feature information of the workpiece to be processed is obtained;
inputting the target polishing standard information and the workpiece characteristic information to be processed into the characteristic analysis logic layer to obtain a polishing scheme of the workpiece to be processed;
and outputting the polishing scheme of the workpiece to be processed as an output result through the output layer.
3. The method of claim 2, wherein the obtaining workpiece characteristic information to be processed comprises:
obtaining a preset convolution feature set according to the target polishing standard information, wherein the preset convolution feature set comprises workpiece flatness features, curvature features and hole defect features;
inputting the standard workpiece image information to be processed into the convolution image network layer as input information to perform feature extraction;
and obtaining output information of the convolution image network layer, wherein the output information comprises the workpiece feature information to be processed conforming to the preset convolution feature set.
4. The method of claim 1, wherein the method comprises:
performing analysis effect verification on the applicable workpiece polishing analysis model to obtain model analysis accuracy;
if the model analysis accuracy does not reach the preset analysis accuracy, obtaining model analysis deviation based on the difference value between the model analysis accuracy and the preset analysis accuracy;
and based on a PSO algorithm and the model analysis deviation degree, iteratively updating the applicable workpiece polishing analysis model to obtain the applicable workpiece polishing optimization analysis model.
5. The method of claim 4, wherein said obtaining an applicable workpiece polishing optimization analysis model comprises:
constructing a particle optimization space according to the model training parameters of the applicable workpiece polishing analysis model;
initializing the particle optimization space to obtain particle swarm constraint parameters, and iteratively calculating a particle swarm fitness function according to the model analysis deviation degree and the particle swarm constraint parameters;
when a preset termination condition is reached, a first output result of the particle swarm fitness function is obtained, wherein the first output result comprises optimal result particles;
and mapping the optimal result particles to the applicable workpiece polishing analysis model for optimization training to obtain the applicable workpiece polishing optimization analysis model.
6. A robot control system based on a flexible sensory neural network, the system comprising:
the robot forming module is used for the flexible polishing robot and comprises a visual information acquisition module, a force axis sensor module, a data interaction module and a motion trail control module;
the visual information acquisition module is used for acquiring images of the workpiece to be processed through the visual information acquisition module to obtain image information of the workpiece to be processed;
the data interaction module is used for uploading the image information of the workpiece to be processed to a polishing robot control system for interaction analysis based on the data interaction module;
the model output module is used for acquiring target polishing standard information, the polishing robot control system inputs the target polishing standard information and the image information of the workpiece to be processed into an applicable workpiece polishing analysis model to obtain a model output result, and the model output result comprises a workpiece polishing scheme to be processed, wherein the model output result comprises: the denoising filter unit is used for denoising and filtering the image information of the workpiece to be processed based on an image filtering algorithm to obtain denoising image information of the workpiece to be processed; the equalization processing unit is used for carrying out gray level equalization processing on the denoising workpiece image information to be processed to obtain average workpiece image information to be processed; the image enhancement unit is used for enhancing the image of the workpiece image information to be processed with the mean value based on the generation countermeasure network, and then dividing the enhanced image into image grids to obtain the standard workpiece image information to be processed, and further comprises: the feature classification unit is used for classifying the basic information of the workpiece to be processed through a polished workpiece feature decision tree to obtain polished workpiece classification feature information; the parameter calibration unit is used for calibrating the workpiece to be processed according to the classification characteristic information of the polished workpiece to determine the calibration parameters of the polished workpiece; the model calling unit is used for calling the applicable workpiece polishing analysis model from a workpiece polishing analysis model library based on the polishing workpiece calibration parameters; the model output unit is used for inputting the target polishing standard information and the standard workpiece image information to be processed into the applicable workpiece polishing analysis model and outputting the workpiece polishing scheme to be processed;
the motion trail control module is used for sending the workpiece polishing scheme to be processed to the flexible polishing robot through the data interaction module, and controlling the flexible polishing robot to polish the workpiece according to the workpiece polishing scheme to be processed through the motion trail control module;
the force axis sensor module is used for acquiring workpiece polishing force information in the workpiece polishing process through the force axis sensor module, and performing force compensation analysis according to the workpiece polishing force information to acquire action force compensation information;
and the polishing track correction module is used for capturing the gesture according to the visual information acquisition module to obtain polishing gesture space information, and analyzing the motion force compensation information and the polishing gesture space information through a flexible sensing network to correct the polishing track of the workpiece to be processed in a polishing scheme.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202210588502.5A CN114905515B (en) | 2022-05-26 | 2022-05-26 | Robot control method and system based on flexible perception neural network |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202210588502.5A CN114905515B (en) | 2022-05-26 | 2022-05-26 | Robot control method and system based on flexible perception neural network |
Publications (2)
Publication Number | Publication Date |
---|---|
CN114905515A CN114905515A (en) | 2022-08-16 |
CN114905515B true CN114905515B (en) | 2023-05-12 |
Family
ID=82768939
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202210588502.5A Active CN114905515B (en) | 2022-05-26 | 2022-05-26 | Robot control method and system based on flexible perception neural network |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN114905515B (en) |
Families Citing this family (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN116525517B (en) * | 2023-06-19 | 2023-10-13 | 苏州鸿安机械股份有限公司 | Positioning control method and system for conveying semiconductor wafers |
Family Cites Families (11)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN105643636B (en) * | 2016-04-13 | 2018-03-09 | 广州文冲船厂有限责任公司 | A kind of robot sanding apparatus |
CN107838920A (en) * | 2017-12-20 | 2018-03-27 | 芜湖哈特机器人产业技术研究院有限公司 | A kind of robot polishing Force control system and method |
CN108509947A (en) * | 2018-01-29 | 2018-09-07 | 佛山市南海区广工大数控装备协同创新研究院 | A kind of automatic identification polishing process based on artificial neural network |
CN108388702A (en) * | 2018-01-30 | 2018-08-10 | 河南工程学院 | Engineering ceramics electrical discharge machining effect prediction method based on PSO neural networks |
CN108972573A (en) * | 2018-06-12 | 2018-12-11 | 浙江大学 | A kind of industrial robot automation wheel hub polishing system and method based on characteristics of image identification |
CN109732450B (en) * | 2019-02-27 | 2021-02-26 | 重庆理工大学 | Blade polishing and grinding method based on neural network |
CN111975579B (en) * | 2020-07-29 | 2022-06-14 | 华南理工大学 | Robot constant-force polishing system based on polishing model and iterative algorithm |
US20220032461A1 (en) * | 2020-07-31 | 2022-02-03 | GrayMatter Robotics Inc. | Method to incorporate complex physical constraints in path-constrained trajectory planning for serial-link manipulator |
CN112388511B (en) * | 2020-09-23 | 2021-11-12 | 广东博智林机器人有限公司 | Control method and device of polishing robot, electronic equipment and storage medium |
CN112171458B (en) * | 2020-11-27 | 2021-03-16 | 大捷智能科技(广东)有限公司 | Intelligent mold polishing platform and polishing method |
CN112757057A (en) * | 2021-01-19 | 2021-05-07 | 武汉海默机器人有限公司 | Intelligent manual-teaching-free grinding and polishing method and system integrating visual depth analysis |
-
2022
- 2022-05-26 CN CN202210588502.5A patent/CN114905515B/en active Active
Also Published As
Publication number | Publication date |
---|---|
CN114905515A (en) | 2022-08-16 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN104915957B (en) | A kind of matching antidote for improving industrial robot 3D vision accuracy of identification | |
CN111079561B (en) | Robot intelligent grabbing method based on virtual training | |
CN110223345B (en) | Point cloud-based distribution line operation object pose estimation method | |
CN114905515B (en) | Robot control method and system based on flexible perception neural network | |
CN108247637B (en) | Industrial robot arm vision anti-collision control method | |
CN112819772B (en) | High-precision rapid pattern detection and recognition method | |
CN107194371B (en) | User concentration degree identification method and system based on hierarchical convolutional neural network | |
CN109143855B (en) | Visual servo control method of unmanned gyroplane based on fuzzy SARSA learning | |
CN113034600B (en) | Template matching-based texture-free planar structure industrial part identification and 6D pose estimation method | |
CN111881772A (en) | Multi-mechanical arm cooperative assembly method and system based on deep reinforcement learning | |
JP2019057250A (en) | Work-piece information processing system and work-piece recognition method | |
CN116525517B (en) | Positioning control method and system for conveying semiconductor wafers | |
CN115026660B (en) | CCD-based grinding process intelligent control method and system | |
CN117523555A (en) | Aircraft part defect detection method based on self-encoder | |
CN117227247A (en) | Intelligent positioning control method for carton processing | |
CN106127147B (en) | A kind of face depth texture restorative procedure based on three-dimensional data | |
CN115157272B (en) | Automatic programming system based on visual scanning | |
CN111553437A (en) | Neural network based image classification method | |
CN113808206B (en) | Typesetting system and method based on vision tracking robot | |
Juang et al. | A wheeled mobile robot path-tracking system based on image processing and adaptive CMAC | |
EP3910595A1 (en) | Reinforcement learning based conveyoring control | |
Rost et al. | Knowledge-based configuration of image processing algorithms | |
Nag et al. | Generating Vectors from Images using Multi-Stage Edge Detection for Robotic Artwork | |
CN118357934A (en) | Precise positioning method and device for manipulator based on visual positioning | |
CN117671011B (en) | AGV positioning precision improving method and system based on improved ORB algorithm |
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 |