CN109571528A - Activation lacking mechanical finger and finger tips trajectory predictions method - Google Patents
Activation lacking mechanical finger and finger tips trajectory predictions method Download PDFInfo
- Publication number
- CN109571528A CN109571528A CN201811630106.4A CN201811630106A CN109571528A CN 109571528 A CN109571528 A CN 109571528A CN 201811630106 A CN201811630106 A CN 201811630106A CN 109571528 A CN109571528 A CN 109571528A
- Authority
- CN
- China
- Prior art keywords
- finger
- joint
- finger joint
- telescopic rod
- lacking mechanical
- 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
- 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
- B25J9/10—Programme-controlled manipulators characterised by positioning means for manipulator elements
- B25J9/104—Programme-controlled manipulators characterised by positioning means for manipulator elements with cables, chains or ribbons
-
- 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
- B25J9/1664—Programme controls characterised by programming, planning systems for manipulators characterised by motion, path, trajectory planning
Abstract
The present invention provides a kind of activation lacking mechanical finger and finger tips trajectory predictions method.It realizes the rotation of key rope and rotates the effect combined with drive lacking, manpower can be imitated to be grabbed, has adaptive ability, entire finger structure is simple, it is low to manufacture processing cost, transmits torque and power using key rope, capacity usage ratio is high, it just can be good at reaching the envelope crawl to object by simple control method, the trajectory predictions method computational efficiency based on activation lacking mechanical finger of the present invention is higher, prediction is more accurate, reduces rate of false alarm, improves sensitivity.
Description
Technical field
The invention belongs to manipulator technical field more particularly to a kind of activation lacking mechanical finger and finger tips trajectory predictions
Method.
Background technique
Drive lacking refers to that independent control variable number is less than freedom degree number, is saving energy, is reducing cost, mitigation weight
Amount, enhancing system flexibility etc. all drive superior more completely.Under-actuated systems structure is simple, is convenient for whole power
Credit analysis and test.Simultaneously because the nonlinearity of under-actuated systems, Parameter Perturbation, multi objective control require and control amount by
The reasons such as limit, under-actuated systems are again complicated enough, convenient for studying and verifying the validity of various algorithms.
The uncertainty for needing to consider robot when robot lacks the necessary information for executing required by task, can by property
The uncertain factor of robot system is divided into three classes: Parameter uncertainties sexual factor, the i.e. uncertainty of principle model parameter,
Such as D-H parameter, sensor function；The system performance that nonparametric uncertain factor, i.e. principle model cannot describe or ignore,
As high-order shakes mode, mechanical system return difference etc.；Stochastic uncertainty factor, such as the duplicate measurements error of measuring system, driving
Think highly of multiple position error etc..
In addition to this, current activation lacking mechanical finger need to install a large amount of sensor, to obtain activation lacking mechanical finger
Seized condition, such as joint rotation angle signal is needed to carry out trajectory predictions, but joint rotation angle jitter and poor repeatability.
And the trajectory predictions based on current activation lacking mechanical finger have focused largely on and analyze mobile object spatial position, thus in advance
Mobile object the next position information is surveyed, is that speed is slow and not accurate enough.
Summary of the invention
In order to solve the above technical problems, the present invention provides a kind of activation lacking mechanical finger and finger tips trajectory predictions side
Method.
The present invention adopts the following technical scheme:
In some alternative embodiments, a kind of activation lacking mechanical finger is provided, comprising: key rope, sequentially axis connection is close
Finger joint, middle finger joint, remote finger joint, and the telescopic rod group of variable bound is carried out to each finger joint；Key rope one end and finger
Finger tip connection, the other end successively bypass the driving axis connection after three joint shafts of finger with motor；The nearly finger joint, middle finger joint
And pressure sensor is arranged in the working face of remote finger joint.
In some alternative embodiments, the telescopic rod group includes: that the first telescopic rod, the second telescopic rod and third are flexible
Bar；First telescopic rod one end and underactuated manipulator palm are hinged, and the other end and the working face of the nearly finger joint are hinged；Institute
It states second telescopic rod one end and the working face of the nearly finger joint is hinged, the working face of the other end and the middle finger joint is hinged；It is described
Third telescopic rod one end and the working face of the middle finger joint are hinged, and the other end and the working face of the remote finger joint are hinged.
In some alternative embodiments, each telescopic rod in the telescopic rod group is by pipe sleeve, action pipe and spring
Composition；Described action pipe one end is protruded into the cavity of the pipe sleeve and is connected via the bottom of chamber of the spring and the cavity of the pipe sleeve
It connects.
In some alternative embodiments, guide wheel is set on the nearly finger joint and the middle finger joint, in the remote finger joint
Fixing terminal is set, winding wheel is set in the drive shaft of the motor；One end of the key rope is connect with the fixing terminal, separately
One end is wrapped on the winding wheel after crossing through three joint shafts of the guide wheel and finger.
In some alternative embodiments, the activation lacking mechanical finger, further includes: be arranged in underactuated manipulator hand
The connector sleeve of palm, the motor are arranged on the connector sleeve, and directive wheel, one end of the key rope are arranged on the connector sleeve
It is connect with the fixing terminal, the other end crosses through after three joint shafts of the guide wheel and finger again via the guiding
Wheel is wrapped on the winding wheel after being oriented to twice.
In some alternative embodiments, the present invention provides a kind of finger tips trajectory predictions method, comprising:
Establish the end orbit prediction model returned based on Gaussian process；
The pressure sensor pressure data collected of activation lacking mechanical finger and the current value of motor are obtained, it will be described
The input value of pressure data and current value as the end orbit prediction model solves optimal solution to obtain underactuated manipulator
The angle value of each finger joint referred to；
Wherein, the activation lacking mechanical finger includes: key rope, sequentially the nearly finger joint of axis connection, middle finger joint, remote finger joint, with
And the telescopic rod group of variable bound is carried out to each finger joint, the finger tip of key rope one end and finger connects, the other end successively around
Cross the driving axis connection after three joint shafts of finger with the motor, the working face of the nearly finger joint, middle finger joint and remote finger joint
Pressure sensor is set.
In some alternative embodiments, the finger tips trajectory predictions method, further includes: define activation lacking mechanical
The judgment basis of finger grip stability；Wherein, using the contact condition of contact phase finish time as underactuated manipulator
Refer to the judgment basis of grasp stability, the judgment basis be when emulation starts each finger joint according to given trace persistent movement,
When finger joint and object contact, the constraint condition is added later and continues for finger joint stop motion immediately relevant to contact point
Kinematics Simulation, until each finger joint is contacted with object or track is completed.
In some alternative embodiments, described establish based on the end orbit prediction model that Gaussian process returns includes:
Choose gaussian kernel function；
A linear regression model (LRM) with Gaussian noise based on bayesian theory is defined, form performance is as follows:
F (x)=xTw；
Y=f (x)+ε；
Wherein, x is input vector, and w is the weight vectors of linear model, and f is functional value, and y is expressed as observed object value, ε
Indicate independent Gaussian noise, ε meets the standardized normal distribution that mean value is 0；
The value for solving the solution weight vectors w of linear model obtains the observation probability density of known parameters, form performance
It is as follows:
Wherein, | z | indicate the European length of vector z, X indicates observation；
Regression forecasting is carried out using Posterior probability distribution, giving input vector is x ' ∈ Rn, then its corresponding functional value
For y ', then the probability distribution of y ' are as follows:
Wherein, A=σn -2XXT+∑p -1；
A square of index covariance function is given, as follows:
Consider Gaussian noise, covariance function indicated are as follows:
cov(f(xp), f (xq))=K (xp, xq)+σn2I；
Wherein, I indicates the unit matrix of NxN, K (xp, xq) it is expressed as nuclear matrix；
Assuming that the functional value f of a data-oriented, then its Posterior probability distribution are as follows:
Wherein, (X, y) is sample data, then obtains an edge likelihood based on function f, as follows:
P (y | x, θ, K)=∫ p (y | X, f, θ, K) p (f | X, θ, K) df；
The extreme point of hyper parameter θ maximal margin likelihood is solved, as follows:
θ '=argmaxp (θ | X, f, K)；
L (θ)=p (θ | X, f, K) is enabled, then obtains following formula:
The extreme point of above formula can be acquired by the local derviation to each hyper parameter.
In some alternative embodiments, the foundation is also wrapped based on the end orbit prediction model that Gaussian process returns
It includes:
A Gaussian process is defined to describe regression function distribution, carries out Bayesian inference, the Gauss in function space
The property of process determines by mean function and covariance function, such as following formula:
Wherein, x, x ' ∈ RdFor any stochastic variable, k (x, x ') is kernel function, thus Gaussian process be defined as f (x)~
GP (m (x), k (x, x ')).
The utility model has the advantages that can adapt to nonparametric uncertain factor and random uncertain factor brought by of the invention；It is real
The rotation of key rope is showed and has rotated the effect combined with drive lacking, manpower can be imitated and grabbed, have adaptive ability, entire hand
Refer to that structure is simple, manufacture processing cost is low, transmits torque and power using key rope, capacity usage ratio is high, by simply controlling
Method just can be good at reaching the envelope crawl to object, based on the trajectory predictions method of activation lacking mechanical finger of the present invention
It is more efficient, prediction is more accurate, reduces rate of false alarm, improves sensitivity.
For the above and related purposes, one or more embodiments include being particularly described below and in claim
In the feature that particularly points out.Certain illustrative aspects are described in detail in the following description and the annexed drawings, and its instruction is only
Some modes in the utilizable various modes of the principle of each embodiment.Other benefits and novel features will be under
The detailed description in face is considered in conjunction with the accompanying and becomes obvious, the disclosed embodiments be all such aspects to be included and they
Be equal.
Detailed description of the invention
Fig. 1 is the structural schematic diagram of activation lacking mechanical finger of the present invention；
Fig. 2 is the main view of activation lacking mechanical finger of the present invention；
Fig. 3 is that key rope of the present invention moves towards schematic diagram；
Fig. 4 is the position view of motor of the present invention, connector sleeve and directive wheel；
Fig. 5 is the variation schematic diagram of activation lacking mechanical finger of the present invention in the process of grasping；
Fig. 6 is the connection schematic diagram of pipe sleeve of the present invention, action pipe and spring；
Fig. 7 is the schematic diagram for the drive lacking hand control system model that the present invention establishes.
Specific embodiment
The following description and drawings fully show specific embodiments of the present invention, to enable those skilled in the art to
Practice them.Other embodiments may include structure, logic, it is electrical, process and other change.Embodiment
Only represent possible variation.Unless explicitly requested, otherwise individual components and functionality is optional, and the sequence operated can be with
Variation.The part of some embodiments and feature can be included in or replace part and the feature of other embodiments.This hair
The range of bright embodiment includes equivalent obtained by the entire scope of claims and all of claims
Object.
As shown in Figures 1 to 3, in some illustrative embodiments, a kind of activation lacking mechanical finger is provided, comprising: key rope
1, nearly finger joint 2, middle finger joint 3, remote finger joint 4, telescopic rod group, motor 6, pressure sensor 7, connector sleeve 8, winding wheel 11.It realizes
The rotation of key rope 1 rotates the effect combined with drive lacking, can imitate manpower and be grabbed, have adaptation function.
Nearly finger joint 2, middle finger joint 3, remote finger joint 4 sequentially axis connection, nearly finger joint 2 and underactuated manipulator palm axis connection,
In, activation lacking mechanical finger further include: three joint shafts, i.e., close joint shaft 2a, middle joint shaft 3a, remote nodal axisn 4a.
Telescopic rod group carries out variable bound to each finger joint, comprising: the first telescopic rod 51, the second telescopic rod 52 and third are stretched
Contracting bar 53.First telescopic rod, 51 one end and underactuated manipulator palm are hinged, and the other end and the working face of nearly finger joint 2 are hinged；The
The quantity of two telescopic rods 52 is two, and one end and the working face of nearly finger joint 2 are hinged, and the working face of the other end and middle finger joint 3 is cut with scissors
It connects；53 one end of third telescopic rod and the working face of middle finger joint 3 are hinged, and the other end and the working face of remote finger joint 4 are hinged.Wherein, work
Refer to the one side of activation lacking mechanical finger crawl object as face.
Activation lacking mechanical finger is provided with key rope 1, to control mechanical finger crawl.The finger tip of key rope 1 one end and finger
Connection, the other end successively bypass the driving axis connection after three joint shafts of finger with motor 6, and the finger tip of finger is remote finger joint
4 end, three finger joints share a motor 6 and drive.
Nearly finger joint 2 with guide wheel 9 is set on middle finger joint 3, fixing terminal 10 is set in remote finger joint 4, in the drive shaft of motor 6
Winding wheel 11 is set.One end of key rope 1 is connect with fixing terminal 10, and the other end crosses through three passes of guide wheel 9 and finger
It is wrapped in after nodal axisn on winding wheel 11.When motor 6 rotates 11 drive key rope 1 of winding wheel, the driving force of motor 6 can be via key rope
1 is transmitted to each finger joint, and three finger joints of finger can rotate accordingly according to the rigidity ratio of each telescopic rod in telescopic rod group
Angle, if three finger joints are not all contacted with external object at this time, the angle value of angle value and three joint shafts that motor 6 rotates
Unique corresponding, i.e., end orbit is certain.If any finger joint is contacted with external object, the track of finger tip is unpredictable, this is deficient
Drive the characteristic of finger inherently.When motor 6 rotate release key rope 1 when, each finger joint can telescopic rod group limitation next time
Multiple initial position.In above process, in conjunction with joint locking device, whether according to the enabling of actually required control finger-joint,
It can achieve the effect that finger tip track is controllable.
As shown in figure 4, winding wheel 11 is used to fixing and storing key rope 1, and by cooperating with winding wheel cap 12,11 quilt of winding wheel
It is fixed on the output shaft of motor 6, and the power output of motor 6 is passed into key rope 1, it is dynamic to complete crawl for driving manipulator
Make.The effect of connector sleeve 8 is that entire motor drive module is fixed on underactuated manipulator palm, guarantees motor drive module
Output driving power that can be stable.Motor 6 be arranged on connector sleeve 8, on connector sleeve 8 be arranged directive wheel 13, since finger tip through
It is connect by each finger joint and one end of key rope 1 for being passed through from connector sleeve 8 with fixing terminal 10, the other end crosses through guide wheel 9
And it is wrapped on winding wheel 11 after three joint shafts of finger, then after being oriented to twice via institute's directive wheel.
Directive wheel 13 is fixed on running wheel bracket 14 by pin shaft and circlip for shaft, directive wheel 13 and running wheel bracket 14
Running wheel rack module is collectively constituted, running wheel rack module major function is correct guidance linking motor driven part and finger section
The key rope 1 divided, normally can transmit power and be unlikely to knot.
As shown in fig. 6, each telescopic rod in telescopic rod group, i.e. the first telescopic rod 51, the second telescopic rod 52 and third are stretched
Contracting bar 53 is made of pipe sleeve 15, action pipe 16 and spring 17.16 one end of action pipe is protruded into the cavity of pipe sleeve 15 and via bullet
Spring 17 is connect with the bottom of chamber of the cavity of pipe sleeve, to realize the restricted movement of action pipe 16.In the pass of activation lacking mechanical finger
Mechanical hard limit is equipped at section, to guarantee that mechanical finger will not the excess of stroke.
Pressure sensor 7 is arranged in the working face of nearly finger joint 2, middle finger joint 3 and remote finger joint 4.
The process of single finger envelope crawl object can be described as:
Firstly, finger tips, that is, finger tip has determining track before activation lacking mechanical finger does not detect object；
Then, when wherein some finger joint detects object, the angle in each joint is not just distributed according still further to joint stiffness,
Since the stress of finger interior at this time is complicated, the movement of each finger joint no longer has certain rule；
Finally, motor 6 continues to drive activation lacking mechanical finger movement, the finger joint for having contacted body surface can be in driving force
It is slided under effect along body surface, motor 6 stops driving when three finger joints all detect object, at this time activation lacking mechanical
Finger forms complete envelope to object, completes envelope crawl.
Activation lacking mechanical finger just can be good at reaching the envelope crawl to object by simple control method, entirely
The crawl task process of manipulator can be described as:
Firstly, the original state that activation lacking mechanical finger keeps it to grab；
Secondly, be equipped with the mechanical arm of mechanical finger according to certain mode, such as the crawl target that obtains of machine vision
Pose, stop motion after underactuated manipulator to be moved to the position contacted with object external outline nearby；
Finally, the motor 6 of activation lacking mechanical finger is started to work, activation lacking mechanical finger starts close to object, until owing
Driving manipulator refers to complete envelope object.
As shown in figure 5, activation lacking mechanical finger prediction planning demonstration: being original state first；It is attached to then move to target
Part；Then, nearly 2 contact target object of finger joint, nearly finger joint joint is locked, obtains the driving current letter of contact force signal and motor
Number；Then, remote 4 contact target object of finger joint, remote 4 joint of finger joint are slided in target without lock, obtain contact force signal and motor
Driving current signal；Finally, 3 contact target object of middle finger joint, 3 joint of middle finger joint is locked, and remote 4 joint of finger joint is locked, is connect
Touch force signal and driving current signal.
Finger tips trajectory predictions are directed to the output motion profile that given contact point solves driving unit.In order to grasp in advance
The end orbit of activation lacking mechanical finger prevents the security risk of underactuated manipulator to take effective actions, and proposes one
The end orbit prediction model that kind is returned based on Gaussian process carries out analogue simulation to existing end orbit, to predict
End orbit route, effective trajectory planning play an important role in tracking and early warning contingency management.
The present invention provides a kind of finger tips trajectory predictions method, comprising:
Firstly, establishing the end orbit prediction model returned based on Gaussian process；
Then, the judgment basis of activation lacking mechanical finger grasp stability is defined；
Finally, the pressure sensor pressure data collected of activation lacking mechanical finger and the current value of motor are obtained,
Using pressure data and current value as the input value of end orbit prediction model, optimal solution is solved to obtain activation lacking mechanical finger
Each finger joint angle value, by the angle value of each finger joint can obtain crawl track.
In order to guarantee to grab the computational efficiency of trajectory planning, driven using the contact condition of contact phase finish time as deficient
The judgment basis of motivation tool finger grip stability.In order to realize the purpose, crawl emulation only needs to consider contact phase
Robot movement state, detailed process is as follows: each finger joint is according to given trace persistent movement when starting for emulation, when direct and object
Body finger joint stop motion immediately relevant to contact point when contacting, the constraint condition is added later and continues Kinematics Simulation,
Until each finger joint contacted with object or track complete.
In view of motor driving current signals and joint rotation angle jitter and poor repeatability, in order to reduce rate of false alarm, mention
It is highly sensitive, it is necessary to the reference model and signal noise model of more accurate contact detection.For this purpose, the present invention uses Gauss mistake
Journey return come back approach contact detection reference model and signal noise model prediction go out end orbit route.
It establishes based on the end orbit prediction model that Gaussian process returns and includes:
Gaussian process recurrence is a kind of new machine learning algorithm based on Bayesian network, not only has interpretation strong
Bayesian Network Inference ability, while adaptive place the problems such as be provided with the small sample of support vector machines, non-linear, higher-dimension
Reason ability is the research hotspot in machine learning field, can preferably solve the problems, such as trajectory predictions.
Gaussian Profile is a kind of continuity random distribution model, when stochastic variable is single argument x, under Gaussian Profile meets
Formula:
Wherein, μ is mean value, σ2It is variance.
If stochastic variable x is multidimensional, indicated with vector D, Joint Gaussian distribution such as following formula:
Wherein, μ is the mean vector of multi-C vector D, and ∑ is the covariance matrix of DxD.
From function space angle, a Gaussian process is defined to describe regression function distribution, directly in function space
Bayesian inference is carried out, Gaussian process is the set that arbitrary finite stochastic variable all has Joint Gaussian distribution, and property is complete
It is determined entirely by mean function and covariance function, such as following formula:
Wherein, x, x ' ∈ RaFor any stochastic variable, k (x, x ') is kernel function, therefore Gaussian process may be defined as: f (x)
~GP (m (x), k (x, x ')) usually pre-processes data in order to succinct on symbol, its mean function is made to be equal to 0.
Kernel function is the core content that Gaussian process returns, and effect is that high dimension vector is calculated to calculate to low-dimensional vector to turn
It changes, reduces computational complexity.
The form for the gaussian kernel function that the present invention chooses is as follows:
Wherein, k (| | x-xc | |) is defined as the dull letter of Euclidean distance between any point x to a certain center xc in space
Number, xc are kernel function center, and σ is the width parameter of function, control the radial effect range of function.
From weight space angle, it is contemplated that a linear regression mould with Gaussian noise based on bayesian theory
Type, form performance are as follows:
F (x)=xTw；
Y=f (x)+ε；
Wherein, x is input vector, and w is the weight vectors of linear model, and f is functional value, and y is expressed as observed object value, ε
Indicate independent Gaussian noise, ε meets the standardized normal distribution that mean value is 0, and form is as follows:
ε~N (0, σn 2)。
This linear regression problem is the value to solve the weight vectors w of linear model, available according to independence assumption
The observation probability density of known parameters, form performance are as follows:
Wherein, X indicates observation.
Regression forecasting is carried out using Posterior probability distribution, giving input vector is x ' ∈ Rn, then its corresponding functional value
For y ', then the probability distribution of y ' are as follows:
Above formula is the distribution function of y ', wherein A=σn -2XXT+∑p -1；X indicates observation.
GP modeling purpose be find be capable of it is good fit observation data covariance function and hyper parameter give for convenience
A fixed square of index covariance function, as follows:
In view of the presence of Gaussian noise, covariance function is indicated are as follows:
cov(f(xp), f (xq))=K (xp, xq)+σn 2I；
Wherein, I indicates the unit matrix of NxN, K (xp, xq) it is expressed as nuclear matrix.
For each covariance function K, there is some hyper parameter θ, it is now assumed that the functional value f of a data-oriented, then its
Posterior probability distribution are as follows:
Wherein, (X, y) is sample data, then obtains an edge likelihood based on function f, as follows:
P (y | x, θ, K)=∫ p (y | X, f, θ, K) p (f | X, θ, K) df；
It is realized by solving the extreme point of hyper parameter θ maximal margin likelihood:
θ '=argmaxp (θ | X, f, K)；
Selection asks the extreme point of its logarithm to maximize p (θ | X, f, K).
L (θ)=p (θ | X, f, K) is enabled, then obtains following formula:
The extreme point of above formula can be acquired by the local derviation to each hyper parameter.
The present invention establishes the end orbit prediction model and other homing methods, such as nerve net returned based on Gaussian process
The output predicted value of network, support vector machines etc. is different, Gaussian process returns the distribution situation of output predicted value, i.e. Gaussian Profile
Mean value and variance.Advantage are as follows: predicted value is probability, therefore can calculate experience confidence interval, then according to these information,
Some area-of-interest is fitted prediction again；It is multi-functional, different core can be specified for system, be mentioned by the complexity of system
For different IPs, but it also can specify specific core.
As shown in fig. 7, this prediction technique obtains drive lacking hand control system model by Gaussian process recurrence, adopt thus
Drive lacking hand control system is described with incremental model, as follows:
δs[k]=fa(s[k], u[k])+ga(s[k], u[k])；
In formula: [k] is sample point data number；s[k]System state amount is controlled for drive lacking hand；u[k]For drive lacking manual control
System control amount processed；δs[k]System state amount increment is controlled for drive lacking hand；fa(s[k], u[k]) it is principle model；ga(s[k],
u[k]) it is principle model error.
To obtain principle model error ga(s[k], u[k]), using the residual error of Gaussian process regression approach principle model, definition
The training set of the regression problem of [k] number sampled point are as follows:
Simultaneous end orbit prediction model and drive lacking hand control system model, obtain drive lacking hand control system model
Output, i.e. mean value and variance:
In formula,The mean value of system mode incremental forecasting is controlled for drive lacking hand；For drive lacking hand control system shape
The variance of state incremental forecasting；ug|kFor Gaussian process return prediction distribution in training set x=[s[k], u[k]] be under the conditions of it is equal
Value output；σg[k]For Gaussian process return prediction distribution in training set x=[s[k], u[k]] be under the conditions of variance output.
To make a certain performance indicator optimal while given prediction result, as shown in fig. 7, drive lacking hand control system model
The control amount for being sent to robot is obtained by solving optimal control problem, solid line indicates optimum control process, and dotted line indicates
Roll learning process.
In order to distinguish the inner sensor and outer sensor system state amount obtained of drive lacking hand control system, by system
State statement are as follows:
s[k]=[q[k]；p[k]]；
In formula, q[k]The system state amount obtained for inner sensor；p[k]The system state amount obtained for outer sensor.
It should also be appreciated by one skilled in the art that various illustrative logical boxs, mould in conjunction with the embodiments herein description
Electronic hardware, computer software or combinations thereof may be implemented into block, circuit and algorithm steps.In order to clearly demonstrate hardware and
Interchangeability between software surrounds its function to various illustrative components, frame, module, circuit and step above and carries out
It is generally described.Hardware is implemented as this function and is also implemented as software, depends on specific application and to entire
The design constraint that system is applied.Those skilled in the art can be directed to each specific application, be realized in a manner of flexible
Described function, still, this realization decision should not be construed as a departure from the scope of protection of this disclosure.
Claims (9)
1. activation lacking mechanical finger characterized by comprising key rope, the sequentially nearly finger joint of axis connection, middle finger joint, remote finger joint, with
And the telescopic rod group of variable bound is carried out to each finger joint；
The connection of the finger tip of key rope one end and finger, the other end successively bypass the driving after three joint shafts of finger with motor
Axis connection；
Pressure sensor is arranged in the working face of the nearly finger joint, middle finger joint and remote finger joint.
2. activation lacking mechanical finger according to claim 1, which is characterized in that the telescopic rod group includes: first flexible
Bar, the second telescopic rod and third telescopic rod；
First telescopic rod one end and underactuated manipulator palm are hinged, and the other end and the working face of the nearly finger joint are hinged；
Second telescopic rod one end and the working face of the nearly finger joint are hinged, and the working face of the other end and the middle finger joint is cut with scissors
It connects；
Third telescopic rod one end and the working face of the middle finger joint are hinged, and the working face of the other end and the remote finger joint is cut with scissors
It connects.
3. activation lacking mechanical finger according to claim 2, which is characterized in that each telescopic rod in the telescopic rod group
It is made of pipe sleeve, action pipe and spring；Described action pipe one end protrude into the cavity of the pipe sleeve and via the spring with
The bottom of chamber of the cavity of the pipe sleeve connects.
4. activation lacking mechanical finger according to claim 1 or 3, which is characterized in that the nearly finger joint and the middle finger joint
Upper setting guide wheel is arranged fixing terminal in the remote finger joint, winding wheel is arranged in the drive shaft of the motor；The key rope
One end is connect with the fixing terminal, the other end cross through be wrapped in after three joint shafts of the guide wheel and finger it is described
On winding wheel.
5. activation lacking mechanical finger according to claim 4, which is characterized in that further include: it is arranged in underactuated manipulator
Connector sleeve on palm, the motor are arranged on the connector sleeve, are arranged directive wheel on the connector sleeve, and the one of the key rope
End is connect with the fixing terminal, and the other end is led via described again after crossing through three joint shafts of the guide wheel and finger
It is wrapped on the winding wheel after being oriented to twice to wheel.
6. finger tips trajectory predictions method characterized by comprising
Establish the end orbit prediction model returned based on Gaussian process；
The pressure sensor pressure data collected of activation lacking mechanical finger and the current value of motor are obtained, by the pressure
The input value of data and current value as the end orbit prediction model solves optimal solution to obtain activation lacking mechanical finger
The angle value of each finger joint；
Wherein, the activation lacking mechanical finger includes: key rope, sequentially the nearly finger joint of axis connection, middle finger joint, remote finger joint and right
Each finger joint carries out the telescopic rod group of variable bound, and the finger tip of key rope one end and finger connects, and the other end is successively around receiving and distributing
Driving axis connection after three joint shafts referred to the motor, the working face setting of the nearly finger joint, middle finger joint and remote finger joint
Pressure sensor.
7. finger tips trajectory predictions method according to claim 6, which is characterized in that further include:
Define the judgment basis of activation lacking mechanical finger grasp stability；
Wherein, using the contact condition of contact phase finish time as activation lacking mechanical finger grasp stability judgement according to
According to, the judgment basis is that each finger joint is according to given trace persistent movement when emulation starts, when finger joint is contacted with object, with
The stop motion immediately of the relevant finger joint in contact point, the constraint condition is added later and continues Kinematics Simulation, until each finger
Section is contacted with object or track is completed.
8. finger tips trajectory predictions method according to claim 7, which is characterized in that described establish is based on Gaussian process
The end orbit prediction model of recurrence includes:
Choose gaussian kernel function；
A linear regression model (LRM) with Gaussian noise based on bayesian theory is defined, form performance is as follows:
F (x)=xTw；
Y=f (x)+ε；
Wherein, x is input vector, and w is the weight vectors of linear model, and f is functional value, and y is expressed as observed object value, and ε is indicated
Independent Gaussian noise, ε meet the standardized normal distribution that mean value is 0；
The value for solving the solution weight vectors w of linear model, obtains the observation probability density of known parameters, and form shows such as
Under:
Wherein, | z | indicate the European length of vector z, X indicates observation；
Regression forecasting is carried out using Posterior probability distribution, giving input vector is x ' ∈ Rn, then its corresponding functional value is y ',
The then probability distribution of y ' are as follows:
Wherein, A=σn -2XXT+∑p -1；
A square of index covariance function is given, as follows:
Consider Gaussian noise, covariance function indicated are as follows:
cov(f(xp), f (xq))=K (xp, xq)+σn 2I；
Wherein, I indicates the unit matrix of NxN, K (xp, xq) it is expressed as nuclear matrix；
Assuming that the functional value f of a data-oriented, then its Posterior probability distribution are as follows:
Wherein, (X, y) is sample data, then obtains an edge likelihood based on function f, as follows:
P (y | x, θ, K)=∫ p (y | X, f, θ, K) p (f | X, θ, K) df；
The extreme point of hyper parameter θ maximal margin likelihood is solved, as follows:
θ '=argmaxp (θ | X, f, K)；
L (θ)=p (θ | X, f, K) is enabled, then obtains following formula:
The extreme point of above formula can be acquired by the local derviation to each hyper parameter.
9. finger tips trajectory predictions method according to claim 8, which is characterized in that described establish is based on Gaussian process
The end orbit prediction model of recurrence further include:
A Gaussian process is defined to describe regression function distribution, carries out Bayesian inference, the Gaussian process in function space
Property determined by mean function and covariance function, such as following formula:
Wherein, x, x ' ∈ RdFor any stochastic variable, k (x, x ') is kernel function, therefore Gaussian process is defined as f (x)~GP (m
(x), k (x, x ')).
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811630106.4A CN109571528A (en) | 2018-12-29 | 2018-12-29 | Activation lacking mechanical finger and finger tips trajectory predictions method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811630106.4A CN109571528A (en) | 2018-12-29 | 2018-12-29 | Activation lacking mechanical finger and finger tips trajectory predictions method |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN109571528A true CN109571528A (en) | 2019-04-05 |
Family
ID=65932325
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201811630106.4A Pending CN109571528A (en) | 2018-12-29 | 2018-12-29 | Activation lacking mechanical finger and finger tips trajectory predictions method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109571528A (en) |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110170989A (en) * | 2019-05-05 | 2019-08-27 | 北京空间飞行器总体设计部 | A kind of rope drive drive lacking grasping mechanism Parameters design |
| CN110501421A (en) * | 2019-07-24 | 2019-11-26 | 武汉大学 | A kind of track profiling method of detection based on mechanical arm |
| CN111645096A (en) * | 2020-06-13 | 2020-09-11 | 南通大学 | Slider promotes type robot finger structure |
| CN117140504A (en) * | 2023-08-01 | 2023-12-01 | 四川大学 | N-link mechanical arm control method based on incremental model predictive control |
| CN117140504B (en) * | 2023-08-01 | 2024-05-10 | 四川大学 | N-link mechanical arm control method based on incremental model predictive control |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106903682A (en) * | 2017-04-05 | 2017-06-30 | 哈尔滨工业大学 | A kind of controllable modularization under-actuated finger of end orbit |
| CN106945069A (en) * | 2017-04-05 | 2017-07-14 | 哈尔滨工业大学 | Three controllable three finger joint underactuated manipulators of finger of a kind of finger tips track |
| CN206445813U (en) * | 2016-03-18 | 2017-08-29 | 杜宇 | A kind of drive lacking humanoid dexterous arm device |
| US9914214B1 (en) * | 2016-02-22 | 2018-03-13 | X Development Llc | Preshaping for underactuated fingers |
| CN207606868U (en) * | 2017-12-25 | 2018-07-13 | 南京工程学院 | A kind of simple activation lacking mechanical finger |
-
2018
- 2018-12-29 CN CN201811630106.4A patent/CN109571528A/en active Pending
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US9914214B1 (en) * | 2016-02-22 | 2018-03-13 | X Development Llc | Preshaping for underactuated fingers |
| CN206445813U (en) * | 2016-03-18 | 2017-08-29 | 杜宇 | A kind of drive lacking humanoid dexterous arm device |
| CN106903682A (en) * | 2017-04-05 | 2017-06-30 | 哈尔滨工业大学 | A kind of controllable modularization under-actuated finger of end orbit |
| CN106945069A (en) * | 2017-04-05 | 2017-07-14 | 哈尔滨工业大学 | Three controllable three finger joint underactuated manipulators of finger of a kind of finger tips track |
| CN207606868U (en) * | 2017-12-25 | 2018-07-13 | 南京工程学院 | A kind of simple activation lacking mechanical finger |
Non-Patent Citations (2)
| Title |
|---|
| 曲轶松: "基于高斯过程的数据处理的研究", 《中国优秀硕士学位论文全文数据库》 * |
| 李励耘: "基于高斯过程的抓取规划方法研究", 《中国优秀硕士学位论文全文数据库》 * |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110170989A (en) * | 2019-05-05 | 2019-08-27 | 北京空间飞行器总体设计部 | A kind of rope drive drive lacking grasping mechanism Parameters design |
| CN110501421A (en) * | 2019-07-24 | 2019-11-26 | 武汉大学 | A kind of track profiling method of detection based on mechanical arm |
| CN111645096A (en) * | 2020-06-13 | 2020-09-11 | 南通大学 | Slider promotes type robot finger structure |
| CN117140504A (en) * | 2023-08-01 | 2023-12-01 | 四川大学 | N-link mechanical arm control method based on incremental model predictive control |
| CN117140504B (en) * | 2023-08-01 | 2024-05-10 | 四川大学 | N-link mechanical arm control method based on incremental model predictive control |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109571528A (en) | Activation lacking mechanical finger and finger tips trajectory predictions method | |
| CN110750096B (en) | Mobile robot collision avoidance planning method based on deep reinforcement learning in static environment | |
| JP5759495B2 (en) | Procedure memory learning and robot control | |
| CN108828944B (en) | Encoder fault diagnosis system and method based on improved PSO and SVM | |
| CN105259786B (en) | The inertial parameter discrimination method and device of target to be identified | |
| CN110216649A (en) | The control method of robot manipulating task system and robot manipulating task system | |
| CN105818129A (en) | Humanoid hand control system based on data glove | |
| CN106329399B (en) | A kind of control method and controller of transmission line of electricity bolt fastening machine people | |
| CN110000787A (en) | A kind of control method of super redundant mechanical arm | |
| CN112276944A (en) | Man-machine cooperation system control method based on intention recognition | |
| Lin et al. | Reinforcement learning without ground-truth state | |
| CN114378811A (en) | Force and torque guided robotic assembly | |
| WO2020257263A1 (en) | Systems and methods for solving geosteering inverse problems in downhole environments using a deep neural network | |
| Gutzeit et al. | The besman learning platform for automated robot skill learning | |
| CN116494247A (en) | Mechanical arm path planning method and system based on depth deterministic strategy gradient | |
| Boltov et al. | Performance Evaluation of Real-Time System for Vision-Based Navigation of Small Autonomous Mobile Robots | |
| CN111590575B (en) | Robot control system and method | |
| Paudel | Learning for robot decision making under distribution shift: A survey | |
| CN111203883A (en) | Self-learning model prediction control method for robot electronic component assembly | |
| CN112809675A (en) | Method for automatically capturing space debris by using super-redundant mechanical arm based on reinforcement learning algorithm | |
| CN109976188A (en) | A kind of cricket control method and system based on Timed Automata | |
| Heyu et al. | Impedance control method with reinforcement learning for dual-arm robot installing slabstone | |
| Konidaris et al. | Sensorimotor abstraction selection for efficient, autonomous robot skill acquisition | |
| Horiuchi et al. | A Wireless Wearable Sensor for Pointing and Arm-Gesture Recognition | |
| CN108927806A (en) | A kind of industrial robot learning method applied to high-volume repeatability processing |
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: 20190405 |
|
| RJ01 | Rejection of invention patent application after publication |