CN112077841B - Multi-joint linkage method and system for manipulator precision of elevator robot arm - Google Patents
Multi-joint linkage method and system for manipulator precision of elevator robot arm Download PDFInfo
- Publication number
- CN112077841B CN112077841B CN202010795671.7A CN202010795671A CN112077841B CN 112077841 B CN112077841 B CN 112077841B CN 202010795671 A CN202010795671 A CN 202010795671A CN 112077841 B CN112077841 B CN 112077841B
- Authority
- CN
- China
- Prior art keywords
- model
- robot arm
- control instruction
- training
- arm
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Images
Classifications
-
- B—PERFORMING OPERATIONS; TRANSPORTING
- B25—HAND TOOLS; PORTABLE POWER-DRIVEN TOOLS; MANIPULATORS
- B25J—MANIPULATORS; CHAMBERS PROVIDED WITH MANIPULATION DEVICES
- B25J9/00—Programme-controlled manipulators
- B25J9/16—Programme controls
- B25J9/1602—Programme controls characterised by the control system, structure, architecture
- B25J9/161—Hardware, e.g. neural networks, fuzzy logic, interfaces, processor
-
- B—PERFORMING OPERATIONS; TRANSPORTING
- B25—HAND TOOLS; PORTABLE POWER-DRIVEN TOOLS; MANIPULATORS
- B25J—MANIPULATORS; CHAMBERS PROVIDED WITH MANIPULATION DEVICES
- B25J9/00—Programme-controlled manipulators
- B25J9/02—Programme-controlled manipulators characterised by movement of the arms, e.g. cartesian coordinate type
- B25J9/023—Cartesian coordinate type
-
- B—PERFORMING OPERATIONS; TRANSPORTING
- B25—HAND TOOLS; PORTABLE POWER-DRIVEN TOOLS; MANIPULATORS
- B25J—MANIPULATORS; CHAMBERS PROVIDED WITH MANIPULATION DEVICES
- B25J9/00—Programme-controlled manipulators
- B25J9/16—Programme controls
- B25J9/1628—Programme controls characterised by the control loop
- B25J9/163—Programme controls characterised by the control loop learning, adaptive, model based, rule based expert control
Landscapes
- Engineering & Computer Science (AREA)
- Robotics (AREA)
- Mechanical Engineering (AREA)
- Automation & Control Theory (AREA)
- Physics & Mathematics (AREA)
- Artificial Intelligence (AREA)
- Evolutionary Computation (AREA)
- Fuzzy Systems (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Manipulator (AREA)
Abstract
The invention provides a multi-joint linkage method and a multi-joint linkage system for the arm manipulation precision of an elevator robot, which belong to the field of information science and technology and robots, and are used for constructing and training an internal model based on a neural network, wherein the internal model comprises a forward model and a reverse model, the forward model contains the mapping relation from the joint angle of the robot arm to the direction in a Cartesian space, and the reverse model contains the mapping relation from the direction in the Cartesian space to the joint angle of the robot arm; the robot arm generates a control instruction according to the acquired target relative position based on the trained internal model, and predicts the motion direction of the arm according to the control instruction, thereby improving the precision of the robot arm.
Description
The technical field is as follows:
the invention belongs to the field of information science and technology and robots, and particularly relates to a multi-joint linkage method and system for the arm control precision of an elevator robot.
Background art:
the robot has been widely researched since its birth, and the popularity of the discussion has never been reduced, and the body image frequency of the robot appears in various literature and film and television works. In recent years, due to the heat of fire of artificial intelligence, robots closely related to the artificial intelligence also get more attention and go more into the visual field of people, more and more researchers are invested in research work related to the robots, and many people assume that the robots can replace most of the work of human beings and live together with the human beings in the future intelligent society. In fact, robots have been used in many places at present and replace part of the work of humans, such as assembly robots, disaster area rescue robots, etc., which are widely used in industry. In the future and the present stage of robots, the manipulation of the arm of the robot is a crucial issue, for example, to replace workers to operate parts on a production line, or to replace people with tools in daily life. Thus, arm manipulation of robots has become a research focus of great interest in the field of robotics.
However, in the operation of the robot arm, the problem of insufficient precision of the robot arm is often faced, for example, when parts are assembled by using the robot arm in industry, the robot arm is required to have high precision, and since the steering engine for controlling the robot arm mainly comprises a steering wheel, a reduction gear set and the like, the angle control of the steering engine is discrete, and the discrete can cause that the angle of the robot arm reaching the target position is actually inaccessible. To take an extreme example, as shown in fig. 1, a two-degree-of-freedom robot arm is shown, with an asterisk indicating the target position, assuming that the robot arm joints can only actually be controlled {0 °,30 °,60 °,90 °, and the angle of both joints to the target is 40 °, as shown in fig. 1 (a). When a control command of 40 degrees is sent to the steering engine, the steering engine can only reach 30 degrees, as shown in a (b) diagram in fig. 1, and obviously, a better solution exists that the first joint is 30 degrees, and the second joint is 60 degrees, as shown in a (c) diagram in fig. 1. As can be seen from this example, in this case, the multi-joint robot arm may have a possibility of improving the accuracy.
The invention content is as follows:
the invention aims to provide a method and a system for improving the precision of a robot arm through multi-joint linkage.
In order to achieve the purpose, the technical scheme adopted by the invention is as follows:
a multi-joint linkage method for the control precision of an arm of an elevator robot comprises the following steps:
constructing an internal model based on a neural network, wherein the internal model comprises a forward model and a reverse model, the forward model contains a mapping relation from a robot arm joint angle to a direction in a Cartesian space, and the reverse model contains a mapping relation from the direction in the Cartesian space to the robot arm joint angle;
training the inner model, including training a forward model and training a reverse model;
starting a robot arm based on a trained inner model, acquiring the direction of a target relative to the tail end of the arm based on a sensor, and generating a control instruction for the robot arm to move towards the target direction through a reverse model; and predicting the motion direction of the tail end of the robot arm after the control instruction is executed through a forward model, if the predicted motion direction is consistent with the target direction, executing the control instruction, and otherwise, detecting the target direction again.
Further, the method for training the internal model comprises the following steps: and respectively training a forward model and a reverse model of the inner model through a motion direction-control instruction pair of the training data, wherein the motion direction of the training data is input into the reverse model to generate a control instruction, and the control instruction of the training data is input into the forward model to predict the motion direction of the robot arm after execution.
Further, the motion direction-control command pair of the training data is obtained by randomly giving an angle transformation of less than 5 ° to each joint of the robot arm, and recording the displacement of the robot arm end and the joint angles before and after the motion.
Further, the control instruction is used for controlling the difference of the joint angles before and after the movement of the robot arm.
Further, the fact that the moving direction is consistent with the target direction means that the difference value between the moving direction and the target direction is smaller than a preset threshold value.
Further, the sensor comprises a camera and a laser radar.
A multi-joint linkage system for the control precision of an arm of a hoist robot comprises the robot arm, a sensor and an internal model based on a neural network; the internal model comprises a forward model and a reverse model, wherein the forward model contains a mapping relation from a robot arm joint angle to a direction in a Cartesian space, and the reverse model contains a mapping relation from the direction in the Cartesian space to the robot arm joint angle; the precision of the inner model is improved through training; acquiring the direction of a target relative to the tail end of the robot arm through the sensor, and generating a control instruction of the robot arm moving towards the target direction through a reverse model; and predicting the motion direction of the tail end of the robot arm after the control instruction is executed through a forward model, if the predicted motion direction is consistent with the target direction, executing the control instruction, and otherwise, detecting the target direction again.
Compared with the prior art, the invention has the beneficial effects that:
according to the invention, when the robot arm carries out target positioning and tracking, the traditional absolute position is changed into the relative position relative to the robot arm, so that the more accurate arm control capability can be further obtained under the condition that the precision of the robot arm is limited, the average error is reduced to a certain extent, and a new thought is provided for the robot arm to execute fine arm control tasks such as part assembly, circuit welding and the like.
Drawings
Fig. 1 is a schematic diagram of a two-degree-of-freedom robot arm.
FIG. 2 is a frame diagram of multi-joint linkage lifting accuracy.
Fig. 3 is a schematic diagram of relative positions.
Fig. 4 is a schematic view of an approach path.
The specific implementation mode is as follows:
it has been shown that the approach of human infants consists of two phases, first of all, by means of a large approach of proprioception, which is independent of vision and only depends on the perception of the infant's own arm, i.e. proprioception, and which is coarse-grained and only moves the end of the arm to the vicinity of the target position. Then, the infant accurately adjusts the position of the end of the arm according to the hand-eye coordination, and moves the hand to the target position according to the motion model established inside, thereby reaching the target position. The invention mainly refers to the second section of approach of the human baby, namely the accurate adjustment stage, and the approach process of the human baby is used for reference so as to improve the approach accuracy of the multi-joint linkage robot arm.
In the invention, motion control of the robot arm is developed by acquiring motion skills of the human baby arm, and the concept of the tail end of the robot arm and the direction in a Cartesian space is established in random turbulence (motor babbling) of the robot arm, namely, the robot knows how to move given a target motion direction and knows which direction to move given a control command. In the process, an internal model of the robot arm motion is formed respectively, and comprises a forward model (mapping of the robot arm joint angle to the direction in the Cartesian space) and a reverse model (mapping of the direction in the Cartesian space to the robot arm joint angle). Then, the direction of the target relative to the robot arm is obtained through a sensor, a robot arm control command (delta joint angle, namely the difference between the joint angles before and after movement) moving towards the direction is generated through a reverse model, and whether the control command can meet the requirement or not is predicted through a forward model. And finally, executing the control instruction meeting the requirement. The framework of the method is as shown in fig. 2, the robot obtains the relative position of the tail end of the arm of the robot and the target through sensing information, then obtains a control instruction through a reverse model, predicts the motion direction of the tail end after the control instruction is executed through a forward model, if the difference value between the predicted direction and the target direction is smaller than a certain threshold value, the robot executes the control instruction, and otherwise, detects the relative position again.
The invention builds a robot inner model development frame by using the development of the arm motor skills of the human infants. The following describes an embodiment, and in combination with the accompanying drawings, a multi-joint linkage method for improving the handling accuracy of an arm of a robot according to the present invention is specifically described.
(1) Relative position: the arm manipulation of the multi-joint arm linkage in the method is based on the relative position of the end of the arm relative to the robot and the target, not the absolute position of the target. As shown in fig. 3, the upper part is the approach of absolute position, i.e. the robot directly obtains the spatial position of the target by a sensor such as a camera, etc., and the lower part is the approach based on relative position, i.e. the robot obtains the direction of the target relative to the end of the robot arm by a sensor, and then drives the end of the arm to move toward the target direction. The approach based on the absolute position is greatly influenced by the sensing error, the inaccurate absolute position of the target can cause the inaccurate arriving position of the final robot arm end, and the arm control based on the relative position can reduce the error of the target. Thus, the multi-joint linkage arm manipulation of the following method is based on relative position.
(2) Acquiring training data: and (3) attaching a mark to the tail end of the robot arm, randomly giving a small angle (less than 5 degrees) change to each joint of the robot arm, and recording the displacement of the mark and the joint angle before and after movement. The displacement is normalized into a unit vector (namely the motion direction), and the difference between the joint angles before and after motion is a control command, so as to obtain training data, wherein the training data is a motion direction-control command pair.
(3) Training an inner model: based on the neural network model, a forward model mapping from the current joint angle and motion direction to the control command and a backward model mapping from the current joint angle and control command to the direction of motion are trained, respectively. The reverse model is directly used for obtaining a control instruction, and the forward model is used for predicting whether the result of the model is accurate or not. The motion direction-control instruction pair of the training data is used for training, wherein the reverse model takes the motion direction of the training data as input and takes the control instruction as output, the forward model is opposite to the reverse model, and takes the control instruction of the training data as input to predict the motion direction.
(4) The multi-joint linkage frame: the direction of the target position relative to the tail end of the arm is obtained based on the robot vision, and in order to avoid partial errors brought by a sensor, the absolute position is directly replaced by the relative position; calculating a control instruction of the arm through a reverse model, before executing the instruction, predicting whether the direction after executing the instruction is consistent with the target direction by using a forward model, if so, executing the instruction, and judging the direction of the next target from the tail end of the arm according to vision; and if the target direction is inconsistent with the target direction, detecting the target direction again. Compared with the prior art, the robot arm has the advantages that the relative position is used for replacing the absolute position, and meanwhile, the approach of the tail end of the robot arm is realized by combining a forward model and a reverse model.
In order to verify the effectiveness of the method, the invention carries out a simulation experiment of approaching on a two-dimensional plane on a robot arm with two degrees of freedom. The precision of a joint steering engine of the robot arm is 2 degrees, the length of an upper arm is 10cm, and the length of a lower arm is 8 cm. 80000 training data sets were collected by the robot babbling of the robot arm, and the results in 100 proximity experiments are shown in Table 1.
TABLE 1
| Kinematics | The method of the invention | |
| Mean error (cm) | 0.137 | 0.134 |
The approaching path is shown in fig. 4, the abscissa and ordinate respectively represent the x and y values on the two-dimensional plane, wherein "x" represents the target position, and "·" represents the moving path of the end of the robot arm, and the three diagrams (a), (b) and (c) in fig. 4 are respectively the approaching examples using the method of the present invention, and it can be seen that the method of the present invention can successfully make the end of the robot arm approach the target object.
The above embodiments are only intended to illustrate the technical solution of the present invention, but not to limit it, and a person skilled in the art can modify the technical solution of the present invention or substitute it with an equivalent, and the protection scope of the present invention is subject to the claims.
Claims (8)
1. A multi-joint linkage method for the control precision of an arm of an elevator robot is characterized by comprising the following steps:
constructing an internal model based on a neural network, wherein the internal model comprises a forward model and a reverse model, the forward model contains a mapping relation from a robot arm joint angle to a direction in a Cartesian space, and the reverse model contains a mapping relation from the direction in the Cartesian space to the robot arm joint angle;
training the inner model, including training a forward model and training a reverse model; the method for training the internal model comprises the following steps: training a forward model and a reverse model of an inner model through a motion direction-control instruction pair of training data, inputting the motion direction of the training data into the reverse model to output a control instruction, inputting the control instruction of the training data into the forward model to output the motion direction of the executed robot arm, and finishing training until the output motion direction-control instruction pair is consistent with the input motion direction-control instruction pair;
starting a robot arm based on a trained inner model, acquiring the direction of a target relative to the tail end of the arm through a sensor, and generating a control instruction for the robot arm to move towards the target direction through a reverse model; and predicting the motion direction of the tail end of the robot arm after the control instruction is executed through a forward model, if the predicted motion direction is consistent with the target direction, executing the control instruction, and otherwise, detecting the target direction again.
2. The method of claim 1, wherein the motion direction-control command pairs of the training data are obtained by recording the displacement of the end of the robot arm and the joint angles before and after the motion by randomly assigning each joint of the robot arm an angle change of less than 5 °.
3. The method of claim 1, wherein the control instructions are for controlling a difference between a front and rear joint angle of the robot arm movement.
4. The method of claim 1, wherein the direction of motion is consistent with the target direction by a difference less than a predetermined threshold.
5. The method of claim 1, wherein the sensor comprises a camera, a lidar.
6. A multi-joint linkage system for the control precision of an arm of a hoist robot is characterized by comprising a robot arm, a sensor and an internal model based on a neural network; the internal model comprises a forward model and a reverse model, wherein the forward model contains a mapping relation from a robot arm joint angle to a direction in a Cartesian space, and the reverse model contains a mapping relation from the direction in the Cartesian space to the robot arm joint angle; the accuracy of the inner model is improved through training, and the method for training the inner model comprises the following steps: respectively training a forward model and a reverse model of the inner model through a motion direction-control instruction pair of training data, wherein the motion direction of the training data is input into the reverse model to generate a control instruction, and the control instruction of the training data is input into the forward model to predict the motion direction of the robot arm after execution; acquiring the direction of a target relative to the tail end of the robot arm through the sensor, and generating a control instruction of the robot arm moving towards the target direction through a reverse model; and predicting the motion direction of the tail end of the robot arm after the control instruction is executed through a forward model, if the predicted motion direction is consistent with the target direction, executing the control instruction, and otherwise, detecting the target direction again.
7. The system of claim 6, wherein the training data is a motion direction-control command pair obtained by randomly assigning an angle transformation of less than 5 ° to each joint of the robot arm, and recording the displacement of the end of the robot arm and the joint angles before and after the motion.
8. The system of claim 6, wherein the sensor comprises a camera, a lidar.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010795671.7A CN112077841B (en) | 2020-08-10 | 2020-08-10 | Multi-joint linkage method and system for manipulator precision of elevator robot arm |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010795671.7A CN112077841B (en) | 2020-08-10 | 2020-08-10 | Multi-joint linkage method and system for manipulator precision of elevator robot arm |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN112077841A CN112077841A (en) | 2020-12-15 |
| CN112077841B true CN112077841B (en) | 2022-02-11 |
Family
ID=73735366
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202010795671.7A Active CN112077841B (en) | 2020-08-10 | 2020-08-10 | Multi-joint linkage method and system for manipulator precision of elevator robot arm |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN112077841B (en) |
Families Citing this family (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116079730B (en) * | 2023-02-06 | 2024-06-25 | 北京大学 | Control method and system for operation precision of arm of elevator robot |
Family Cites Families (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP5949242B2 (en) * | 2012-07-11 | 2016-07-06 | セイコーエプソン株式会社 | Robot system, robot, robot control apparatus, robot control method, and robot control program |
| CN103278164B (en) * | 2013-06-13 | 2015-11-18 | 北京大学深圳研究生院 | Robot bionic paths planning method and emulation platform under a kind of complicated dynamic scene |
| CN107972027B (en) * | 2016-10-25 | 2020-11-27 | 深圳光启合众科技有限公司 | Robot positioning method and device and robot |
| JP2018116589A (en) * | 2017-01-20 | 2018-07-26 | Kddi株式会社 | State identification device, program and method using changed image group of object image |
| CN108326850B (en) * | 2018-01-10 | 2021-07-06 | 温州大学 | Method and system for robot to accurately move mechanical arm to reach specified position |
| CN110181504B (en) * | 2018-02-22 | 2021-03-02 | 北京猎户星空科技有限公司 | Method and device for controlling mechanical arm to move and control equipment |
| CN110587596A (en) * | 2019-07-30 | 2019-12-20 | 武汉恒新动力科技有限公司 | Multi-axis configuration device remote control method and device, terminal equipment and storage medium |
| CN110524544A (en) * | 2019-10-08 | 2019-12-03 | 深圳前海达闼云端智能科技有限公司 | A kind of control method of manipulator motion, terminal and readable storage medium storing program for executing |
-
2020
- 2020-08-10 CN CN202010795671.7A patent/CN112077841B/en active Active
Also Published As
| Publication number | Publication date |
|---|---|
| CN112077841A (en) | 2020-12-15 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109848983B (en) | Method for guiding robot to cooperatively work by high-compliance person | |
| CN108838991B (en) | Autonomous humanoid double-arm robot and tracking operation system thereof for moving target | |
| US9981389B2 (en) | Robotics platforms incorporating manipulators having common joint designs | |
| Li et al. | Vision-based robotic manipulation of flexible PCBs | |
| CN110421556B (en) | Track planning method for real-time collision avoidance of redundant double-arm service robot and stable operation method | |
| Luo et al. | Real time human motion imitation of anthropomorphic dual arm robot based on Cartesian impedance control | |
| CN111230873B (en) | Teaching learning-based collaborative handling control system and method | |
| CN115469576B (en) | Teleoperation system based on human-mechanical arm heterogeneous motion space hybrid mapping | |
| CN113829343B (en) | Real-time multitasking and multi-man-machine interaction system based on environment perception | |
| Agravante et al. | Visual servoing for the REEM humanoid robot's upper body | |
| CN114571469A (en) | Zero-space real-time obstacle avoidance control method and system for mechanical arm | |
| CN112077841B (en) | Multi-joint linkage method and system for manipulator precision of elevator robot arm | |
| Ren et al. | Integrated task sensing and whole body control for mobile manipulation with series elastic actuators | |
| CN111515928A (en) | Mechanical arm motion control system | |
| Han et al. | Grasping control method of manipulator based on binocular vision combining target detection and trajectory planning | |
| Li et al. | Multi-joint active collision avoidance for robot based on depth visual perception | |
| Mu et al. | Cartesian space robot manipulator clamping movement in ROS simulation and experiment | |
| Li et al. | Human-robot collaborative manipulation with the suppression of human-caused disturbance | |
| Teke et al. | Real-time and robust collaborative robot motion control with Microsoft Kinect® v2 | |
| Ranjan et al. | Identification and control of NAO humanoid robot to grasp an object using monocular vision | |
| Ikeda et al. | On-line optimization of avoidance ability for redundant manipulator | |
| Jiao et al. | Vision based cable assembly in constrained environment | |
| Zhou et al. | Visual servo control system of 2-DOF parallel robot | |
| CN114888768A (en) | Mobile duplex robot cooperative grabbing system and method based on multi-sensor fusion | |
| Zhou et al. | A cooperative shared control scheme based on intention recognition for flexible assembly manufacturing |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |