CN112306060B - Training gait control method based on deep learning - Google Patents
Training gait control method based on deep learning Download PDFInfo
- Publication number
- CN112306060B CN112306060B CN202011108791.1A CN202011108791A CN112306060B CN 112306060 B CN112306060 B CN 112306060B CN 202011108791 A CN202011108791 A CN 202011108791A CN 112306060 B CN112306060 B CN 112306060B
- Authority
- CN
- China
- Prior art keywords
- training
- robot
- data
- deep learning
- gait
- 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.)
- Expired - Fee Related
Links
- 230000005021 gait Effects 0.000 title claims abstract description 55
- 238000000034 method Methods 0.000 title claims abstract description 32
- 238000013135 deep learning Methods 0.000 title claims abstract description 30
- 239000011248 coating agent Substances 0.000 claims abstract description 23
- 238000000576 coating method Methods 0.000 claims abstract description 23
- 230000005291 magnetic effect Effects 0.000 claims description 46
- XLYOFNOQVPJJNP-UHFFFAOYSA-N water Substances O XLYOFNOQVPJJNP-UHFFFAOYSA-N 0.000 claims description 45
- 238000010521 absorption reaction Methods 0.000 claims description 33
- 230000005389 magnetism Effects 0.000 claims description 32
- 238000005461 lubrication Methods 0.000 claims description 26
- 229920000742 Cotton Polymers 0.000 claims description 22
- 239000000835 fiber Substances 0.000 claims description 19
- 239000004005 microsphere Substances 0.000 claims description 17
- 239000012528 membrane Substances 0.000 claims description 15
- 230000008447 perception Effects 0.000 claims description 15
- 238000013136 deep learning model Methods 0.000 claims description 12
- 238000013527 convolutional neural network Methods 0.000 claims description 6
- 238000004088 simulation Methods 0.000 claims description 3
- 239000000126 substance Substances 0.000 claims description 3
- 230000002159 abnormal effect Effects 0.000 abstract description 13
- 230000001050 lubricating effect Effects 0.000 abstract description 13
- 230000000694 effects Effects 0.000 abstract description 12
- 239000003921 oil Substances 0.000 description 27
- 239000010687 lubricating oil Substances 0.000 description 9
- 239000000463 material Substances 0.000 description 8
- 238000005299 abrasion Methods 0.000 description 5
- 238000000354 decomposition reaction Methods 0.000 description 5
- 238000001514 detection method Methods 0.000 description 3
- 230000005489 elastic deformation Effects 0.000 description 3
- 239000007789 gas Substances 0.000 description 3
- 238000013473 artificial intelligence Methods 0.000 description 2
- 238000013528 artificial neural network Methods 0.000 description 2
- 210000004556 brain Anatomy 0.000 description 2
- 238000001125 extrusion Methods 0.000 description 2
- 230000002349 favourable effect Effects 0.000 description 2
- 230000008595 infiltration Effects 0.000 description 2
- 238000001764 infiltration Methods 0.000 description 2
- 238000010801 machine learning Methods 0.000 description 2
- 238000010586 diagram Methods 0.000 description 1
- 239000007884 disintegrant Substances 0.000 description 1
- 230000005294 ferromagnetic effect Effects 0.000 description 1
- 239000007769 metal material Substances 0.000 description 1
- 230000008450 motivation Effects 0.000 description 1
- 230000035515 penetration Effects 0.000 description 1
- 239000002243 precursor Substances 0.000 description 1
- 230000001960 triggered effect Effects 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G05—CONTROLLING; REGULATING
- G05D—SYSTEMS FOR CONTROLLING OR REGULATING NON-ELECTRIC VARIABLES
- G05D1/00—Control of position, course, altitude or attitude of land, water, air or space vehicles, e.g. using automatic pilots
- G05D1/02—Control of position or course in two dimensions
- G05D1/021—Control of position or course in two dimensions specially adapted to land vehicles
- G05D1/0259—Control of position or course in two dimensions specially adapted to land vehicles using magnetic or electromagnetic means
-
- G—PHYSICS
- G05—CONTROLLING; REGULATING
- G05D—SYSTEMS FOR CONTROLLING OR REGULATING NON-ELECTRIC VARIABLES
- G05D1/00—Control of position, course, altitude or attitude of land, water, air or space vehicles, e.g. using automatic pilots
- G05D1/02—Control of position or course in two dimensions
- G05D1/021—Control of position or course in two dimensions specially adapted to land vehicles
- G05D1/0212—Control of position or course in two dimensions specially adapted to land vehicles with means for defining a desired trajectory
- G05D1/0221—Control of position or course in two dimensions specially adapted to land vehicles with means for defining a desired trajectory involving a learning process
-
- G—PHYSICS
- G05—CONTROLLING; REGULATING
- G05D—SYSTEMS FOR CONTROLLING OR REGULATING NON-ELECTRIC VARIABLES
- G05D1/00—Control of position, course, altitude or attitude of land, water, air or space vehicles, e.g. using automatic pilots
- G05D1/02—Control of position or course in two dimensions
- G05D1/021—Control of position or course in two dimensions specially adapted to land vehicles
- G05D1/0276—Control of position or course in two dimensions specially adapted to land vehicles using signals provided by a source external to the vehicle
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Automation & Control Theory (AREA)
- Remote Sensing (AREA)
- Radar, Positioning & Navigation (AREA)
- Aviation & Aerospace Engineering (AREA)
- Theoretical Computer Science (AREA)
- Computing Systems (AREA)
- Biophysics (AREA)
- General Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Data Mining & Analysis (AREA)
- General Engineering & Computer Science (AREA)
- Computational Linguistics (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Evolutionary Computation (AREA)
- Electromagnetism (AREA)
- Biomedical Technology (AREA)
- Artificial Intelligence (AREA)
- Life Sciences & Earth Sciences (AREA)
- Health & Medical Sciences (AREA)
- Toys (AREA)
- Manipulator (AREA)
Abstract
The invention discloses a training gait control method based on deep learning, belonging to the technical field of robot control, which can construct a gait model of a target robot on the basis of deep learning to realize the all-round walking of the robot and collect various related data, wherein the gait model comprises wear data of a training coating pre-coated at a joint of the robot, the joint wear is firstly introduced into a training learning sample as objective data, the interference factor and error control of the gait control of the robot are judged through abnormal pressure or wear, more accurate wear data can be obtained in a mode of directly positioning a worn area, and the predictable wear in the training process or even the actual motion process can be reduced through lubricating and protecting a worn part, thereby realizing the flexible and accurate control of the gait of the robot, the gait control effect of the robot is greatly improved.
Description
Technical Field
The invention relates to the technical field of robot control, in particular to a training gait control method based on deep learning.
Background
Gait refers to the motion process of each leg of the robot according to a certain sequence and track, and is an important factor for ensuring the stable operation of the walking mechanism. The translational gait means that the robot always keeps the body to translate when walking. The fixed point turning gait refers to the gait of the robot body rotating around a certain axis. Gait planning involves determining the onset and termination of the motion of the bearing phase, the swing phase. In order to maintain the stability of the body during the gait generation process, the robot is required to be ensured to be in a supporting phase state at least three feet during the walking process. Meanwhile, in order to ensure that the prototype has better stability in the walking process, the adjacent walking feet cannot be in the swing phase state at the same time, namely the adjacent feet of the robot cannot start to swing at the same time.
Deep learning is one of machine learning, and machine learning is a must-pass path for implementing artificial intelligence. The concept of deep learning is derived from the research of artificial neural networks, and a multi-layer perceptron comprising a plurality of hidden layers is a deep learning structure. Deep learning forms a more abstract class or feature of high-level representation properties by combining low-level features to discover a distributed feature representation of the data. The motivation for studying deep learning is to build neural networks that simulate the human brain for analytical learning, which mimics the mechanism of the human brain to interpret data such as images, sounds, text, and the like.
Although the deep learning is widely applied to the field of artificial intelligence, gait control interference factors of the robot are complex, so that the requirement of the traditional deep learning on data is very strict, and particularly, the data is difficult to train and learn by aiming at the abrasion of joints of the robot, so that errors of the gait control of the robot are difficult to effectively improve.
Disclosure of Invention
1. Technical problem to be solved
Aiming at the problems in the prior art, the invention aims to provide a training gait control method based on deep learning, which can construct a gait model of a target robot on the basis of the deep learning to realize the all-around walking of the robot and collect various related data, wherein the training gait control method comprises the wear data of a training coating pre-coated at the joint of the robot, firstly introduces the joint wear into a training learning sample as objective data, judges the interference factor and error control of the gait control of the robot through abnormal pressure or wear, can obtain more accurate wear data in a mode of directly positioning a worn area, and can reduce the foreseeable wear in the training process and even the actual motion process by lubricating and protecting the worn part, thereby realizing the flexible and accurate control of the gait of the robot, meanwhile, the feedback information of the wear data is utilized to assist error control and abnormity detection, and the gait control effect of the robot is greatly improved.
2. Technical scheme
In order to solve the above problems, the present invention adopts the following technical solutions.
A training gait control method based on deep learning comprises the following steps:
s1, constructing a gait model of the target robot on the simulation platform to realize the all-around walking of the robot;
s2, collecting motion requirement data, motion target data, various state data of the robot, wear data and external environment data, storing the data into a training database, and providing a training sample;
s3, constructing a deep learning framework based on the convolutional neural network, and training the convolutional neural network by using training samples in a training database to obtain a deep learning model;
s4, acquiring real-time motion requirement data, motion target data, various state data of the robot, wear data and external environment data, inputting the data into the deep learning model, and acquiring an output result of the deep learning model;
and S5, generating a control signal in real time according to the output result of the deep learning model by using the constructed gait model to control the gait of the robot in real time.
Furthermore, the gait model of the target robot adopts a simplified 6-freedom-degree connecting rod model, and the foot part adopts a plane sole structure.
Further, the state data of the robot comprises self shape data, motion state data and stress data of each component of the robot.
Further, the wear data is of a training coating, and the training coating is applied to each joint of the robot.
Further, be connected with a plurality of evenly distributed's perception hemisphere on the training coating surface, perception hemisphere internal surface is connected with inlays the pipe that holds in the training coating, still inlay the volume of disintegrating post that is connected with a plurality of evenly distributed in the training coating, and the volume of disintegrating post is located and holds between the pipe, hold and be connected with many evenly distributed's water guide fibre between the volume of disintegrating post, the abnormal pressure and the wearing and tearing of joint department are carried out the perception through the perception hemisphere, then according to the perception feedback to holding in the pipe, borrow the water guide fibre to trigger its adjacent volume of disintegrating post's the action of disintegrating, not only can acquire wearing and tearing data, can lubricate the protection to joint department simultaneously.
Further, the perception hemisphere is including magnetism inhale skin, elasticity hemisphere and extension rod, and magnetism inhale outer cover and connect in elasticity hemisphere surface, the extension rod is connected with magnetism and is inhaled outer inner, and the extension rod runs through elasticity hemisphere and extends to and hold intraductally, magnetism inhale the skin can adsorb the lubricated microballon of magnetism with the cooperation of the volume of disintegration post after the disintegration, can avoid the lubricated microballon of magnetism to drop on the one hand and be difficult to retrieve, and can accurate positioning wearing and tearing region, on the other hand can lubricate the protection to wearing and tearing region through the lubricated microballon of magnetism, the elasticity hemisphere can utilize elastic deformation ability to provide the extension action of extension rod to holding intraductally, and then trigger follow-up procedure.
Further, it is connected with a plurality of sponge water absorption balls corresponding with the water guide fiber to inlay in the holding tube, and the sponge water absorption balls run through the holding tube and extend to the sponge water absorption balls, and different sponge water absorption balls correspond different abnormal pressure and degree of wear, and can not receive the interference each other, can release absorptive moisture after the sponge water absorption balls receive the extrusion, then carry to the volume of breaking up post through the water guide fiber and force it to take place corresponding disintegration.
Further, the disintegration beam includes a plurality of disintegration units corresponding with the water guide fiber, the disintegration unit includes water-proof membrane piece, a plurality of magnetic lubrication microballons and collapses and deblocks, the piece that disintegrates is filled between a pair of water-proof membrane piece, and the magnetic lubrication microballon inlays in the piece that disintegrates, and water-proof membrane piece plays the effect of keeping apart moisture infiltration, avoids mutual interference between the adjacent disintegration unit, and the piece that disintegrates can react rapidly after meetting moisture and release a large amount of gases, not only can break the water-proof membrane piece in the outside, and on the other hand can drive the magnetic lubrication microballon and remove to the external world, is favorable to the magnetic lubrication microballon to be inhaled the better absorption of outer layer and lubricate the protection by the magnetism.
Further, the magnetic lubrication microsphere comprises a magnetic shell, an oil absorption cotton layer and a core ball which expands when being heated, wherein the core ball which expands when being heated is arranged inside the magnetic shell, the oil absorption cotton layer is wrapped between the core ball which expands when being heated and the magnetic shell, a plurality of uniformly distributed holes are formed in the magnetic shell, the magnetic shell has magnetism and can provide lubrication protection effect, the core ball which expands when being heated can extrude the oil absorption cotton layer to release absorbed lubricating oil, so that the lubrication protection effect of the magnetic lubrication microsphere is improved, the heat source of the core ball which expands when being heated has two parts, one part is from heat generated by friction, and the other part is from heat released when the disintegration block reacts.
Furthermore, the cotton layer surface of oil absorption is connected with a plurality of oil material matters of leading that correspond with the hole, and leads the oil material and extend to and trade oilhole mouth department, not only can derive the lubricating oil that the cotton layer of oil absorption released the magnetism shell outside and lubricate, can suck back lubricating oil back after the cotton layer of oil absorption reconversion simultaneously, reduces the loss of consumption.
3. Advantageous effects
Compared with the prior art, the invention has the advantages that:
(1) the scheme can construct a gait model of a target robot on the basis of deep learning to realize the all-around walking of the robot and collect various related data, wherein the gait model comprises wear data of a training coating pre-coated at a joint of the robot, joint wear is firstly introduced into a training and learning sample as objective data, interference factors and error control of the gait control of the robot are judged through abnormal pressure or wear, more accurate wear data can be obtained in a mode of directly positioning a worn area, and predictable wear in the training process or even in the actual motion process can be reduced by lubricating and protecting a worn part, so that the gait of the robot can be flexibly and accurately controlled, and error control and abnormal detection are assisted by feedback information of the wear data, the gait control effect of the robot is greatly improved.
(2) Be connected with a plurality of evenly distributed's perception hemisphere on the training coating surface, perception hemisphere internal surface is connected with inlays the pipe that holds in the training coating, still inlay in the training coating and be connected with a plurality of evenly distributed's the volume of disintegrating post, and the volume of disintegrating post is located and holds between the pipe, hold and be connected with many evenly distributed's water guide fiber between the volume of disintegrating post, the abnormal pressure and the wearing and tearing of department to the joint are carried out the perception through the perception hemisphere, then according to the perception feedback to holding in the pipe, trigger the action of disintegrating of its adjacent volume of disintegrating post by means of water guide fiber, not only can acquire wearing and tearing data, can lubricate the protection to joint department simultaneously.
(3) The perception hemisphere is including magnetism inhale the skin, elasticity hemisphere and extension rod, and magnetism inhale outer cover and connect in elasticity hemisphere surface, the extension rod is connected with magnetism and is inhaled outer inner, and the extension rod runs through the elasticity hemisphere and extends to in the holding tube, magnetism is inhaled the skin and can be adsorbed the lubricated microballon of magnetism with the cooperation of the volume of disintegration after the disintegration, can avoid the lubricated microballon of magnetism to drop on the one hand and be difficult to retrieve, and can accurate positioning wearing and tearing region, on the other hand can lubricate the protection to wearing and tearing region through the lubricated microballon of magnetism, the elasticity hemisphere can utilize elastic deformation ability, thereby provide the extension action of extension rod in to holding tube, and then trigger follow.
(4) The holding tube is internally embedded with a plurality of sponge water absorption balls corresponding to the water guide fibers, the sponge water absorption balls penetrate through the holding tube and extend to the sponge water absorption balls, different sponge water absorption balls correspond to different abnormal pressures and abrasion degrees and cannot be interfered with each other, the sponge water absorption balls can release absorbed moisture after being extruded, and then the sponge water absorption balls are conveyed to the disintegration amount column through the water guide fibers to force the disintegration to correspondingly disintegrate.
(5) The disintegration beam includes a plurality of disintegration units corresponding with the water guide fibre, the disintegration unit includes water proof membrane piece, a plurality of magnetic lubrication microballons and collapses the piece, the piece that disintegrates is filled between a pair of water proof membrane piece, and the magnetic lubrication microballon is inlayed in the piece that disintegrates, water proof membrane piece plays the effect of keeping apart moisture infiltration, avoid mutual interference between the adjacent disintegration unit, the piece that disintegrates can react rapidly after meetting moisture and release a large amount of gases, not only can break the water proof membrane piece in the outside, on the other hand can drive the magnetic lubrication microballon and remove to the external world, be favorable to the magnetic lubrication microballon to be inhaled the better absorption of outer layer and lubricate the protection by magnetism.
(6) The magnetic lubricating microsphere comprises a magnetic shell, an oil absorbing cotton layer and a core ball expanding when being heated, wherein the core ball expanding when being heated is arranged inside the magnetic shell, the oil absorbing cotton layer is wrapped between the core ball expanding when being heated and the magnetic shell, a plurality of uniformly distributed holes are formed in the magnetic shell, the magnetic shell has magnetism and can provide lubricating protection effect, the core ball expanding when being heated can extrude the oil absorbing cotton layer to release absorbed lubricating oil, so that the lubricating protection effect of the magnetic lubricating microsphere is improved, the heat source of the core ball expanding when being heated has two parts, one part is from heat generated by friction, and the other part is from heat released when the disintegrating block reacts.
(7) The cotton layer surface of oil absorption is connected with a plurality of oil material of leading that correspond with the hole, and leads the drill way department that the oil material extends to the hole, not only can derive the lubricating oil that the cotton layer of oil absorption released the magnetism shell outside and lubricate, can suck back lubricating oil back at the cotton layer reconversion of oil absorption simultaneously, reduces the loss of consumption.
Drawings
FIG. 1 is a schematic flow diagram of the present invention;
FIG. 2 is a schematic structural view of a training coating of the present invention;
FIG. 3 is a schematic view of a sensing hemisphere according to the present invention;
FIG. 4 is a top view of the inventive precursor solution column;
FIG. 5 is a schematic structural view of a magnetic lubricating microsphere of the present invention;
FIG. 6 is a schematic view of the lubricating protection state of the training coating according to the present invention.
The reference numbers in the figures illustrate:
1 sensing hemisphere, 11 magnetic outer layer, 12 elastic hemisphere, 13 extension rod, 2 containing tube, 3 water guide fiber, 4 decomposer volume column, 41 water-proof membrane, 42 magnetic lubricating microsphere, 421 magnetic shell, 422 oil absorption cotton layer, 423 core ball expanding by heat, 424 holes, 425 oil guide substance, 43 disintegration block, 5 sponge water absorption ball.
Detailed Description
The technical solution in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention; it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all embodiments, and all other embodiments obtained by those skilled in the art without any inventive work are within the scope of the present invention.
In the description of the present invention, it should be noted that the terms "upper", "lower", "inner", "outer", "top/bottom", and the like indicate orientations or positional relationships based on those shown in the drawings, and are only for convenience of description and simplification of description, but do not indicate or imply that the referred device or element must have a specific orientation, be constructed in a specific orientation, and be operated, and thus should not be construed as limiting the present invention. Furthermore, the terms "first" and "second" are used for descriptive purposes only and are not to be construed as indicating or implying relative importance.
In the description of the present invention, it should be noted that, unless otherwise explicitly specified or limited, the terms "mounted," "disposed," "sleeved/connected," "connected," and the like are to be construed broadly, e.g., "connected," which may be fixedly connected, detachably connected, or integrally connected; can be mechanically or electrically connected; they may be connected directly or indirectly through intervening media, or they may be interconnected between two elements. The specific meanings of the above terms in the present invention can be understood in specific cases to those skilled in the art.
Example 1:
referring to fig. 1, a training gait control method based on deep learning includes the following steps:
s1, constructing a gait model of the target robot on the simulation platform to realize the all-around walking of the robot;
s2, collecting motion requirement data, motion target data, various state data of the robot, wear data and external environment data, storing the data into a training database, and providing a training sample;
s3, constructing a deep learning framework based on the convolutional neural network, and training the convolutional neural network by using training samples in a training database to obtain a deep learning model;
s4, acquiring real-time motion requirement data, motion target data, various state data of the robot, wear data and external environment data, inputting the data into the deep learning model, and acquiring an output result of the deep learning model;
and S5, generating a control signal in real time according to the output result of the deep learning model by using the constructed gait model to control the gait of the robot in real time.
The gait model of the target robot adopts a simplified 6-freedom-degree connecting rod model, and the foot adopts a planar sole structure.
The state data of the robot comprises the self form data, the motion state data and the stress data of each part of the robot.
The wear data is of a training coating applied at each joint of the robot.
Please refer to fig. 2, the outer surface of the training coating is connected with a plurality of sensing hemispheres 1 which are uniformly distributed, the inner surface of the sensing hemispheres 1 is connected with a containing pipe 2 which is embedded in the training coating, a plurality of decomposition quantity columns 4 which are uniformly distributed are also embedded and connected in the training coating, the decomposition quantity columns 4 are positioned between the containing pipes 2, a plurality of water guide fibers 3 which are uniformly distributed are connected between the containing pipes 2 and the decomposition quantity columns 4, the sensing hemispheres 1 sense abnormal pressure and abrasion of joints, and then the abnormal pressure and the abrasion are fed back to the containing pipes 2 according to the sensing, the decomposition action of the adjacent decomposition quantity columns 4 is triggered by the water guide fibers 3, and not only abrasion data can be acquired, but also the joints can be lubricated and protected.
Please refer to fig. 3, the sensing hemisphere 1 includes magnetism and inhales outer 11, elasticity hemisphere 12 and extension rod 13, and magnetism is inhaled outer 11 cover and is connected in elasticity hemisphere 12 surface, extension rod 13 is inhaled outer 11 inner with magnetism and is connected, and extension rod 13 runs through elasticity hemisphere 12 and extends to in holding pipe 2, magnetism is inhaled outer 11 and can be adsorbed magnetic lubrication microsphere 42 with the cooperation of the solution volume post 4 after disassembling, can avoid magnetic lubrication microsphere 42 to drop on the one hand and be difficult to retrieve, and can accurate positioning wearing and tearing region, on the other hand can lubricate the protection to wearing and tearing region through magnetic lubrication microsphere 42, elasticity hemisphere 12 can utilize elastic deformation ability, thereby provide the extension action of extension rod 13 in to holding pipe 2, and then trigger follow-up procedure.
It has a plurality of sponge that correspond with water guide fiber 3 to inlay in holding pipe 2 and absorbs water ball 5, and sponge absorbs water ball 5 runs through and holds pipe 2 and extend to sponge and absorb water in ball 5, different sponge absorb water ball 5 and correspond different abnormal pressure and degree of wear, and can not receive the interference each other, absorb water ball 5 and receive to release absorptive moisture after the extrusion at the sponge, then carry to the disintegration volume post 4 through water guide fiber 3 and force it to take place corresponding disintegration.
Referring to fig. 4, the disintegration measure column 4 includes a plurality of disintegration units corresponding to the water guide fiber 3, each disintegration unit includes a water-stop membrane 41, a plurality of magnetic lubrication microspheres 42 and a disintegration block 43, the disintegration block 43 is made of the existing effervescent disintegrant, the disintegration block 43 is filled between a pair of water-stop membranes 41, and the magnetic lubrication microspheres 42 are embedded in the disintegration block 43, the water-stop membranes 41 play a role in isolating water penetration, so as to avoid mutual interference between adjacent disintegration units, the disintegration block 43 can rapidly react to release a large amount of gas after encountering water, not only can the water-stop membranes 41 on the outer side be flushed, but also can drive the magnetic lubrication microspheres 42 to move to the outside, thereby facilitating the magnetic lubrication microspheres 42 to be better adsorbed by the magnetic outer layer 11 for lubrication protection.
Referring to fig. 5-6, the magnetic lubricating microsphere 42 includes a magnetic outer shell 421, an oil absorbing cotton layer 422 and a core ball 423 that expands when heated, the magnetic outer shell 421 is made of ferromagnetic metal material, the core ball 423 that expands when heated is made of thermal expansion material, the core ball 423 that expands when heated is disposed inside the magnetic outer shell 421, the oil absorbing cotton layer 422 is wrapped between the core ball 423 that expands when heated and the magnetic outer shell 421, the magnetic outer shell 421 is formed with a plurality of uniformly distributed holes 424, the magnetic outer shell 421 has magnetism and can provide lubrication protection, the core ball 423 that expands when heated can extrude the oil absorbing cotton layer 422 to release absorbed lubricating oil, thereby improving the lubrication protection effect of the magnetic lubricating microsphere 42, the heat source of the core ball 423 that expands when heated has two parts, one part is from heat generated by friction, and the other part is from heat released when the block 43 reacts.
The outer surface of the oil absorption cotton layer 422 is connected with a plurality of oil guiding materials 425 corresponding to the holes 424, the oil guiding materials 425 extend to the openings of the holes 424, the oil guiding materials 425 can guide lubricating oil released by the oil absorption cotton layer 422 out of the magnetic shell 421 for lubrication, and meanwhile the lubricating oil can be sucked back in a reverse mode after the oil absorption cotton layer 422 recovers, and consumption loss is reduced.
The invention can construct a gait model of a target robot on the basis of deep learning to realize the all-round walking of the robot and collect various related data, wherein the data comprises the wear data of a training coating pre-coated at the joint of the robot, the joint wear is firstly introduced into a training and learning sample as objective data, the interference factor and the error control of the gait control of the robot are judged by abnormal pressure or wear, more accurate wear data can be obtained by directly positioning a worn area, and the foreseeable wear in the training process or even the actual motion process can be reduced by lubricating and protecting a worn part, thereby realizing the flexible and accurate control of the gait of the robot, and simultaneously utilizing the feedback information of the wear data to assist the error control and the abnormal detection, the gait control effect of the robot is greatly improved.
Claims (9)
1. A training gait control method based on deep learning is characterized in that: the method comprises the following steps:
s1, constructing a gait model of the target robot on the simulation platform to realize the all-around walking of the robot;
s2, collecting motion requirement data, motion target data, various state data of the robot, wear data and external environment data, storing the data into a training database, and providing a training sample, wherein the wear data is wear data of a training coating, and the training coating is coated on each joint of the robot;
s3, constructing a deep learning framework based on the convolutional neural network, and training the convolutional neural network by using training samples in a training database to obtain a deep learning model;
s4, acquiring real-time motion requirement data, motion target data, various state data of the robot, wear data and external environment data, inputting the data into the deep learning model, and acquiring an output result of the deep learning model;
and S5, generating a control signal in real time according to the output result of the deep learning model by using the constructed gait model to control the gait of the robot in real time.
2. The deep learning-based training gait control method according to claim 1, characterized in that: the gait model of the target robot adopts a simplified 6-degree-of-freedom connecting rod model, and the feet adopt a planar sole structure.
3. The deep learning-based training gait control method according to claim 1, characterized in that: the state data of the robot comprises the self form data, the motion state data and the stress data of each part of the robot.
4. The deep learning-based training gait control method according to claim 1, characterized in that: be connected with a plurality of evenly distributed's perception hemisphere (1) on the training coating surface, perception hemisphere (1) internal surface is connected with inlays holds pipe (2) in the training coating, still inlay disintegration volume post (4) that are connected with a plurality of evenly distributed in the training coating, and disintegration volume post (4) are located and hold between pipe (2), hold and be connected with many evenly distributed's water guide fiber (3) between pipe (2) and disintegration volume post (4).
5. The deep learning-based training gait control method according to claim 4, characterized in that: perception hemisphere (1) is including magnetism outer layer (11), elasticity hemisphere (12) and extension rod (13) of inhaling, and magnetism is inhaled outer layer (11) and is covered and connect in elasticity hemisphere (12) surface, extension rod (13) are connected with magnetism outer layer (11) inner, and extension rod (13) run through elasticity hemisphere (12) and extend to hold in pipe (2).
6. The deep learning-based training gait control method according to claim 5, characterized in that: the containing pipe (2) is internally embedded and connected with a plurality of sponge water absorption balls (5) corresponding to the water guide fibers (3), and the sponge water absorption balls (5) penetrate through the containing pipe (2) and extend into the sponge water absorption balls (5).
7. The deep learning-based training gait control method according to claim 4, characterized in that: the disintegration column (4) comprises a plurality of disintegration units corresponding to the water guide fibers (3), each disintegration unit comprises a water-resisting membrane (41), a plurality of magnetic lubrication microspheres (42) and a disintegration block (43), the disintegration block (43) is filled between a pair of water-resisting membranes (41), and the magnetic lubrication microspheres (42) are embedded in the disintegration block (43).
8. The deep learning-based training gait control method according to claim 7, characterized in that: magnetic lubrication microballon (42) is including magnetism shell (421), oil absorption cotton layer (422) and meet thermal expansion core ball (423), meet thermal expansion core ball (423) and locate inside magnetism shell (421), and oil absorption cotton layer (422) parcel is met between thermal expansion core ball (423) and magnetism shell (421), a plurality of evenly distributed's hole (424) have been seted up on magnetism shell (421).
9. The deep learning-based training gait control method according to claim 8, characterized in that: a plurality of oil guiding substances (425) corresponding to the holes (424) are connected to the outer surface of the oil absorption cotton layer (422), and the oil guiding substances (425) extend to the openings of the holes (424).
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202011108791.1A CN112306060B (en) | 2020-10-16 | 2020-10-16 | Training gait control method based on deep learning |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202011108791.1A CN112306060B (en) | 2020-10-16 | 2020-10-16 | Training gait control method based on deep learning |
Publications (2)
Publication Number | Publication Date |
---|---|
CN112306060A CN112306060A (en) | 2021-02-02 |
CN112306060B true CN112306060B (en) | 2021-06-25 |
Family
ID=74327876
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202011108791.1A Expired - Fee Related CN112306060B (en) | 2020-10-16 | 2020-10-16 | Training gait control method based on deep learning |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN112306060B (en) |
Citations (11)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
EP1136193A2 (en) * | 2000-03-21 | 2001-09-26 | Sony Corporation | Humanoid robot communicating with body language |
CN101320251A (en) * | 2008-07-15 | 2008-12-10 | 华南理工大学 | Robot ambulation control method based on confirmation learning theory |
CN102591344A (en) * | 2012-03-05 | 2012-07-18 | 中国人民解放军国防科学技术大学 | Time and position control method of four-footed bionic robot |
CN103978484A (en) * | 2014-03-04 | 2014-08-13 | 上海大学 | Efficient and high-precision leg type gait planning method based on planet wheel mechanism legged robot |
CN104965517A (en) * | 2015-07-07 | 2015-10-07 | 张耀伦 | Robot cartesian space trajectory planning method |
CN105117545A (en) * | 2015-08-20 | 2015-12-02 | 河北工业大学 | Method for walking motion simulation experiments of lower extremity prostheses |
CN105760835A (en) * | 2016-02-17 | 2016-07-13 | 天津中科智能识别产业技术研究院有限公司 | Gait segmentation and gait recognition integrated method based on deep learning |
CN108009680A (en) * | 2017-11-30 | 2018-05-08 | 航天科工智能机器人有限责任公司 | Humanoid robot gait's planing method based on multi-objective particle swarm algorithm |
CN109288650A (en) * | 2018-07-31 | 2019-02-01 | 电子科技大学 | The independent used movable lower limb training of wearer and auxiliary intelligent apparatus |
CN110785268A (en) * | 2017-06-28 | 2020-02-11 | 谷歌有限责任公司 | Machine learning method and device for semantic robot grabbing |
CN111568705A (en) * | 2020-05-28 | 2020-08-25 | 连云港市第二人民医院(连云港市临床肿瘤研究所) | Motion detection system and detection method for lower limb robot of disabled person |
Family Cites Families (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JP5456588B2 (en) * | 2010-06-07 | 2014-04-02 | 本田技研工業株式会社 | Control device for legged mobile robot |
US8914151B2 (en) * | 2011-07-05 | 2014-12-16 | The State Of Oregon Acting By And Through The State Board Of Higher Education On Behalf Of Oregon State University | Apparatus and method for legged locomotion integrating passive dynamics with active force control |
KR101508973B1 (en) * | 2013-05-14 | 2015-04-07 | 한국과학기술연구원 | Gait rehabilitation having passive mechanism for shifting center of fravity |
-
2020
- 2020-10-16 CN CN202011108791.1A patent/CN112306060B/en not_active Expired - Fee Related
Patent Citations (11)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
EP1136193A2 (en) * | 2000-03-21 | 2001-09-26 | Sony Corporation | Humanoid robot communicating with body language |
CN101320251A (en) * | 2008-07-15 | 2008-12-10 | 华南理工大学 | Robot ambulation control method based on confirmation learning theory |
CN102591344A (en) * | 2012-03-05 | 2012-07-18 | 中国人民解放军国防科学技术大学 | Time and position control method of four-footed bionic robot |
CN103978484A (en) * | 2014-03-04 | 2014-08-13 | 上海大学 | Efficient and high-precision leg type gait planning method based on planet wheel mechanism legged robot |
CN104965517A (en) * | 2015-07-07 | 2015-10-07 | 张耀伦 | Robot cartesian space trajectory planning method |
CN105117545A (en) * | 2015-08-20 | 2015-12-02 | 河北工业大学 | Method for walking motion simulation experiments of lower extremity prostheses |
CN105760835A (en) * | 2016-02-17 | 2016-07-13 | 天津中科智能识别产业技术研究院有限公司 | Gait segmentation and gait recognition integrated method based on deep learning |
CN110785268A (en) * | 2017-06-28 | 2020-02-11 | 谷歌有限责任公司 | Machine learning method and device for semantic robot grabbing |
CN108009680A (en) * | 2017-11-30 | 2018-05-08 | 航天科工智能机器人有限责任公司 | Humanoid robot gait's planing method based on multi-objective particle swarm algorithm |
CN109288650A (en) * | 2018-07-31 | 2019-02-01 | 电子科技大学 | The independent used movable lower limb training of wearer and auxiliary intelligent apparatus |
CN111568705A (en) * | 2020-05-28 | 2020-08-25 | 连云港市第二人民医院(连云港市临床肿瘤研究所) | Motion detection system and detection method for lower limb robot of disabled person |
Non-Patent Citations (1)
Title |
---|
A neural network solution for bipedal gait synthesis;D.M.A. Lee 等;《IEEE Xplore》;20020806;第763-769页 * |
Also Published As
Publication number | Publication date |
---|---|
CN112306060A (en) | 2021-02-02 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
US9445711B2 (en) | System and method to magnetically actuate a capsule endoscopic robot for diagnosis and treatment | |
ES2875333T3 (en) | Emotional / behavioral / psychological state estimation system | |
TWI648031B (en) | Cardiac monitoring system | |
Kunze et al. | Towards recognizing tai chi-an initial experiment using wearable sensors | |
DE69939446D1 (en) | Motion responsive device for performing tests on motile animals | |
US9980672B2 (en) | Single-chambered sweat rate monitoring sensor | |
CN112306060B (en) | Training gait control method based on deep learning | |
Huang et al. | Directional sweat transport and breathable sandwiched electrodes for electrocardiogram monitoring system | |
Cui et al. | The study of a remote-controlled gastrointestinal drug delivery and sampling system | |
Akhavanhezaveh et al. | Diagnosing gait disorders based on angular variations of knee and ankle joints utilizing a developed wearable motion sensor | |
Yao et al. | An RNN-LSTM enhanced compact and affordable micro force sensing system for interventional continuum robots with interchangeable end-effector instruments | |
Castañeda et al. | Knee joint angle monitoring system based on inertial measurement units for human gait analysis | |
Bao et al. | A wearable multimode system with soft sensors for lower limb activity evaluation and rehabilitation | |
De Pasquale | Biomechanical energy harvesting: design, testing, and future trends in healthcare and human-machines interfacing | |
CN107997766A (en) | Gait tester | |
CN201135440Y (en) | Minitype measurement device for recording the human walking information | |
US11965876B1 (en) | Biomimetic interface device and methods of using the same | |
CN105455794A (en) | Wearing article | |
CN113081275B (en) | Force feedback device and application thereof | |
CN209421917U (en) | A kind of mask for positron radionuclide count detection instrument | |
Zou et al. | Flexible Multimodal Sensor Based on Double‐network Hydrogel for Human and Robotic Applications | |
CN110169763A (en) | A kind of digital medical health parameters monitoring device | |
CN107874743A (en) | One kind record conscious mouse respiratory movement and habituation are with removing habituation device | |
Owida et al. | Emerging Development of Auto‐Charging Sensors for Respiration Monitoring | |
Peuscher et al. | The" Glass Lung"-A Lifelike Electromechanical Lung Simulator |
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 | ||
CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20210625 |