CN108527371A - A kind of Dextrous Hand planing method based on BP neural network - Google Patents
A kind of Dextrous Hand planing method based on BP neural network Download PDFInfo
- Publication number
- CN108527371A CN108527371A CN201810343226.XA CN201810343226A CN108527371A CN 108527371 A CN108527371 A CN 108527371A CN 201810343226 A CN201810343226 A CN 201810343226A CN 108527371 A CN108527371 A CN 108527371A
- Authority
- CN
- China
- Prior art keywords
- neural network
- function
- planing method
- dextrous hand
- output
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
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
- B25J9/1612—Programme controls characterised by the hand, wrist, grip control
-
- B—PERFORMING OPERATIONS; TRANSPORTING
- B25—HAND TOOLS; PORTABLE POWER-DRIVEN TOOLS; MANIPULATORS
- B25J—MANIPULATORS; CHAMBERS PROVIDED WITH MANIPULATION DEVICES
- B25J15/00—Gripping heads and other end effectors
- B25J15/0009—Gripping heads and other end effectors comprising multi-articulated fingers, e.g. resembling a human hand
-
- B—PERFORMING OPERATIONS; TRANSPORTING
- B25—HAND TOOLS; PORTABLE POWER-DRIVEN TOOLS; MANIPULATORS
- B25J—MANIPULATORS; CHAMBERS PROVIDED WITH MANIPULATION DEVICES
- B25J9/00—Programme-controlled manipulators
- B25J9/16—Programme controls
- B25J9/1656—Programme controls characterised by programming, planning systems for manipulators
- B25J9/1661—Programme controls characterised by programming, planning systems for manipulators characterised by task planning, object-oriented languages
Abstract
The present invention provides a kind of planing methods of humanoid dextrous hand, include the following steps:Step S1, BP neural network training make error function decline along gradient direction by the amendment of amendment and threshold value to network weight;The input of the BP neural network is to capture the affine not bending moment of target, is exported as crawl targeted species corresponding with crawl gesture;Step S2 identifies unknown pattern, judges corresponding grasp mode by trained BP neural network and provide each joint angles of apery hand.The method of the present invention disclosure satisfy that crawl and operational requirements of the robot for the tool of different type, size.
Description
Technical field
The present invention relates to a kind of multiple degrees of freedom humanoid dextrous hand mode of operation planing methods, especially in the process of grasping can
The operation planning for enough adapting to different crawl targets, belongs to robot system technical field.
Background technology
Robot is as one of greatest invention of 20th century mankind, by long-term development, applies in various fields, especially
It is in dangerous field, including space, water and subterranean resource exploitation etc., instead of or assist the mankind work.In face of complicated labor
Power driven tools, robot need the hand of a pair of dexterity to complete various tasks.
Recent decades emerge large quantities of mechanical Dextrous Hands for being integrated with most advanced science and technology both at home and abroad, these mechanical Dextrous Hands
Some have been successfully applied in military and space field, or even have some mechanical Dextrous Hand to take the lead in marching toward industry market and artifucial limb neck
Domain.The theoretical consummate simple and flexible of these Dextrous Hands or complicated and changeable.
Aaron M.Dollar and Robert D.Howe proposed a kind of very simple manipulator in 2010 and set
Concept is counted, although this manipulator can not complete fine operation, this design concept enables to manipulator to crawl
Target show high adaptability.German Space Agency has developed apery hand DLR Hand Arm System, this hand
Flexibility is very high, but thus brings 38 motors and complicated kinematics burden.The robot astronaut that NASA is researched and developed
Robonaut 2 possesses the manipulator of 12 degree of freedom of a pair of.Shadow companies started the five fingers humanoid dexterous of research and development in 2004
Hand amounts to 24 joints close to human hand in shape.Humanoid dextrous hand mechanism is complicated, and degree of freedom is more, completes crawl or was operating
Cheng Zhong, it is a difficult point to carry out planning to each joint.
Invention content
The purpose of the present invention is to provide a kind of planing methods based on BP neural network, humanoid dextrous hand can be made as people
Hand equally uses miscellaneous tool, completes operation task..
Technical scheme is as follows.
A kind of planing method of humanoid dextrous hand, includes the following steps:
Step S1, BP neural network training make error function along ladder by the amendment of amendment and threshold value to network weight
Direction is spent to decline;The input of the BP neural network is to capture the affine not bending moment of target, and it is corresponding with crawl gesture to export
Crawl targeted species;
Step S2, identify unknown pattern, by trained BP neural network judge corresponding grasp mode and to
Go out each joint angles of apery hand.
Preferably, the BP neural network includes input layer, hidden layer, output layer.
Preferably, the transforming function transformation function of the hidden layer is nonlinear function.
Preferably, the nonlinear function is S type functions or hyperbolic tangent function.
Preferably, the transforming function transformation function of the output layer is nonlinear or linear.
Preferably, by carrying out feature extraction to image, not displacement feature is obtained;These features are inputted into network respectively,
Specimen sample training is carried out, sample is then carried out and completely trains;The output of the BP networks is to belong to the degree of membership of each type objects,
The maximum output node of its output valve corresponds to the specific object category of an output, obtains recognition result.
Preferably, the input of the BP neural network includes tool types t, the minimum envelop radius of circle in the tool plan
R and tool folding condition b.
Preferably, the output of the BP neural network is defined as position of the imitation human finger point end with respect to wrist, apery hand
Refer to each joint angles to solve to obtain by the computation of inverse- kinematics.
Preferably, the transfer function between the input layer and the hidden layer is log-sigmoid functions, described implicit
Transfer function between layer and the output layer is linear function, f (x)=x.
Preferably, in an iterative process, study of the step length changing method to BP neural network is used according to the variation of error function
Rate η is adjusted:
Wherein E (n) is the error function of BP neural network.
By above technical scheme, the present invention is capable of providing multiple degrees of freedom humanoid dextrous hand mode of operation in the process of grasping
Planning, meet robot for different type, size tool crawl and operational requirements.
Description of the drawings
Fig. 1 is the BP neural network configuration schematic diagram of the present invention.
Fig. 2 (a)-Fig. 2 (b) is to capture long tubular object schematic diagram using the planing method control Dextrous Hand of the present invention.
Fig. 3 (a)-Fig. 3 (b) is the planing method control Dextrous Hand crawl square objects schematic diagram using the present invention.
Fig. 4 (a)-Fig. 4 (b) is the planing method control Dextrous Hand crawl plate-shaped body schematic diagram using the present invention.
Specific implementation mode
The planing method of humanoid dextrous hand provided by the invention includes the following steps:
Step S1, BP neural network training make error function along ladder by the amendment of amendment and threshold value to network weight
Direction is spent to decline;The input of the BP neural network is to capture the affine not bending moment of target, and it is corresponding with crawl gesture to export
Crawl targeted species;
Step S2, identify unknown pattern, by trained BP neural network judge corresponding grasp mode and to
Go out each joint angles of apery hand.
The typical structure for the BP neural network that the present invention uses is as shown in Figure 1.The transforming function transformation function of wherein hidden layer is generally
Nonlinear function, such as S type functions or hyperbolic tangent function.The transforming function transformation function of output layer can be nonlinear, can also be
It is linear, depending on this needs by inputting, exporting mapping relations.
The guiding theory of BP neural network study is that the amendment of amendment and threshold value to network weight makes error function edge
Gradient direction declines.The tool identification neural network input be tool affine not bending moment, export for crawl gesture phase
Corresponding tool kind.
Three node layer of BP networks is expressed as:Input layer has M node, hidden layer to have Q node, ωjkBe input layer and
Connection weight between hidden layer node, output layer have L node, ωijIt is the connection weight between hidden layer and output node layer
Value, the input of hidden layer and output node layer are the weighted sums of the output of previous node layer, and the incentive degree of each node is by it
Excitation function determine.
By carrying out feature extraction to image, not displacement feature is obtained.These features are inputted into network respectively, carry out sample
Then sample train is carried out sample and is completely trained.As continuous 5 E<When 0.01, training terminates.When frequency of training is more than 5000 times
When be failure to train.It is possible that deviation, causes the feature of object cannot be accurate when in practice due to Shape Feature Extraction
It obtains, in order to increase the robustness of system, the output of BP networks is to belong to the degree of membership of each type objects, and output valve is maximum defeated
Go out the specific object category that contact corresponds to an output, obtains recognition result.If maximum value be less than threshold value 0.6, to object into
Second of feature extraction of row simultaneously identifies, to further determine that target.If the maximum value of output is thrown away less than 0.6, judge in scene
There is no target to be identified.If BP neural network output node exists simultaneously multiple maximum values, refuse to judge.
The determination of BP neural network model parameter is illustrated below.
Illustrate input definition first.
When carrying out space station instrument and meter installation and maintenance, need to use a variety of different tools, and tool kind is determined
Determine the gesture of apery hand crawl and operation.Therefore, tool is classified, tool types t (t=1,2,3) is used as BP nerve nets
One input of network.It is clipper tool when wherein t=1, is bevel-type tool when t=2, when t=3 is gun-type tool.
And even same tool, due to the difference of size, the stretching degree of palm is also different.Therefore, by tool sizes
Also as an input of neural network.Since mainly in plane sizes, there were significant differences for tool sizes, difference in thickness is little, because
This input of this tool sizes can be set as the minimum envelop radius of circle R in the tool plan.Minimum envelop circle only envelope apery
Hand needs the place captured, i.e., clipper tool is only using handle size as input, the size without considering pliers point.
In addition, the folding condition of clipper tool can also influence the angle of finger grip, therefore, by tool folding condition b
Input as neural network.6 grades will be divided by 0 ° to 180 °, and be equal to 0 to the 6 folding angles for representing tool, b with b
Tool is represented when equal to 0 to be closed completely, for bevel-type and gun-type tool, b 0.
Illustrate output definition below.
The output of neural network is defined as position of the imitation human finger point end with respect to wrist, and each joint angles of finger are then
It solves to obtain by the computation of inverse- kinematics.
Illustrate that network structure is set below.
The determination of input layer number:Input layer includes tool kind, minimum envelop radius of circle folding condition, therefore input layer
Number of nodes m be 3.
The determination of output layer number of nodes:Output layer number of nodes is related to apery hand practical structures.It, will to simplify neural network
Output is set as the folding condition of apery hand operating gesture and hand.According to the classification to robot astronaut's mode of operation, to each
A operating gesture is encoded, and has 15 types;By the folding condition of hand by being closed completely dividing for completely separable progress 1~6
Grade.Therefore output layer number of nodes n=16.
The determination of node in hidden layer:According to the Neural network empirical functional expression of single layer hidden layerWithBetween determining that node in hidden layer s, wherein a are 1 to 10
Integer.Rule of thumb, node in hidden layer is set as 10.
The determination of transfer function:Transfer function between input layer and hidden layer is log-sigmoid functions, hidden layer
Transfer function between output layer is linear function, f (x)=x.
Illustrate that network weight adjusts below.
The method of the present invention uses step length changing method to BP nerve nets in an iterative process, according to the variation of error function E (n)
The learning rate η adjustment of network, expression formula are as follows.
In order to ensure to learn to effectively correct every time, accelerate convergence process, therefore limit log-sigmoid functions
Output, expression formula are as follows.
The initialization value of BP neural network parameter is:α=0.8;ωij、θjIt is randomly choosed between 0~1.
The training process for the BP neural network that the present invention uses is described below.
Training sample when apery hand operates different tools is obtained using three-dimensional software.First with three-dimensional visualization entity
Simulation softward Autodesk Inventor establish the threedimensional model of tool and apery hand, then acquire out and grab in the three-dimensional model
Pose of the end of each finger tip of apery hand relative to wrist when taking the tool of variety classes, size, folding condition.By these samples
This implementation completely training.As continuous 5 E<When 0.01, training terminates.It is considered as failure to train when frequency of training is more than 5000 times.
In order to increase the robustness of system, the output of BP networks is to belong to the degree of membership of each grasp mode, if output valve maximum value is less than
Threshold value 0.6 then carries out the selection of second of grasp mode, to further determine that target;If output maximum value still less than 0.6,
Then judge that the tool is not similar with known means;If BP neural network output node exists simultaneously multiple maximum values, refusal is sentenced
It is disconnected.
The experimentation of the Dextrous Hand planing method control crawl based on BP neural network using the present invention is described below.
Tool and apery hand model are established in adams softwares, after tool shape, size and state be input to matlab
In software, pose of the finger tip relative to wrist is obtained by trained nerve network system, is obtained by the computation of inverse- kinematics
Go out each articulation angle of finger.Then the angle being calculated is input in adams, control apery hand is moved, complete
At grasping manipulation.
It is that the long cartridge type object of selection carries out crawl emulation experiment shown in Fig. 2 (a)-Fig. 2 (b).In experimental tool, hand
The tools such as electricity, handle are mostly cylindrical shape.
From Fig. 2 (a) as can be seen that when Dextrous Hand captures long cartridge type object, each joint of finger can carry out
Bending can largely fit to long cartridge type body surface to reach finger member.As can be seen that finger exists from Fig. 2 (b)
In motion process, each joint angles variation is continuous, and saltus step does not occur.
It is that selection square objects carry out crawl emulation experiment shown in Fig. 3 (a)-Fig. 3 (b).Square objects are in experimental tool
In relatively conventional, such as box.
As can be seen that when it is square objects to capture target from Fig. 3 (a), finger is without image of Buddha snatching cylindrical object one
Sample makes finger joint as far as possible close to target object, and therefore, finger is captured according to the posture that neural network obtains so that target object is most
Possible envelope is within the scope of finger grip, as shown in Fig. 3 (b).
It is that selection plate-shaped body carries out crawl emulation experiment shown in Fig. 4 (a)-Fig. 4 (b).In common tool, plate object
Body occurs frequently as operation object, such as plank, brick.
As can be seen that being directed to plate-shaped body from Fig. 4 (a), finger can not completely will be including target envelope.To capture
More securely, by the way of making end finger joint be contacted as far as possible with target.At crawl end it can be seen from Fig. 4 (b)
Phase, finger tips articulation angle is close to 180 degree.
Claims (10)
1. a kind of planing method of humanoid dextrous hand, includes the following steps:
Step S1, BP neural network training, by the amendment of amendment and threshold value to network weight, makes error function along gradient side
To decline;The input of the BP neural network is to capture the affine not bending moment of target, is exported as gesture is corresponding grabs with crawl
Take targeted species;
Step S2 identifies unknown pattern, judges corresponding grasp mode by trained BP neural network and provide imitative
Each joint angles of human hand.
2. the planing method of humanoid dextrous hand according to claim 1, which is characterized in that the BP neural network includes defeated
Enter layer, hidden layer, output layer.
3. the planing method of humanoid dextrous hand according to claim 2, which is characterized in that the transforming function transformation function of the hidden layer
For nonlinear function.
4. the planing method of humanoid dextrous hand according to claim 3, which is characterized in that the nonlinear function is S types
Function or hyperbolic tangent function.
5. the planing method of humanoid dextrous hand according to claim 2, which is characterized in that the transforming function transformation function of the output layer
It is nonlinear or linear.
6. the planing method of humanoid dextrous hand according to claim 2, which is characterized in that carried by carrying out feature to image
It takes, obtains not displacement feature;These features are inputted into network respectively, carry out specimen sample training, sample is then carried out and completely instructs
Practice;The output of the BP networks is to belong to the degree of membership of each type objects, and the maximum output node of output valve corresponds to an output
Specific object category, obtain recognition result.
7. the planing method of humanoid dextrous hand according to claim 6, which is characterized in that the input of the BP neural network
Including the minimum envelop radius of circle R and tool folding condition b in tool types t, the tool plan.
8. the planing method of humanoid dextrous hand according to claim 6, which is characterized in that the output of the BP neural network
It is defined as position of the imitation human finger point end with respect to wrist, each joint angles of imitation human finger are solved by the computation of inverse- kinematics
It arrives.
9. the planing method of humanoid dextrous hand according to claim 6, which is characterized in that the input layer is implied with described
Transfer function between layer is log-sigmoid functions, and the transfer function between the hidden layer and the output layer is linear
Function, f (x)=x.
10. the planing method of humanoid dextrous hand according to claim 6, which is characterized in that in an iterative process, according to accidentally
The variation of difference function is adjusted the learning rate η of BP neural network using step length changing method:
Wherein E (n) is the error function of BP neural network.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201810343226.XA CN108527371A (en) | 2018-04-17 | 2018-04-17 | A kind of Dextrous Hand planing method based on BP neural network |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201810343226.XA CN108527371A (en) | 2018-04-17 | 2018-04-17 | A kind of Dextrous Hand planing method based on BP neural network |
Publications (1)
Publication Number | Publication Date |
---|---|
CN108527371A true CN108527371A (en) | 2018-09-14 |
Family
ID=63481050
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201810343226.XA Pending CN108527371A (en) | 2018-04-17 | 2018-04-17 | A kind of Dextrous Hand planing method based on BP neural network |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN108527371A (en) |
Cited By (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN111993416A (en) * | 2020-07-30 | 2020-11-27 | 浙江大华技术股份有限公司 | Method, equipment, system and device for controlling movement of mechanical arm |
CN112809680A (en) * | 2021-01-26 | 2021-05-18 | 山东大学 | Robot grabbing mode selection method and system based on multiple constraint conditions |
CN113478478A (en) * | 2021-06-16 | 2021-10-08 | 北京控制工程研究所 | Uncertain object-oriented perception execution interaction natural compliance human-like manipulation method |
CN116652940A (en) * | 2023-05-19 | 2023-08-29 | 兰州大学 | Human hand imitation precision control method and device, electronic equipment and storage medium |
Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20060015216A1 (en) * | 2003-10-08 | 2006-01-19 | Masato Ito | Information processing apparatus and method, program storage medium, and program |
CN1785608A (en) * | 2005-11-10 | 2006-06-14 | 上海大学 | Control platform of multifinger mechanical skillful closed ring real time action |
KR20070072314A (en) * | 2005-12-31 | 2007-07-04 | 고려대학교 산학협력단 | Method for controlling pose of robot with using neural network, recording medium thereof, apparatus for controlling pose of robot with using neuron-network and robot therewith |
CN106078748A (en) * | 2016-06-13 | 2016-11-09 | 北京理工大学 | A kind of robot based on hands, eye, arm coordinated manipulation captures the control method of object |
CN106393102A (en) * | 2015-07-31 | 2017-02-15 | 发那科株式会社 | Machine learning device, robot system, and machine learning method |
CN107030722A (en) * | 2017-06-16 | 2017-08-11 | 深圳市大寰机器人科技有限公司 | A kind of robot delicate |
-
2018
- 2018-04-17 CN CN201810343226.XA patent/CN108527371A/en active Pending
Patent Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20060015216A1 (en) * | 2003-10-08 | 2006-01-19 | Masato Ito | Information processing apparatus and method, program storage medium, and program |
CN1785608A (en) * | 2005-11-10 | 2006-06-14 | 上海大学 | Control platform of multifinger mechanical skillful closed ring real time action |
KR20070072314A (en) * | 2005-12-31 | 2007-07-04 | 고려대학교 산학협력단 | Method for controlling pose of robot with using neural network, recording medium thereof, apparatus for controlling pose of robot with using neuron-network and robot therewith |
CN106393102A (en) * | 2015-07-31 | 2017-02-15 | 发那科株式会社 | Machine learning device, robot system, and machine learning method |
CN106078748A (en) * | 2016-06-13 | 2016-11-09 | 北京理工大学 | A kind of robot based on hands, eye, arm coordinated manipulation captures the control method of object |
CN107030722A (en) * | 2017-06-16 | 2017-08-11 | 深圳市大寰机器人科技有限公司 | A kind of robot delicate |
Non-Patent Citations (6)
Title |
---|
孙毅刚等: ""一种三维物体的视觉识别方法研究"", 《中国民航学院学报》 * |
孟令启等: "《中厚板轧机测量测试与力学行为建模》", 31 August 2006 * |
张蕾等: ""采用神经网络算法的多指机械手织物抓取规划"", 《纺织学报》 * |
贵州电网有限责任公司: "《电动汽车充电管理云平台与智能充电技术》", 31 December 2017, 贵州大学出版社有限责任公司 * |
赵妍: "《面向大数据的挖掘方法研究》", 31 July 2016, 电子科技大学出版社 * |
黄媛婕: ""机器人多指手预抓取模式分类方法研究"", 《中国优秀硕士学位论文全文数据库(电子期刊)信息科技辑》 * |
Cited By (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN111993416A (en) * | 2020-07-30 | 2020-11-27 | 浙江大华技术股份有限公司 | Method, equipment, system and device for controlling movement of mechanical arm |
CN112809680A (en) * | 2021-01-26 | 2021-05-18 | 山东大学 | Robot grabbing mode selection method and system based on multiple constraint conditions |
CN113478478A (en) * | 2021-06-16 | 2021-10-08 | 北京控制工程研究所 | Uncertain object-oriented perception execution interaction natural compliance human-like manipulation method |
CN113478478B (en) * | 2021-06-16 | 2022-08-12 | 北京控制工程研究所 | Uncertain object-oriented perception execution interaction natural compliance human-like manipulation method |
CN116652940A (en) * | 2023-05-19 | 2023-08-29 | 兰州大学 | Human hand imitation precision control method and device, electronic equipment and storage medium |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN108527371A (en) | A kind of Dextrous Hand planing method based on BP neural network | |
CN108972494B (en) | Humanoid manipulator grabbing control system and data processing method thereof | |
Li et al. | Asymmetric bimanual control of dual-arm exoskeletons for human-cooperative manipulations | |
Yang et al. | Biologically inspired motion modeling and neural control for robot learning from demonstrations | |
CN108638054B (en) | Control method for intelligent explosive disposal robot five-finger dexterous hand | |
CN106346485B (en) | The Non-contact control method of bionic mechanical hand based on the study of human hand movement posture | |
Navarro et al. | Haptic object recognition for multi-fingered robot hands | |
CN106909216A (en) | A kind of Apery manipulator control method based on Kinect sensor | |
Wu et al. | Coordinated control of a dual-arm robot for surgical instrument sorting tasks | |
Su et al. | Robust grasping for an under-actuated anthropomorphic hand under object position uncertainty | |
Neha et al. | Grasp analysis of a four-fingered robotic hand based on matlab simmechanics | |
Hinwood et al. | Towards the design of a human-inspired gripper for textile manipulation | |
Lin et al. | Grasp mapping using locality preserving projections and knn regression | |
Kang et al. | A grasp abstraction hierarchy for recognition of grasping tasks from observation | |
Zhang et al. | FunctionalGrasp: Learning Functional Grasp for Robots via Semantic Hand-Object Representation | |
Rodriguez et al. | Learning postural synergies for categorical grasping through shape space registration | |
Zhu et al. | A robotic semantic grasping method for pick-and-place tasks | |
Jiang et al. | An inverse-kinematics table-based solution of a humanoid robot finger with nonlinearly coupled joints | |
Xinyu et al. | Research on robot dexterous grasping for aerospace product parts | |
TAŞAR et al. | Modeling, Simulation and Control of Prosthetic Hand using Sim Mechanics | |
Wei et al. | A Wearable Robotic Hand for Hand-over-Hand Imitation Learning | |
CN111300421A (en) | Mapping method applied to simulation of actions of both hands of humanoid robot | |
Vatsal et al. | Augmenting vision-based grasp plans for soft robotic grippers using reinforcement learning | |
Yao et al. | Analysis and grasp strategy modeling for underactuated multi-fingered robot hand | |
Krechetov et al. | Approach to the study of kinematics and modeling grip of 22 DOF anthropomorphic gripping manipulator |
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 | ||
RJ01 | Rejection of invention patent application after publication |
Application publication date: 20180914 |
|
RJ01 | Rejection of invention patent application after publication |