CN112025242A - Mechanical arm hole searching method based on multilayer perceptron - Google Patents
Mechanical arm hole searching method based on multilayer perceptron Download PDFInfo
- Publication number
- CN112025242A CN112025242A CN202010876074.7A CN202010876074A CN112025242A CN 112025242 A CN112025242 A CN 112025242A CN 202010876074 A CN202010876074 A CN 202010876074A CN 112025242 A CN112025242 A CN 112025242A
- Authority
- CN
- China
- Prior art keywords
- mechanical arm
- force
- jack
- hole searching
- workpiece
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Images
Classifications
-
- B—PERFORMING OPERATIONS; TRANSPORTING
- B23—MACHINE TOOLS; METAL-WORKING NOT OTHERWISE PROVIDED FOR
- B23P—METAL-WORKING NOT OTHERWISE PROVIDED FOR; COMBINED OPERATIONS; UNIVERSAL MACHINE TOOLS
- B23P19/00—Machines for simply fitting together or separating metal parts or objects, or metal and non-metal parts, whether or not involving some deformation; Tools or devices therefor so far as not provided for in other classes
-
- 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
-
- 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
- B25J9/1687—Assembly, peg and hole, palletising, straight line, weaving pattern movement
Landscapes
- Engineering & Computer Science (AREA)
- Mechanical Engineering (AREA)
- Robotics (AREA)
- Manipulator (AREA)
Abstract
The invention discloses a mechanical arm hole searching method based on a multilayer sensor, which realizes mechanical arm hole searching by combining a top layer hole searching track planner model based on the multilayer sensor and a bottom layer force-position hybrid controller; the input of the top layer hole searching track planner model is force/moment information generated by a workpiece contact jack, and the output is the next action direction. Because the method is based on the multilayer perceptron, and in the process of collecting data, under the condition of the same position change, the force/moment characteristic change is more obvious, and the method has better anti-interference capability and higher success rate in the actual application stage of hole searching after neural network training; the invention has certain universality on a common industrial robot platform, does not need manual intervention, effectively improves the assembly efficiency, and has better adaptability to shaft hole assembly tasks.
Description
Technical Field
The invention belongs to the field of shaft hole assembly, and particularly relates to a mechanical arm hole searching method based on a multilayer sensor.
Background
Industrial robots, which are "pearls at the top of manufacturing crown", have been widely used in modern industrial automation fields such as automobile, 3C manufacturing, and ship manufacturing. The characteristics of high precision and long-time work enable the industrial robot to be commonly used for assisting human beings to complete certain work with high precision, high working strength and high repeatability. In some fields with severe working environment and threatened safety of human, the robot is widely applied. With the aging of the population and the increasing shortage of labor in China, the function of the industrial robot is more important. China pays much attention to the role of industrial robots in the manufacturing industry, and the industrial robots are already listed as key development objects in the national development planning of '2025 made in China'.
The application of the industrial robot in the field of shaft hole assembly is one of hot spots in the field of industrial robot research, and the industrial robot is widely applied in the fields of automobile tire assembly, large part assembly in the aviation industry, electronic component 3C production lines and the like. The problems to be considered by the industrial robot in the hole searching stage in the shaft hole assembly are mainly shaft hole alignment and contact force control. The alignment of the shaft holes mainly refers to the process of adjusting the position and the posture of the tail end of the mechanical arm to align the workpiece shaft clamped by the tail end of the mechanical arm with the positions of the jacks, so that relative deviation between the shaft holes is eliminated; the contact force control of the mechanical arm refers to controlling the force between a workpiece shaft clamped at the tail end of the mechanical arm and the surface where the hole is located. In the hole searching process, the mechanical arm can be in contact with the external environment, the tail end of the mechanical arm can be clamped by the small contact force to separate from the surface of the hole, the workpiece or the mechanical arm can be damaged due to the large contact force, and therefore the tail end force of the mechanical arm is controlled to be very important.
In summary, the hole searching method based on artificial intelligence in the shaft hole assembly mainly includes a hole searching control method based on a multilayer sensor. The method is based on the design of a top layer hole searching controller of a visual sensor and a multilayer perceptron (MLP), and the rough positioning of the jack position is obtained through the visual sensor and is close to the jack position. In the accurate positioning adjustment process, the mapping relation between the information obtained by the force sensor and the hole searching direction is obtained by training the multilayer perceptron, so that the motion control strategy of the next period of the mechanical arm is obtained. However, in the data acquisition process of the method, the contact acquisition method that the workpiece clamped at the tail end of the mechanical arm is relatively parallel to the surface of the position of the jack has the advantages that the acquired force/torque information has small change and unobvious characteristics under the condition that the change of the stepping position is small, and the success rate and the anti-interference capability are limited due to the interference of noise and contact force jitter in the real environment. With the rapid development of the artificial intelligence technology and the increasing application of the artificial intelligence technology in the industrial robot environment, the neural network training method is used, the input data has great influence on the training effect of the whole model, and the data is collected more effectively, so that the control effect and the success rate of the whole model are improved.
Disclosure of Invention
The invention aims to provide a mechanical arm hole searching method based on a multilayer sensor, aiming at the defects of the prior art.
The purpose of the invention is realized by the following technical scheme: a mechanical arm hole searching method based on a multilayer sensor is realized by combining a top layer hole searching track planner model based on the multilayer sensor and a bottom layer force position hybrid controller; the input of the top layer hole searching track planner model is force/moment information generated by a workpiece contact jack, and the output is the next action direction;
the force/moment information is obtained by the following steps: firstly, roughly positioning the tail end of a mechanical arm to be close to the position of a jack, and adjusting the mechanical arm to rotate a wrist joint to enable the lower surface of an assembly workpiece to form a certain angle alpha with the plane of the jack position; then the mechanical arm moves vertically downwards, and the tail end generates a contact force F; after each contact, the mechanical arm rotates the workpiece, so that the wrist joint rotates by the same angle alpha around the x axis of the wrist joint, the workpiece has the tendency of returning to the plane perpendicular to the position of the jack, the workpiece has the tendency of translational motion and the moment generated by rotational motion, and the force and moment data at the moment are collected to obtain force/moment information;
the next action direction comprises upward, downward, leftward and rightward.
Further, the angle alpha is 5-10 degrees.
Further, the contact force F is 10-15N.
Further, the top-level search hole trajectory planner model is trained by the following steps:
(1) data acquisition: firstly, the tail end of the mechanical arm is controlled to move to the center of the jack position, then the mechanical arm is controlled to traverse a series of discrete points around the center of the jack, and force/moment information [ F ] of each point is obtainedx,Fy,Mx,My](ii) a Wherein, Fx,FyContact forces in the x, y directions, Mx,MyThe moment on the x axis and the y axis of the wrist joint;
(2) data marking: marking each point acquired in the step (1) as a category label of the next action direction required for reaching the jack position, specifically: subtracting the position data of the jack center from the position data of each point to obtain the relative position d of each pointx、dy、dzLabeling the data collected by each discrete point according to the following rules:
if d isy＜dxAnd d isy＜-dxThe label category is 0, which represents moving upwards;
if d isy＞dxAnd d isy＞-dxThe label category is 1, which represents downward movement;
if d isy＞-dxAnd d isy＜dxLabel category is 2, representing left movement;
if d isy＞dxAnd d isy＜-dxThe label category is 3, representing a right shift.
(3) And (3) training a multilayer perceptron according to the data obtained in the steps (1) and (2) to obtain a top-layer hole searching track planner model.
Further, the bottom layer force position hybrid controller adopts impedance control in the direction perpendicular to the plane of the jack, and the expected force is set to be F.
Further, the top-level search hole trajectory planner model uses a ReLU function as an activation function, using a back propagation algorithm to learn training network parameters.
Further, the back propagation algorithm is an Adam algorithm.
The invention has the beneficial effects that: the present invention uses an improved data acquisition method, first moving the robotic arm vertically downward, producing a contact force of 15N. Meanwhile, after each contact, the mechanical arm rotates the tail end workpiece to rotate along the anticlockwise direction by taking the contact point as a center, namely rotates a tiny angle around the x axis of the wrist, so that the workpiece has a tendency of returning to a vertical state, and the workpiece has a tendency of not only translational motion but also moment generated by rotational motion. Compared with the method in the thesis, the acquired data has more obvious force characteristics, better anti-interference capability is achieved in the actual application stage of hole searching after neural network training, and the success rate is greatly improved. In the whole hole searching data acquisition process, manual intervention is not needed, the assembly efficiency is effectively improved, and the hole searching data acquisition process has better applicability in the whole application process.
Drawings
FIG. 1 is an overall architecture diagram of an industrial robot and a shaft hole assembly workpiece;
FIG. 2 is a schematic view of a hole searching process;
FIG. 3 is a schematic diagram illustrating a sample scattering point distribution of search data;
FIG. 4 is a data phase end pose schematic;
FIG. 5 is a schematic diagram of a multi-layer perceptron network configuration;
FIG. 6 is a force-bit hybrid controller control block diagram;
in fig. 1, an industrial robot actuator 1, an external force sensor 2, an assembly workpiece 3, a jack position 4, an industrial camera 5;
in fig. 2, an industrial robot initial position 6, a contact state 7, a hole searching stage 8 and an insertion stage 9;
in fig. 3, jack locations 10, jack centers 11, and discrete sampling points 12;
in fig. 4, the original data acquisition method workpiece pose 13, contact point Q14, end workpiece axis center point 15;
in fig. 5, the input layer 16, the hidden layer 17, the output layer 18;
in fig. 6, a top level trajectory planner 19, a robot position loop 20, an impedance controller 21, a robot velocity loop 22, a coordinate transformation 23, a gravity compensation 24, a force sensor 25.
Detailed Description
The invention is further described below with reference to the accompanying drawings.
The structure of the industrial robot shaft hole assembly is shown in fig. 1 and mainly comprises an industrial robot actuator 1, an external force sensor 2, an assembly workpiece 3, a jack position 4 and an industrial camera 5. Wherein, the wrist part at the tail end of the industrial robot executor 1 is provided with an external force sensor 2 which is used for measuring the force/moment information of the contact between the tail end clamping workpiece and the environment. The industrial robot actuator 1 grips the assembly workpiece 3 at the end, and has a task of searching the assembly workpiece 3 to the insertion hole position 4 and inserting the assembly workpiece 3 to the insertion hole position 4. The robot arm can obtain coarse positioning information of the fitting workpiece 3 and the jack position 4 by the guidance of the industrial camera 5, thereby moving the fitting workpiece 3 to the vicinity of the jack position 4. The hole searching method will be described in detail below.
The top layer track planner control method based on the multilayer perceptron (MLP) in shaft hole assembly combines the top layer track planner and the bottom layer force position hybrid controller, and the bottom layer force position hybrid controller enables the tail end of an industrial robot actuator 1 to be kept in safe and stable contact with the environment and is used for hole searching tasks. And (3) obtaining a mapping relation model of force/moment information and the next action direction through training of a multilayer perceptron (MLP), and using the mapping relation model as a top-layer trajectory planner to predict the action direction of the workpiece in the next process. The method is proved to be effective and reliable in the application of high-precision shaft hole assembly tasks.
The steps of the MLP-based top-level trajectory planner training method and the bottom-level force-bit hybrid controller control method in the implementation process are explained in detail below, and the overall hole searching flow chart of the present invention is shown in fig. 2, and comprises the following processes in total: the initial position 6 is reached, the mechanical arm is driven to enable the assembled workpiece and the jack position to be in a contact state 7, the hole searching stage 8 and the inserting stage 9.
The process of obtaining force/moment information and the next action direction through the training of the multilayer perceptron (MLP) comprises the aspects of data acquisition, data labeling, neural network building and training and model storage. The MLP multilayer perceptron also belongs to one of the artificial neural networks, but the multilayer perceptron (MLP) can be applied to the non-linear separable occasions. One set of input vectors is mapped into another set of output vectors after being processed by the multi-layer perceptron, and the input vectors are transmitted to the hidden layer and then to the output layer for gradual processing. MLP is used to train a four-classifier with the input being force/torque data collected by the end-of-arm sensor and the output being the direction of the hole search taken next by the arm.
In the data acquisition phase, as shown in fig. 3, the sampling object includes a jack position 10, a jack center 11, and discrete sampling points 12. The data acquisition process for the jack location 10 is as follows: firstly, the mechanical arm is controlled to move to the center 11 of the jack (accurate positioning), namely, the shaft hole has no relative pose deviation, then the mechanical arm is controlled to traverse a series of discrete sampling points 12 around the hole, and the state information of each discrete sampling point 12 is recorded. According to practical application experience, the rough positioning error of the jack position 10 is generally within 10mm, so the method controls the mechanical arm to traverse a square area which takes the hole as the center and has the side length of 20 mm. Where the step increment traversed is 0.2mm, 101 x 101 pieces of data can be finally collected for the training task of the MLP multi-layer perceptron-based top-level trajectory planner.
In particular, as shown in fig. 4, the wrist joint of the robot arm rotates at a slight angle to the plane of the receptacle location 10 as the robot arm traverses the discrete sampling points 12. The robot arm is first moved vertically downward to bring the assembled workpiece into contact with the socket location at contact point Q14, resulting in a contact force of 15N. Meanwhile, after each contact, the mechanical arm rotates the tail end workpiece to rotate along the anticlockwise direction by taking the contact point Q14 as a center, namely rotates a tiny angle around the x axis of the wrist joint, so that the workpiece has a tendency of returning to a vertical state, the workpiece has a tendency of translational motion and moment generated by rotational motion, and force/moment data at the moment are collected and recorded. The characteristic change of the force/moment data acquired by the method is more obvious, and compared with the workpiece pose 13 acquired by the original data acquisition method, the method is more beneficial to training a multilayer perceptron to obtain a top-level planner model with better performance.
A data labeling stage, in which force/moment information [ F ] obtained by each discrete sampling point 12 is used according to the data acquired in the data acquisition stagex,Fy,Mx,My]As input data, and marks the class labels of the sample points according to the position data. The sampling points are divided into four categories: i.e., up, down, left, right. The four categories represent four directions of movement of the robotic arm during the hole search. Fx,FyContact forces in the x, y directions, Mx,MyThe moment on the x-axis and the y-axis of the wrist joint.
The specific method for labeling is as follows, firstly, the initial position data is subtracted from the collected position data to obtain the relative position d in the xyz directionx、dy、dzThe data collected at each discrete sample point 12 is then labeled according to the following rules:
if d isy＜dxAnd d isy＜-dxThe label category is 0, which represents moving upwards;
if d isy＞dxAnd d isy＞-dxThe label category is 1, which represents downward movement;
if d isy＞-dxAnd d isy＜dxLabel category is 2, representing left movement;
if d isy＞dxAnd d isy＜-dxThe label category is 3, representing a right shift.
In the neural network building and training stage, a schematic diagram of a multi-layer perceptron model is shown in fig. 5, and the multi-layer perceptron model comprises an input layer 16, a hidden layer 17 and an output layer 18. Based on a multilayer perceptron (MLP), a neural network with a network structure of 4-100-50-4, namely a four-input four-output classifier model, is constructed. There are two hidden layers 16, each with a number of neurons of 100 and 50, respectively. The input data is 4-dimensional vector Fx,Fy,Mx,My]The output is class labels 0,1,2 and 3 respectivelyRepresenting the next motion direction of the mechanical arm. Meanwhile, normalization processing needs to be carried out on input data, and neural network training is facilitated. The multi-layer perceptron uses a ReLU function as an activation function and uses an Adam algorithm in a back propagation algorithm to learn and train artificial neural network parameters.
After training, the artificial neural network returns a classifier model based on MLP, and the model is stored in a disk file to provide guidance for a subsequent hole searching process.
In the hole searching process, the control block diagram of the force-position hybrid controller is shown in fig. 6, and includes a robot position ring 20, an impedance controller 21, and a robot speed ring 22. The top track planner 19 outputs the action direction of the mechanical arm at the next moment to the robot position ring 20 according to the input of the force sensor 25, so that the workpiece clamped at the tail end of the mechanical arm generates translation motion in the x and y directions. The impedance controller 21 is used for applying a force of 15N in the Z-axis direction at the tail end of the mechanical arm to ensure that the assembly workpiece 3 is stably contacted with the jack position 4. When the mechanical arm is started, the force sensor 25 obtains Z-axis force data, gravity compensation 24 is carried out on the assembly workpiece 3, and the influence of the self weight of the assembly workpiece on high-precision shaft hole assembly is eliminated.
The top-layer planner model and the bottom-layer force position hybrid controller model which are obtained through training and are based on the multilayer perceptron are combined, and therefore hole searching work of high-precision shaft hole assembly can be adapted.
In the process of searching and debugging the hole of the industrial robot, the invention mainly focuses on the debugging of variables such as the learning rate of a multilayer sensor, the parameters of a force position hybrid controller, the searching step length of a workpiece at the tail end of a mechanical arm during data acquisition and the like, and simultaneously needs to acquire the parameters such as the control period, the repeatable precision and the like of the platform of the industrial robot.
Claims (7)
1. A mechanical arm hole searching method based on a multilayer perceptron is characterized in that the hole searching method is realized by combining a top layer hole searching track planner model based on the multilayer perceptron, a bottom layer force and position hybrid controller and the like. The input of the top layer hole searching track planner model is force/moment information generated by a workpiece contact jack, and the output is the next action direction.
The force/moment information is obtained by the following steps: firstly, roughly positioning the tail end of the mechanical arm to be close to the position of the jack, and adjusting the mechanical arm to rotate the wrist joint to enable the lower surface of the assembly workpiece to form a certain angle alpha with the plane of the position of the jack. Then the mechanical arm moves vertically downwards, and the tail end generates a contact force F; after each contact, the mechanical arm rotates the workpiece, so that the wrist joint rotates by the same angle alpha around the x axis of the wrist joint, the workpiece has the tendency of returning to the plane perpendicular to the position of the jack, the workpiece has the tendency of translational motion and the moment generated by rotational motion, and the force and moment data at the moment are collected to obtain force/moment information.
The next action direction comprises upward, downward, leftward and rightward.
2. The mechanical arm hole searching method based on the multilayer perceptron as claimed in claim 1, wherein the angle α is 5-10 degrees.
3. The multi-layer sensor-based mechanical arm hole searching method as claimed in claim 1, wherein the contact force F is 10-15N.
4. The multi-layer perceptron-based mechanical arm hole searching method as claimed in claim 1, wherein the top layer hole searching trajectory planner model is trained by the following steps:
(1) data acquisition: firstly, the tail end of the mechanical arm is controlled to move to the center of the jack position, then the mechanical arm is controlled to traverse a series of discrete points around the center of the jack, and force/moment information [ F ] of each point is obtainedx,Fy,Mx,My](ii) a Wherein, Fx,FyContact forces in the x, y directions, Mx,MyThe moment on the x axis and the y axis of the wrist joint;
(2) data marking: marking each point acquired in the step (1) as a category label of the next action direction required for reaching the jack position, specifically: subtracting the position data of the jack center from the position data of each point to obtain the relative position d of each pointx、dy、dzPush-buttonLabeling the data collected at each discrete point according to the following rules:
if d isy＜dxAnd d isy＜-dxThe label category is 0, which represents moving upwards;
if d isy＞dxAnd d isy＞-dxThe label category is 1, which represents downward movement;
if d isy＞-dxAnd d isy＜dxLabel category is 2, representing left movement;
if d isy＞dxAnd d isy＜-dxThe label category is 3, representing a right shift.
(3) And (3) training a multilayer perceptron according to the data obtained in the steps (1) and (2) to obtain a top-layer hole searching track planner model.
5. The multi-layer perceptron-based robotic arm hole searching method of claim 1, wherein the bottom layer force position mixing controller employs impedance control in a direction perpendicular to a plane of the jack, setting the desired force to F.
6. The multi-layered perceptron-based robotic hole search method of claim 1, wherein the top-level hole search trajectory planner model uses a ReLU function as an activation function, using a back propagation algorithm to learn training network parameters.
7. The multi-layer perceptron-based robotic arm hole searching method of claim 6, wherein the back propagation algorithm is an Adam algorithm.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010876074.7A CN112025242A (en) | 2020-08-27 | 2020-08-27 | Mechanical arm hole searching method based on multilayer perceptron |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010876074.7A CN112025242A (en) | 2020-08-27 | 2020-08-27 | Mechanical arm hole searching method based on multilayer perceptron |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN112025242A true CN112025242A (en) | 2020-12-04 |
Family
ID=73580211
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202010876074.7A Pending CN112025242A (en) | 2020-08-27 | 2020-08-27 | Mechanical arm hole searching method based on multilayer perceptron |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN112025242A (en) |
Cited By (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112847235A (en) * | 2020-12-25 | 2021-05-28 | 山东大学 | Robot step force guiding assembly method and system based on deep reinforcement learning |
| CN113747682A (en) * | 2021-09-07 | 2021-12-03 | 山东大学 | Robot precision assembling system and method |
| CN114147724A (en) * | 2021-12-20 | 2022-03-08 | 上海景吾智能科技有限公司 | Robot force-controlled shaft hole assembling method and system |
| CN114952830A (en) * | 2022-05-13 | 2022-08-30 | 哈尔滨工业大学 | Robot high-precision shaft hole assembly shaft part positioning method based on contact constraint |
| CN115592671A (en) * | 2022-11-03 | 2023-01-13 | 哈尔滨工业大学（Cn） | Robot tail end contact positioning method based on deep learning |
| WO2024156228A1 (en) * | 2023-01-29 | 2024-08-02 | 深圳先进技术研究院 | Peg-in-hole assembly method and system, electronic device, and storage medium |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JPH04348887A (en) * | 1991-05-23 | 1992-12-03 | B L Oototetsuku Kk | Device for inserting rod member into insertion hole and method thereof |
| CN101041220A (en) * | 2006-03-22 | 2007-09-26 | 中国科学院自动化研究所 | Method for realizing the assembly of shaft hole having high-precision by using robot having low precision |
| CN106272416A (en) * | 2016-08-29 | 2017-01-04 | 上海交通大学 | Feel based on power and the robot slender axles Fine Boring system and method for vision |
| CN110293552A (en) * | 2018-03-21 | 2019-10-01 | 北京猎户星空科技有限公司 | Mechanical arm control method, device, control equipment and storage medium |
| CN110449882A (en) * | 2019-08-02 | 2019-11-15 | 珞石（北京）科技有限公司 | The search assembly method of binding force control |
| US20200114506A1 (en) * | 2017-12-05 | 2020-04-16 | Google Llc | Viewpoint invariant visual servoing of robot end effector using recurrent neural network |
-
2020
- 2020-08-27 CN CN202010876074.7A patent/CN112025242A/en active Pending
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JPH04348887A (en) * | 1991-05-23 | 1992-12-03 | B L Oototetsuku Kk | Device for inserting rod member into insertion hole and method thereof |
| CN101041220A (en) * | 2006-03-22 | 2007-09-26 | 中国科学院自动化研究所 | Method for realizing the assembly of shaft hole having high-precision by using robot having low precision |
| CN106272416A (en) * | 2016-08-29 | 2017-01-04 | 上海交通大学 | Feel based on power and the robot slender axles Fine Boring system and method for vision |
| US20200114506A1 (en) * | 2017-12-05 | 2020-04-16 | Google Llc | Viewpoint invariant visual servoing of robot end effector using recurrent neural network |
| CN110293552A (en) * | 2018-03-21 | 2019-10-01 | 北京猎户星空科技有限公司 | Mechanical arm control method, device, control equipment and storage medium |
| CN110449882A (en) * | 2019-08-02 | 2019-11-15 | 珞石（北京）科技有限公司 | The search assembly method of binding force control |
Non-Patent Citations (1)
| Title |
|---|
| 郑养龙: "基于力传感器的双臂机器人轴孔柔顺装配策略与方法研究", 《中国优秀硕士学位论文全文数据库 信息科技辑》 * |
Cited By (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112847235A (en) * | 2020-12-25 | 2021-05-28 | 山东大学 | Robot step force guiding assembly method and system based on deep reinforcement learning |
| CN113747682A (en) * | 2021-09-07 | 2021-12-03 | 山东大学 | Robot precision assembling system and method |
| CN113747682B (en) * | 2021-09-07 | 2023-09-12 | 山东大学 | Robot precise assembly system and method |
| CN114147724A (en) * | 2021-12-20 | 2022-03-08 | 上海景吾智能科技有限公司 | Robot force-controlled shaft hole assembling method and system |
| CN114147724B (en) * | 2021-12-20 | 2024-04-16 | 上海景吾智能科技有限公司 | Robot power control shaft hole assembly method and system |
| CN114952830A (en) * | 2022-05-13 | 2022-08-30 | 哈尔滨工业大学 | Robot high-precision shaft hole assembly shaft part positioning method based on contact constraint |
| CN115592671A (en) * | 2022-11-03 | 2023-01-13 | 哈尔滨工业大学（Cn） | Robot tail end contact positioning method based on deep learning |
| CN115592671B (en) * | 2022-11-03 | 2024-08-06 | 哈尔滨工业大学 | Robot tail end contact positioning method based on deep learning |
| WO2024156228A1 (en) * | 2023-01-29 | 2024-08-02 | 深圳先进技术研究院 | Peg-in-hole assembly method and system, electronic device, and storage medium |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN112025242A (en) | Mechanical arm hole searching method based on multilayer perceptron | |
| CN109382828B (en) | Robot shaft hole assembling system and method based on teaching learning | |
| CN109782601B (en) | Design method of self-adaptive neural network synchronous robust controller of coordinated mechanical arm | |
| CN106584093A (en) | Self-assembly system and method for industrial robots | |
| CN109591019B (en) | Space accurate positioning method for nondeterministic positioning characteristic object | |
| CN113146620A (en) | Binocular vision-based double-arm cooperative robot system and control method | |
| Karayiannidis et al. | “Open sesame!” adaptive force/velocity control for opening unknown doors | |
| CN112192614A (en) | Man-machine cooperation based shaft hole assembling method for nuclear operation and maintenance robot | |
| CN108445764B (en) | Active compliance control strategy of Stewart platform | |
| CN114310954A (en) | Self-adaptive lifting control method and system for nursing robot | |
| Klas et al. | The kit gripper: A multi-functional gripper for disassembly tasks | |
| Hsieh et al. | Deep convolutional generative adversarial network for inverse kinematics of self-assembly robotic arm based on the depth sensor | |
| Liu et al. | Adaptive lead-through teaching control for spray-painting robot with closed control system | |
| Apriaskar et al. | Robotic technology towards industry 4.0: Automatic object sorting robot arm using kinect sensor | |
| CN118288294B (en) | Robot vision servo and man-machine cooperative control method based on image variable admittance | |
| CN115918377B (en) | Control method and control device of automatic tree fruit picking machine and automatic tree fruit picking machine | |
| CN117245666A (en) | Dynamic target quick grabbing planning method and system based on deep reinforcement learning | |
| Zhang et al. | Research on algorithm of humanoid robot arm control system based on fuzzy PID control | |
| CN115741723A (en) | Precision compensation method for multi-degree-of-freedom snake-shaped mechanical arm | |
| Golnazarian et al. | Intelligent industrial robots | |
| Chuang et al. | Automatic vision-based optical fiber alignment using multirate technique | |
| Deng et al. | Design of an all-position intelligent welding robot system | |
| Yong | RESEARCH ON PATH RECOGNITION OF WELDING MANIPULATOR BASED ON AUTOMATIC CONTROL ALGORITHM. | |
| Wan et al. | Learning optimal measurement and control of assembly robot for large-scale heavy-weight parts | |
| Kim et al. | A Study on Real Time Control of Moving Stuff Action Through Iterative Learning for Mobile-Manipulator System |
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 |