High-precision detection method and device for load sensor

By collecting and analyzing load detection data and action parameters on the load sensor, and combining the time stamp of the image sequence for error compensation, the load detection accuracy problem caused by impact and vibration during the grabbing process of the robot arm is solved, and the detection efficiency is improved and precision.

CN118789560BActive Publication Date: 2024-11-19TIANJIN WANZHONG TECH CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202411273255.5
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2024-09-12
Publication Date
2024-11-19
Estimated Expiration
2044-09-12

Smart Images

  • Figure CN118789560B_ABST
    Figure CN118789560B_ABST
Patent Text Reader

Abstract

The present application provides a high-precision detection method and device for load sensors, which relates to the field of mechanical arm detection technology, including: when the mechanical arm grasps a workpiece, using the load sensor to collect data, combining the grasping action parameters to perform impact error analysis, obtaining vibration feature information, adjusting the image sequence in combination with the image timestamp, achieving error compensation, and obtaining high-precision load detection results. This application can solve the problem that the existing technology is affected by shock and vibration, resulting in low accuracy of load detection results, which further affects the detection efficiency of the mechanical arm during the grasping process. It can effectively analyze and compensate for the load detection error caused by the vibration of the mechanical arm, improve the detection efficiency, and further improve the operating accuracy of the mechanical arm in a dynamic environment.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present application relates to the field of mechanical arm detection technology, and in particular to a high-precision detection method and device for a load sensor. Background Art

[0002] In the field of industrial automation, accurate power control is essential to improving product quality and efficiency. As a key component of force control, the accuracy of load sensors directly affects the performance of the entire system.

[0003] At present, due to the influence of shock and vibration, the existing technology leads to low accuracy of load detection results, which further affects the detection efficiency of the robot arm during the grasping process. Summary of the invention

[0004] The purpose of this application is to provide a high-precision detection method and device for load sensors, so as to solve the problem that the load detection results in the prior art are not accurate due to the influence of impact and vibration, which further affects the detection efficiency of the robot arm during the grasping process.

[0005] In view of the above problems, the present application provides a high-precision detection method and device for a load sensor.

[0006] In a first aspect, the present application provides a high-precision detection method for a load sensor, which is implemented by a high-precision detection device for a load sensor, wherein the high-precision detection method for a load sensor comprises: when a target robot arm grasps a target workpiece, based on the load sensor arranged in the target robot arm, first load detection data is collected, as well as grasping action parameters of the target robot arm and a grasping image sequence of grasping the target workpiece; according to the grasping action parameters, load detection impact error analysis is performed to obtain impact error parameters, error compensation is performed on the first load detection data to obtain second load detection data; according to the grasping action parameters, target Vibration analysis of the robotic arm is performed to obtain vibration characteristic information, wherein the vibration characteristic information includes vibration amplitude, vibration period and vibration direction; based on the timestamp of the captured images in the captured image sequence and in combination with the vibration characteristic information, the captured image sequence is adjusted and restored to obtain a restored captured image sequence; based on the restored captured image sequence, a load detection capture error analysis is performed to obtain a capture error parameter, and the second load detection data is error compensated to obtain third load detection data; based on the vibration characteristic information, a load detection vibration error analysis is performed to obtain a vibration error parameter, and the third load detection data is error compensated to obtain fourth load detection data as a high-precision load detection result.

[0007] In a second aspect, the present application further provides a high-precision detection device for a load sensor, which is used to execute the high-precision detection method for a load sensor as described in the first aspect, wherein the high-precision detection device for a load sensor comprises: a first load detection data acquisition module, which is used to collect and obtain first load detection data based on the load sensor arranged in the target robotic arm when the target robotic arm grasps the target workpiece, and collect the grasping action parameters of the target robotic arm and the grasping image sequence of the target workpiece; a second load detection data acquisition module, which is used to perform load detection impact error analysis according to the grasping action parameters, obtain impact error parameters, perform error compensation on the first load detection data, and obtain second load detection data; a vibration characteristic information acquisition module, which is used to perform target detection according to the grasping action parameters. Vibration analysis of the robotic arm to obtain vibration characteristic information, wherein the vibration characteristic information includes vibration amplitude, vibration period and vibration direction; an image restoration module, used to adjust and restore the captured image sequence according to the timestamp of the captured image in the captured image sequence in combination with the vibration characteristic information, so as to obtain a restored captured image sequence; a third load detection data acquisition module, used to perform load detection capture error analysis according to the restored capture image sequence, obtain capture error parameters, perform error compensation on the second load detection data, and obtain third load detection data; a detection result acquisition module, used to perform load detection vibration error analysis according to the vibration characteristic information, obtain vibration error parameters, perform error compensation on the third load detection data, and obtain fourth load detection data as a high-precision load detection result.

[0008] One or more technical solutions provided in this application have at least the following technical effects or advantages:

[0009] When a target manipulator grasps a target workpiece, based on a load sensor arranged in the target manipulator, first load detection data is collected, as well as grasping action parameters of the target manipulator and a grasping image sequence of grasping the target workpiece; load detection impact error analysis is performed according to the grasping action parameters to obtain impact error parameters, and error compensation is performed on the first load detection data to obtain second load detection data; vibration analysis of the target manipulator is performed according to the grasping action parameters to obtain vibration characteristic information, wherein the vibration characteristic information includes vibration amplitude, vibration period and vibration direction; and grasped image adjustment and restoration is performed on the grasped image sequence according to the timestamp of the grasped image in the grasped image sequence in combination with the vibration characteristic information to obtain A restored grasping image sequence is obtained; a load detection grasping error analysis is performed according to the restored grasping image sequence to obtain a grasping error parameter, the second load detection data is error compensated, and the third load detection data is obtained; a load detection vibration error analysis is performed according to the vibration characteristic information to obtain a vibration error parameter, the third load detection data is error compensated, and the fourth load detection data is obtained as a high-precision load detection result. This effectively solves the problem that the load detection result of the prior art is not accurate due to the influence of shock and vibration, which further affects the detection efficiency of the robot arm during the grasping process. The load detection error caused by the vibration of the robot arm can be effectively analyzed and compensated, the detection efficiency is improved, and the operation accuracy of the robot arm in a dynamic environment is further improved.

[0010] The above description is only an overview of the technical solution of the present application. In order to more clearly understand the technical means of the present application, it can be implemented according to the contents of the specification, and in order to make the above and other purposes, features and advantages of the present application more obvious and easy to understand, the specific implementation methods of the present application are specifically cited below. It should be understood that the content described in this section is not intended to identify the key or important features of the embodiments of the present application, nor is it intended to limit the scope of the present application. Other features of the present application will become easy to understand through the following description. BRIEF DESCRIPTION OF THE DRAWINGS

[0011] In order to more clearly illustrate the technical solutions in the present application or the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings in the following description are only exemplary, and for ordinary technicians in this field, other drawings can be obtained based on the provided drawings without paying any creative work.

[0012] Figure 1 A schematic diagram of the process of the high-precision detection method for load sensors in this application;

[0013] Figure 2This is a schematic diagram of the structure of a high-precision detection device for a load sensor in the present application.

[0014] Description of reference numerals:

[0015] A first load detection data acquisition module 11 , a second load detection data acquisition module 12 , a vibration characteristic information acquisition module 13 , an image restoration module 14 , a third load detection data acquisition module 15 , and a detection result acquisition module 16 . DETAILED DESCRIPTION

[0016] The present application provides a high-precision detection method and device for load sensors, thereby solving the problem in the prior art that the load detection results are inaccurate due to the influence of impact and vibration, which further affects the detection efficiency of the robot arm during the grasping process. The application can effectively analyze and compensate for the load detection error caused by the vibration of the robot arm, improve the detection efficiency, and further improve the operating accuracy of the robot arm in a dynamic environment.

[0017] Below, the technical solutions in the present application will be clearly and completely described with reference to the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all of the embodiments of the present application. It should be understood that the present application is not limited to the example embodiments described herein. Based on the embodiments of the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present application. It should also be noted that, for the convenience of description, only the parts related to the present application are shown in the accompanying drawings, rather than all of them.

[0018] For example, please refer to the attached Figure 1 The present application provides a high-precision detection method for a load sensor, wherein the high-precision detection method for a load sensor is applied to a high-precision detection device for a load sensor, and the high-precision detection method for a load sensor specifically comprises the following steps:

[0019] S1: When a target robot arm grasps a target workpiece, first load detection data is acquired based on a load sensor disposed in the target robot arm, and grasping action parameters of the target robot arm and a grasping image sequence of grasping the target workpiece are acquired.

[0020] Specifically, high-precision load sensors are installed at key parts of the robot arm, such as joints and end effectors, to ensure that the sensors are compatible with the structure of the robot arm and will not affect the normal operation of the robot arm. When the robot arm performs a grasping action, the load sensor measures and records the force applied to the workpiece in real time to determine the grasping action parameters that need to be collected, such as position, speed, acceleration, joint angle, etc. These parameters are obtained through encoders, speed sensors and other devices built into the robot arm's control system. Or use external devices such as motion capture systems to track and record the movements of the robot arm. Capture the entire grasping process of the robot arm and the workpiece through a camera. Continuously shoot the grasping process at a certain frame rate, such as 30fps. Ensure that the image has an accurate timestamp for subsequent synchronous analysis with load data and action parameters.

[0021] S2: performing load detection impact error analysis according to the grabbing action parameters to obtain impact error parameters, performing error compensation on the first load detection data to obtain second load detection data.

[0022] Specifically, identify the grasping action parameters, including speed, acceleration, etc. Determine the action stages that may cause impact errors, such as rapid approach to the workpiece, the moment of contact with the workpiece, and rapid withdrawal. Identify mutation points or sharp changes in the action parameters, which are signs of impact errors. Match the load detection data with the time series and analyze the change of force over time. Look for force peaks corresponding to the sudden changes in the action parameters in the load time series, and determine whether these peaks exceed the normal operating force range to determine whether they are impact errors. Estimate the size, duration, and range of impact errors, and use mathematical models such as filters, wavelet transforms, etc. to extract the characteristics of the impact errors. Process the first load detection data according to the characteristics of the impact error. For example, a digital filter can be used to remove or reduce the impact of the impact error. Alternatively, the force peak can be manually adjusted to make it more consistent with the surrounding force data.

[0023] S3: Performing vibration analysis on the target robot arm according to the grasping action parameters to obtain vibration characteristic information, wherein the vibration characteristic information includes vibration amplitude, vibration period and vibration direction.

[0024] Specifically, identify the key sources that may cause vibration of the robot arm, such as joint drive, external interference, structural weaknesses of the robot arm, etc., and select the grasping action parameters for vibration analysis, such as joint angular velocity, angular acceleration, linear velocity and acceleration. Install accelerometers or other vibration sensors at key parts of the robot arm, such as joints and connecting rods. Record the vibration sensor data in real time when the robot arm performs the grasping action, and ensure that the data has an accurate time stamp so that it can be synchronized with the grasping action parameters. Use filters to remove noise from the sensor data, such as high-pass filters to remove low-frequency interference and low-pass filters to remove high-frequency noise. Calculate the root mean square or peak-to-peak value of the vibration signal to determine the vibration amplitude. Determine the main frequency of the vibration through spectrum analysis, such as fast Fourier transform FFT, and thus know the vibration period. Analyze the three-dimensional acceleration data to determine the direction of vibration. Vector analysis or principal component analysis can be used. Parameterize the extracted vibration features, such as vibration amplitude, vibration period and vibration direction. Analyze the relationship between the vibration features and the grasping action parameters to identify which action parameters may cause significant vibration.

[0025] S4: adjusting and restoring the captured images in the captured image sequence according to the timestamps of the captured images in the captured image sequence and combining the vibration characteristic information to obtain a restored captured image sequence.

[0026] Specifically, ensure that each image in the captured image sequence accurately corresponds to the corresponding timestamp. If there is a time deviation, perform time synchronization correction. Map the extracted vibration features, vibration amplitude, period, and direction to the timeline of the image sequence, and analyze the vibration state corresponding to each image. According to the vibration feature information, estimate the displacement of each image due to vibration, and use image processing algorithms such as optical flow method and image registration technology to correct the image displacement. Analyze the degree of blur caused by vibration, and apply deblurring algorithms such as Wiener filtering and blind convolution to restore image clarity. Align the corrected images in time series to ensure the coherence and consistency of the entire sequence, check the restored image sequence to ensure that the image is clear, without displacement and distortion, and save the corrected image sequence as the restored captured image sequence.

[0027] S5: performing load detection grasping error analysis according to the restored grasping image sequence to obtain grasping error parameters, performing error compensation on the second load detection data to obtain third load detection data.

[0028] Specifically, extract key features such as edges, corners, shapes, etc. from the restored grasping image sequence, track the target workpiece in the image sequence, and analyze its motion trajectory and posture changes during the grasping process. Compare the grasping position in the image sequence with the expected grasping path, identify the deviation, and analyze the grasping error caused by inaccurate robot arm movement or workpiece position change. Estimate the size and direction of the grasping error, including position error, angle error, and force error. Based on the results of the visual error analysis, convert the image error into an estimate of the load error. For example, if the image analysis shows that the robot arm deviates downward by a certain angle during grasping, the corresponding load data needs to subtract the additional force component caused by this offset. The compensated load data is recorded as the third load detection data, which should more accurately reflect the actual load of the robot arm during the grasping process.

[0029] S6: Perform load detection vibration error analysis according to the vibration characteristic information to obtain vibration error parameters, perform error compensation on the third load detection data, and obtain fourth load detection data as a high-precision load detection result.

[0030] Specifically, analyze the relationship between vibration characteristics, amplitude, period, direction and load detection data. Determine how vibration affects load readings, for example, vibration may cause fluctuations in load sensor readings. Identify vibration-induced errors in load data, which appear as periodic fluctuations or spikes, and estimate the parameters of vibration errors, including error magnitude, frequency and phase. Select a suitable mathematical model to describe the vibration error based on the vibration characteristics, such as Fourier series, autoregressive model, etc. Use vibration characteristic information to parameterize the error model to ensure that the model can accurately reflect the impact of vibration on load data. Apply the established vibration error model to compensate the third load detection data, and the load reading can be adjusted in real time through the software algorithm to eliminate or reduce the error caused by vibration. Implement compensation algorithms, such as using filters to remove vibration components of specific frequencies, or predicting and compensating for load fluctuations caused by vibration through prediction models. Record the compensated load data as the fourth load detection data. The load detection error caused by the vibration of the manipulator can be effectively analyzed and compensated, thereby improving the accuracy of load detection.

[0031] Furthermore, step S2 of the present application also includes:

[0032] According to the grabbing load test data record of the target robot arm, a sample grabbing action parameter set and a sample impact error parameter set are collected; the sample grabbing action parameter set and the sample impact error parameter set are used to train an impact error analyzer; the impact error analyzer is used to perform load detection impact error analysis on the grabbing action parameters to obtain impact error parameters; the impact error parameters are used to perform error compensation calculation on the first load detection data to obtain second load detection data.

[0033] Specifically, sample grasping motion parameters, including speed, acceleration, position, etc., are extracted from the gripping load test data record of the manipulator. Similarly, parameters related to impact error, such as impact moment, impact force, duration, etc., are extracted from the test data, and the data is cleaned to remove outliers and noise. Standardize or normalize the data. Select a suitable machine learning model to train the impact error analyzer, such as a regression model, a neural network, or a support vector machine. Determine which grasping motion parameters and impact error parameters are input as features into the model. Use a set of sample grasping motion parameters and a set of sample impact error parameters to train the model, and optimize the model parameters through techniques such as cross-validation to prevent overfitting. Use an independent validation set to evaluate the accuracy of the model and adjust the model parameters until satisfactory performance is achieved. Input the actual grasping motion parameters into the trained impact error analyzer. The analyzer outputs impact error parameters, which describe the expected impact error characteristics. Formulate a compensation strategy based on the impact error parameters to determine how to adjust the load data to eliminate the impact effect. Apply the compensation strategy to adjust the first load detection data, and record the compensated load data as the second load detection data.

[0034] Furthermore, step S3 of the present application also includes:

[0035] According to the grasping vibration test data record of the target robotic arm, a sample grasping action parameter set and a sample vibration feature information set are collected, wherein each sample vibration feature information includes a sample vibration amplitude, a sample vibration period and a sample vibration direction; the sample grasping action parameter set and the sample vibration feature information set are used to train a vibration feature analyzer; the vibration feature analyzer is used to perform vibration analysis of the target robotic arm on the grasping action parameters to obtain vibration feature information.

[0036] Specifically, sample grasping motion parameters are extracted from the grasping vibration test data records of the robot arm, including speed, acceleration, joint angle, etc. Similarly, vibration-related feature information is extracted from the test data, including vibration amplitude, vibration period, and vibration direction. Clean the data, remove outliers and noise, and standardize or normalize the data. Select a suitable machine learning model to train the vibration feature analyzer, such as a regression model, a neural network, or a deep learning model. Determine which grasping motion parameters and vibration feature information are input as features into the model. Use a set of sample grasping motion parameters and a set of sample vibration feature information to train the model, and optimize the model parameters through techniques such as cross-validation to prevent overfitting. Use an independent validation set to evaluate the accuracy of the model and adjust the model parameters until satisfactory performance is achieved. Input the actual grasping motion parameters into the trained vibration feature analyzer. The analyzer outputs vibration feature information that describes the expected vibration characteristics, including vibration amplitude, period, and direction.

[0037] Furthermore, step S4 of the present application also includes:

[0038] The starting timestamp information of the target robot arm starting the grasping action is obtained, and multiple image timestamp information is assigned to multiple grasped images in the grasped image sequence according to the starting timestamp information; multiple vibration stroke information of the multiple grasped images is assigned according to the vibration characteristic information and the multiple image timestamp information; and based on the multiple vibration stroke information, the grasped image sequence is adjusted and restored based on the generative adversarial network to obtain a restored grasped image sequence.

[0039] Specifically, when the robot arm starts the grasping action, the starting timestamp is recorded to ensure that the timestamp corresponds accurately to each image in the grasped image sequence. According to the starting timestamp information, an accurate timestamp is assigned to each image in the grasped image sequence to ensure that the timestamp is synchronized with the grasping action parameters and vibration feature information. Analyze the vibration feature information, determine the vibration period and vibration direction, and combine the vibration period to determine the vibration stroke, that is, the position of each grasped image in the grasped image sequence, which corresponds to the timestamp. For example, the vibration stroke of the previous grasped image is 1mm to the right, and the vibration stroke of the next grasped image is 2mm to the right. The vibration stroke can describe the position of the vibration deviation of the robot arm to restore the image. According to the vibration period and vibration direction, a vibration stroke information is assigned to each image in the grasped image sequence. Ensure that the vibration stroke information is synchronized with the image timestamp information. Select a suitable generative adversarial network architecture, such as DCGAN (Deep Convolutional Generative Adversarial Network) or WGAN (Wasserstein GAN). Use the vibration stroke information as the input feature of GAN and the grasped image sequence as the training data of GAN. The GAN model is trained using the vibration stroke information and the captured image sequence. By adjusting the model parameters, the generator can generate images similar to the real captured images based on the vibration stroke information. An independent validation set is used to evaluate the model's generation ability and adjust the model parameters until a satisfactory generation effect is achieved. The trained GAN model is applied to the captured image sequence. According to the vibration stroke information of each image, the image is adjusted to restore the blur or displacement caused by the vibration, and the adjusted image sequence is recorded as the restored captured image sequence.

[0040] Furthermore, the present application also includes:

[0041] According to the grabbing data record of the target robotic arm, a sample grabbing image set and a sample vibration stroke information set are collected, and images of the target robotic arm after vibration stops are collected to obtain a sample restored grabbing image set; based on a generative adversarial network, a grabbing image restorer is constructed, wherein the grabbing image restorer includes a generation channel and an adversarial channel; the sample grabbing image set and the sample vibration stroke information set are used as generation training data, and the sample restored grabbing image set is used as adversarial training data to train the grabbing image restorer until convergence; the multiple vibration stroke information and the multiple grabbing images are input into the grabbing image restorer to restore and generate a restored grabbing image sequence.

[0042] Specifically, a series of sample grasping images are extracted from the grasping data records of the robot arm to ensure that these images can represent the different stages and conditions of the robot arm grasping process. The robot arm vibration test data is analyzed to determine the vibration stroke information of each sample, including vibration amplitude, vibration period and vibration direction. The image after the vibration of the robot arm stops is collected as a reference for the sample restored grasping image set. A suitable generative adversarial network architecture, such as DCGAN or WGAN, is selected to build a grasping image restorer containing a generative channel and an adversarial channel. The generative channel is responsible for generating restored images based on the vibration stroke information, and the adversarial channel is responsible for evaluating the similarity between the generated image and the real restored image. The sample grasping image set and the vibration stroke information set are used as generative training data to train the generative channel so that it can generate images similar to the real restored images. The sample restored grasping image set is used as adversarial training data to train the adversarial channel so that it can distinguish between the real restored images and the generated images. Through iterative training, the model parameters are adjusted so that the generator can generate more and more realistic restored images, while the discriminator can distinguish between the real and generated images more and more accurately. The training process may require multiple iterations until the model converges. A plurality of vibration stroke information and a plurality of grasped images are input into a trained grasped image restorer, and the grasped image restorer generates a restored grasped image sequence according to the input information.

[0043] Furthermore, step S5 of the present application also includes:

[0044] According to the grabbing load test data record of the target robot arm, a set of sample restored grabbing image sequences is collected, and according to the error influence of the posture of the grabbing workpiece in different restored grabbing image sequences on the load detection, a set of sample grabbing error parameters is obtained; using the sample restored grabbing image sequence set and the sample grabbing error parameter set, a grabbing error identifier is trained based on a spatiotemporal graph convolutional network; using the grabbing error identifier, load detection grabbing error analysis is performed on the restored grabbing image sequence to obtain grabbing error parameters; using the grabbing error parameters, error compensation calculation is performed on the second load detection data to obtain third load detection data.

[0045] Specifically, a series of sample restored grasping image sequences are extracted from the gripping load test data records of the robot. Ensure that these image sequences can represent the different postures and conditions of the robot in the grasping process, analyze the restored grasping image sequences, and determine the impact of different grasping workpiece postures on the load detection error. Extract sample grasping error parameters, including the size, direction, and position of the error. Select a suitable spatiotemporal graph convolutional network architecture, such as ST-GCN (Spatial-Temporal Graph Convolutional Network). Construct a grasping error recognizer containing a spatiotemporal graph convolutional layer, which can process sequence data and identify grasping error features in image sequences. The set of sample restored grasping image sequences and the set of grasping error parameters are used as training data to train the recognizer so that it can identify error features in grasping image sequences. Through iterative training, the model parameters are adjusted so that the recognizer can identify grasping errors more and more accurately. The training process requires multiple iterations until the model converges. The restored grasping image sequence is input into the trained grasping error recognizer, which analyzes the input image sequence and outputs the grasping error parameters. The second load detection data is adjusted according to the grasping error parameter, and the adjusted data is recorded as the third load detection data, which can more accurately reflect the actual load condition of the robot arm during the grasping process.

[0046] Furthermore, step S6 of the present application also includes:

[0047] According to the grabbing load test data record of the target robot arm, a sample vibration feature information set is collected, and according to the load detection error obtained from different vibration feature information tests, a sample vibration error parameter set is obtained; the sample vibration feature information set and the sample vibration error parameter set are used to train a vibration error analyzer; the vibration error analyzer is used to perform load detection vibration error analysis on the vibration feature information to obtain vibration error parameters; the vibration error parameters are used to perform error compensation calculation on the third load detection data to obtain fourth load detection data.

[0048] Specifically, a series of sample vibration feature information, including vibration amplitude, vibration period and vibration direction, is extracted from the grab load test data record of the robot arm. The load detection error obtained under different vibration feature information is analyzed, and the sample vibration error parameters, including the size, direction and position of the error, are extracted. A suitable machine learning model is selected to train the vibration error analyzer, such as a regression model, a neural network or a deep learning model. Determine which vibration feature information is input into the model as a feature. The model is trained using a set of sample vibration feature information and a set of sample vibration error parameters, and the model parameters are optimized through cross-validation and other techniques to prevent overfitting. An independent validation set is used to evaluate the accuracy of the model. The actual vibration feature information is input into the trained vibration error analyzer, and the analyzer outputs vibration error parameters, which describe the expected vibration error characteristics. A compensation strategy is formulated based on the vibration error parameters to determine how to adjust the load data to eliminate the impact of vibration. The compensation strategy is applied to adjust the third load detection data, and the compensated load data is recorded as the fourth load detection data.

[0049] In summary, the high-precision detection method for load sensors provided in this application has the following technical effects:

[0050] When a target manipulator grasps a target workpiece, based on a load sensor arranged in the target manipulator, first load detection data is collected, as well as grasping action parameters of the target manipulator and a grasping image sequence of grasping the target workpiece; load detection impact error analysis is performed according to the grasping action parameters to obtain impact error parameters, and error compensation is performed on the first load detection data to obtain second load detection data; vibration analysis of the target manipulator is performed according to the grasping action parameters to obtain vibration characteristic information, wherein the vibration characteristic information includes vibration amplitude, vibration period and vibration direction; and grasped image adjustment and restoration is performed on the grasped image sequence according to the timestamp of the grasped image in the grasped image sequence in combination with the vibration characteristic information to obtain A restored grasping image sequence is obtained; a load detection grasping error analysis is performed according to the restored grasping image sequence to obtain a grasping error parameter, the second load detection data is error compensated, and the third load detection data is obtained; a load detection vibration error analysis is performed according to the vibration characteristic information to obtain a vibration error parameter, the third load detection data is error compensated, and the fourth load detection data is obtained as a high-precision load detection result. This effectively solves the problem that the load detection result of the prior art is not accurate due to the influence of shock and vibration, which further affects the detection efficiency of the robot arm during the grasping process. The load detection error caused by the vibration of the robot arm can be effectively analyzed and compensated, the detection efficiency is improved, and the operation accuracy of the robot arm in a dynamic environment is further improved.

[0051] Embodiment 2: Based on the high-precision detection method for load sensors in the above embodiments, and with the same inventive concept, the present application also provides a high-precision detection device for load sensors, see the attached Figure 2 , the high-precision detection device for the load sensor comprises:

[0052] The first load detection data acquisition module 11 is used to collect and obtain first load detection data based on a load sensor arranged in the target robotic arm when the target robotic arm grasps the target workpiece, as well as to collect grasping action parameters of the target robotic arm and a grasping image sequence of grasping the target workpiece.

[0053] The second load detection data acquisition module 12 is used to perform load detection impact error analysis according to the grabbing action parameters, obtain impact error parameters, perform error compensation on the first load detection data, and obtain second load detection data.

[0054] The vibration characteristic information acquisition module 13 is used to perform vibration analysis on the target robot arm according to the grasping action parameters to obtain vibration characteristic information, wherein the vibration characteristic information includes vibration amplitude, vibration period and vibration direction.

[0055] The image restoration module 14 is used to adjust and restore the captured image sequence according to the timestamps of the captured images in the captured image sequence and in combination with the vibration characteristic information to obtain a restored captured image sequence.

[0056] The third load detection data acquisition module 15 is used to perform load detection grabbing error analysis according to the restored grabbing image sequence, obtain grabbing error parameters, perform error compensation on the second load detection data, and obtain third load detection data.

[0057] The detection result acquisition module 16 is used to perform load detection vibration error analysis according to the vibration characteristic information, obtain vibration error parameters, perform error compensation on the third load detection data, and obtain fourth load detection data as a high-precision load detection result.

[0058] Furthermore, the second load detection data acquisition module 12 in the high-precision detection device for a load sensor is also used for:

[0059] According to the grabbing load test data record of the target robot arm, a sample grabbing action parameter set and a sample impact error parameter set are collected; the sample grabbing action parameter set and the sample impact error parameter set are used to train an impact error analyzer; the impact error analyzer is used to perform load detection impact error analysis on the grabbing action parameters to obtain impact error parameters; the impact error parameters are used to perform error compensation calculation on the first load detection data to obtain second load detection data.

[0060] Furthermore, the vibration characteristic information acquisition module 13 in the high-precision detection device for a load sensor is also used for:

[0061] According to the grasping vibration test data record of the target robotic arm, a sample grasping action parameter set and a sample vibration feature information set are collected, wherein each sample vibration feature information includes a sample vibration amplitude, a sample vibration period and a sample vibration direction; the sample grasping action parameter set and the sample vibration feature information set are used to train a vibration feature analyzer; the vibration feature analyzer is used to perform vibration analysis of the target robotic arm on the grasping action parameters to obtain vibration feature information.

[0062] Furthermore, the image restoration module 14 in the high-precision detection device for load sensors is also used for:

[0063] The starting timestamp information of the target robot arm starting the grasping action is obtained, and multiple image timestamp information is assigned to multiple grasped images in the grasped image sequence according to the starting timestamp information; multiple vibration stroke information of the multiple grasped images is assigned according to the vibration characteristic information and the multiple image timestamp information; and based on the multiple vibration stroke information, the grasped image sequence is adjusted and restored based on the generative adversarial network to obtain a restored grasped image sequence.

[0064] Furthermore, the high-precision detection device for load sensors also includes a restorer training module, which is used to:

[0065] According to the grabbing data record of the target robotic arm, a sample grabbing image set and a sample vibration stroke information set are collected, and images of the target robotic arm after vibration stops are collected to obtain a sample restored grabbing image set; based on a generative adversarial network, a grabbing image restorer is constructed, wherein the grabbing image restorer includes a generation channel and an adversarial channel; the sample grabbing image set and the sample vibration stroke information set are used as generation training data, and the sample restored grabbing image set is used as adversarial training data to train the grabbing image restorer until convergence; the multiple vibration stroke information and the multiple grabbing images are input into the grabbing image restorer to restore and generate a restored grabbing image sequence.

[0066] Furthermore, the third load detection data acquisition module 15 in the high-precision detection device for a load sensor is also used for:

[0067] According to the grabbing load test data record of the target robot arm, a set of sample restored grabbing image sequences is collected, and according to the error influence of the posture of the grabbing workpiece in different restored grabbing image sequences on the load detection, a set of sample grabbing error parameters is obtained; using the sample restored grabbing image sequence set and the sample grabbing error parameter set, a grabbing error identifier is trained based on a spatiotemporal graph convolutional network; using the grabbing error identifier, load detection grabbing error analysis is performed on the restored grabbing image sequence to obtain grabbing error parameters; using the grabbing error parameters, error compensation calculation is performed on the second load detection data to obtain third load detection data.

[0068] Furthermore, the detection result acquisition module 16 in the high-precision detection device for load sensors is also used for:

[0069] According to the grabbing load test data record of the target robot arm, a sample vibration feature information set is collected, and according to the load detection error obtained from different vibration feature information tests, a sample vibration error parameter set is obtained; the sample vibration feature information set and the sample vibration error parameter set are used to train a vibration error analyzer; the vibration error analyzer is used to perform load detection vibration error analysis on the vibration feature information to obtain vibration error parameters; the vibration error parameters are used to perform error compensation calculation on the third load detection data to obtain fourth load detection data.

[0070] The various embodiments in this specification are described in a progressive manner, and each embodiment focuses on the differences from other embodiments. Figure 1 The high-precision detection method and specific examples for load sensors in the first embodiment are also applicable to the high-precision detection device for load sensors in this embodiment. Through the above detailed description of the high-precision detection method for load sensors, those skilled in the art can clearly understand the high-precision detection device for load sensors in this embodiment, so for the sake of brevity of the specification, it will not be described in detail here. For the device disclosed in the embodiment, since it corresponds to the method disclosed in the embodiment, the description is relatively simple, and the relevant parts can be referred to the method part description.

[0071] The above description of the disclosed embodiments enables those skilled in the art to implement or use the present application. Various modifications to these embodiments will be apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present application. Therefore, the present application will not be limited to the embodiments shown herein, but will conform to the widest scope consistent with the principles and novel features disclosed herein.

[0072] Obviously, those skilled in the art can make various changes and modifications to the present application without departing from the spirit and scope of the present application. Thus, if these modifications and variations of the present application belong to the scope of the present application and its equivalent technology, the present application is also intended to include these modifications and variations.

Claims

1. A high-precision detection method for a load sensor, characterized in that: include: When the target manipulator grasps the target workpiece, based on the load sensor arranged in the target manipulator, first load detection data is acquired, as well as grasping action parameters of the target manipulator and a grasping image sequence of grasping the target workpiece; According to the grabbing action parameters, load detection impact error analysis is performed to obtain impact error parameters, and error compensation is performed on the first load detection data to obtain second load detection data; According to the grasping action parameters, a vibration analysis of the target manipulator is performed to obtain vibration characteristic information, wherein the vibration characteristic information includes vibration amplitude, vibration period and vibration direction; According to the timestamps of the captured images in the captured image sequence and in combination with the vibration characteristic information, the captured image sequence is adjusted and restored to obtain a restored captured image sequence; performing a load detection grabbing error analysis according to the restored grabbing image sequence to obtain a grabbing error parameter, performing error compensation on the second load detection data to obtain third load detection data; A load detection vibration error analysis is performed according to the vibration characteristic information to obtain a vibration error parameter, and error compensation is performed on the third load detection data to obtain fourth load detection data as a high-precision load detection result.

2. The high-precision detection method for a load sensor according to claim 1, characterized in that: According to the grabbing action parameters, load detection impact error analysis is performed to obtain impact error parameters, and error compensation is performed on the first load detection data to obtain second load detection data, including: According to the grab load test data record of the target manipulator, a sample grab action parameter set and a sample impact error parameter set are collected; Using the sample grabbing action parameter set and the sample impact error parameter set to train an impact error analyzer; Using the impact error analyzer, load detection impact error analysis is performed on the grasping action parameters to obtain impact error parameters; The impact error parameter is used to perform error compensation calculation on the first load detection data to obtain second load detection data.

3. The high-precision detection method for a load sensor according to claim 1, characterized in that: According to the grasping action parameters, a vibration analysis of the target manipulator is performed to obtain vibration characteristic information, including: According to the grabbing vibration test data record of the target manipulator, a sample grabbing action parameter set and a sample vibration feature information set are collected, wherein each sample vibration feature information includes a sample vibration amplitude, a sample vibration period and a sample vibration direction; Using the sample grabbing motion parameter set and the sample vibration feature information set to train a vibration feature analyzer; The vibration characteristic analyzer is used to perform vibration analysis of the target mechanical arm on the grasping action parameters to obtain vibration characteristic information.

4. The high-precision detection method for a load sensor according to claim 1, characterized in that: According to the timestamps of the captured images in the captured image sequence and in combination with the vibration characteristic information, the captured image sequence is adjusted and restored to obtain a restored captured image sequence, including: Acquire the starting timestamp information of the target robot arm starting the grasping action, and assign multiple image timestamp information to multiple captured images in the captured image sequence according to the starting timestamp information; According to the vibration characteristic information and the timestamp information of the multiple images, a plurality of vibration stroke information of the multiple captured images are allocated and obtained; According to the multiple vibration stroke information, based on a generative adversarial network, the captured image sequence is adjusted and restored to obtain a restored captured image sequence.

5. The high-precision detection method for a load sensor according to claim 4, characterized in that: According to the plurality of vibration stroke information, based on a generative adversarial network, adjusting and restoring the captured image sequence, comprising: According to the grabbing data record of the target mechanical arm, a sample grabbing image set and a sample vibration stroke information set are collected, and an image after the vibration of the target mechanical arm stops is collected to obtain a sample recovery grabbing image set; Based on the generative adversarial network, a captured image restorer is constructed, wherein the captured image restorer includes a generative channel and an adversarial channel; Using the sample captured image set and the sample vibration stroke information set as generated training data, using the sample restored captured image set as adversarial training data, and training the captured image restorer until convergence; The plurality of vibration stroke information and the plurality of captured images are input into the captured image restorer, and restored to generate a restored captured image sequence.

6. The high-precision detection method for a load sensor according to claim 1, characterized in that: Performing load detection grabbing error analysis according to the restored grabbing image sequence to obtain a grabbing error parameter and performing error compensation on the second load detection data, including: According to the grabbing load test data record of the target robot arm, a set of sample restored grabbing image sequences is collected, and according to the error influence of the posture of the grabbing workpiece in different restored grabbing image sequences on the load detection, a set of sample grabbing error parameters is obtained; Using the sample restored captured image sequence set and the sample captured error parameter set, based on the spatiotemporal graph convolutional network, a captured error identifier is trained; Using the grasping error identifier, performing load detection grasping error analysis on the restored grasping image sequence to obtain grasping error parameters; The grabbing error parameter is used to perform error compensation calculation on the second load detection data to obtain third load detection data.

7. The high-precision detection method for a load sensor according to claim 1, characterized in that: Performing load detection vibration error analysis according to the vibration characteristic information to obtain vibration error parameters, and performing error compensation on the third load detection data, including: According to the grabbing load test data record of the target manipulator, a sample vibration characteristic information set is collected, and according to the load detection error obtained by different vibration characteristic information tests, a sample vibration error parameter set is obtained; Using the sample vibration feature information set and the sample vibration error parameter set to train a vibration error analyzer; Using the vibration error analyzer, performing load detection vibration error analysis on the vibration characteristic information to obtain vibration error parameters; The vibration error parameter is used to perform error compensation calculation on the third load detection data to obtain fourth load detection data.

8. A high-precision detection device for a load sensor, characterized in that: The steps for implementing the high-precision detection method for a load sensor according to any one of claims 1 to 7, the high-precision detection device for a load sensor comprising: A first load detection data acquisition module is used to collect and obtain first load detection data based on a load sensor arranged in the target mechanical arm when the target mechanical arm grasps the target workpiece, and to collect grasping action parameters of the target mechanical arm and a grasping image sequence of grasping the target workpiece; A second load detection data acquisition module is used to perform load detection impact error analysis according to the grabbing action parameters, obtain impact error parameters, perform error compensation on the first load detection data, and obtain second load detection data; A vibration characteristic information acquisition module, used to perform vibration analysis of the target mechanical arm according to the grasping action parameters to obtain vibration characteristic information, wherein the vibration characteristic information includes vibration amplitude, vibration period and vibration direction; An image restoration module, configured to adjust and restore the captured images in the captured image sequence according to the timestamps of the captured images in the captured image sequence and in combination with the vibration characteristic information, so as to obtain a restored captured image sequence; a third load detection data acquisition module, configured to perform load detection grabbing error analysis according to the restored grabbing image sequence, obtain grabbing error parameters, perform error compensation on the second load detection data, and obtain third load detection data; The detection result acquisition module is used to perform load detection vibration error analysis according to the vibration characteristic information, obtain vibration error parameters, perform error compensation on the third load detection data, and obtain fourth load detection data as a high-precision load detection result.