CN110327187A - A kind of band priori torque non-model control method of ectoskeleton - Google Patents
A kind of band priori torque non-model control method of ectoskeleton Download PDFInfo
- Publication number
- CN110327187A CN110327187A CN201910618865.7A CN201910618865A CN110327187A CN 110327187 A CN110327187 A CN 110327187A CN 201910618865 A CN201910618865 A CN 201910618865A CN 110327187 A CN110327187 A CN 110327187A
- Authority
- CN
- China
- Prior art keywords
- ectoskeleton
- knee joint
- model
- joint
- torque
- 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.)
- Granted
Links
- 238000000034 method Methods 0.000 title claims abstract description 53
- 210000000629 knee joint Anatomy 0.000 claims abstract description 58
- 210000003141 lower extremity Anatomy 0.000 claims abstract description 28
- 239000013598 vector Substances 0.000 claims description 58
- 230000033001 locomotion Effects 0.000 claims description 31
- 230000005021 gait Effects 0.000 claims description 18
- 238000004458 analytical method Methods 0.000 claims description 15
- 230000000694 effects Effects 0.000 claims description 13
- 210000002414 leg Anatomy 0.000 claims description 11
- 238000004422 calculation algorithm Methods 0.000 claims description 10
- 238000013461 design Methods 0.000 claims description 9
- 239000003550 marker Substances 0.000 claims description 9
- 210000000544 articulatio talocruralis Anatomy 0.000 claims description 7
- 210000000689 upper leg Anatomy 0.000 claims description 6
- 238000013459 approach Methods 0.000 claims description 4
- 210000001624 hip Anatomy 0.000 claims description 4
- 101000911772 Homo sapiens Hsc70-interacting protein Proteins 0.000 claims description 3
- 238000004364 calculation method Methods 0.000 claims description 3
- 238000009795 derivation Methods 0.000 claims description 3
- 210000000474 heel Anatomy 0.000 claims description 3
- 210000001699 lower leg Anatomy 0.000 claims description 3
- 210000002303 tibia Anatomy 0.000 claims description 3
- 210000003371 toe Anatomy 0.000 claims description 3
- 238000012876 topography Methods 0.000 claims description 3
- 230000008859 change Effects 0.000 claims description 2
- 230000006870 function Effects 0.000 description 6
- 239000011159 matrix material Substances 0.000 description 6
- 230000008569 process Effects 0.000 description 6
- 238000004088 simulation Methods 0.000 description 6
- 238000005259 measurement Methods 0.000 description 5
- 241001269238 Data Species 0.000 description 4
- 238000010586 diagram Methods 0.000 description 4
- 238000005070 sampling Methods 0.000 description 4
- 230000001133 acceleration Effects 0.000 description 3
- 230000005484 gravity Effects 0.000 description 3
- 238000012549 training Methods 0.000 description 3
- 230000003044 adaptive effect Effects 0.000 description 2
- 238000013499 data model Methods 0.000 description 2
- 210000002683 foot Anatomy 0.000 description 2
- 210000001503 joint Anatomy 0.000 description 2
- 238000006467 substitution reaction Methods 0.000 description 2
- 230000009471 action Effects 0.000 description 1
- 210000000988 bone and bone Anatomy 0.000 description 1
- 238000011217 control strategy Methods 0.000 description 1
- 238000013501 data transformation Methods 0.000 description 1
- 230000007812 deficiency Effects 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 238000002474 experimental method Methods 0.000 description 1
- 210000004394 hip joint Anatomy 0.000 description 1
- 230000008676 import Effects 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 230000003993 interaction Effects 0.000 description 1
- 210000003127 knee Anatomy 0.000 description 1
- 238000013178 mathematical model Methods 0.000 description 1
- 230000010363 phase shift Effects 0.000 description 1
- 230000035935 pregnancy Effects 0.000 description 1
- URWAJWIAIPFPJE-YFMIWBNJSA-N sisomycin Chemical compound O1C[C@@](O)(C)[C@H](NC)[C@@H](O)[C@H]1O[C@@H]1[C@@H](O)[C@H](O[C@@H]2[C@@H](CC=C(CN)O2)N)[C@@H](N)C[C@H]1N URWAJWIAIPFPJE-YFMIWBNJSA-N 0.000 description 1
- 238000012360 testing method Methods 0.000 description 1
- 238000012795 verification Methods 0.000 description 1
- 239000011800 void material Substances 0.000 description 1
- XLYOFNOQVPJJNP-UHFFFAOYSA-N water Substances O XLYOFNOQVPJJNP-UHFFFAOYSA-N 0.000 description 1
Classifications
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61H—PHYSICAL THERAPY APPARATUS, e.g. DEVICES FOR LOCATING OR STIMULATING REFLEX POINTS IN THE BODY; ARTIFICIAL RESPIRATION; MASSAGE; BATHING DEVICES FOR SPECIAL THERAPEUTIC OR HYGIENIC PURPOSES OR SPECIFIC PARTS OF THE BODY
- A61H3/00—Appliances for aiding patients or disabled persons to walk about
-
- 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
- B25J9/1615—Programme controls characterised by special kind of manipulator, e.g. planar, scara, gantry, cantilever, space, closed chain, passive/active joints and tendon driven manipulators
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61H—PHYSICAL THERAPY APPARATUS, e.g. DEVICES FOR LOCATING OR STIMULATING REFLEX POINTS IN THE BODY; ARTIFICIAL RESPIRATION; MASSAGE; BATHING DEVICES FOR SPECIAL THERAPEUTIC OR HYGIENIC PURPOSES OR SPECIFIC PARTS OF THE BODY
- A61H3/00—Appliances for aiding patients or disabled persons to walk about
- A61H2003/005—Appliances for aiding patients or disabled persons to walk about with knee, leg or stump rests
Landscapes
- Health & Medical Sciences (AREA)
- Engineering & Computer Science (AREA)
- General Health & Medical Sciences (AREA)
- Epidemiology (AREA)
- Robotics (AREA)
- Mechanical Engineering (AREA)
- Orthopedic Medicine & Surgery (AREA)
- Pain & Pain Management (AREA)
- Physical Education & Sports Medicine (AREA)
- Rehabilitation Therapy (AREA)
- Life Sciences & Earth Sciences (AREA)
- Animal Behavior & Ethology (AREA)
- Public Health (AREA)
- Veterinary Medicine (AREA)
- Rehabilitation Tools (AREA)
Abstract
The present invention is a kind of band priori torque non-model control method of ectoskeleton, scanning laser range finder is fixed on the human body any position between user waist and knee joint by this method, perpendicular to landform in front of horizontal Surface scan, adjust the target graduation position of laser range finder, make its alignment underface i.e. standpoint of human body, landform is identified using scanning laser range finder；Knee joint angle appropriate is selected according to landform, knee joint ectoskeleton is controlled using the MFA control mode with priori torque, avoids the problem of the foundation of complex model and model inaccuracy in lower limb exoskeleton control.
Description
Technical field
The present invention relates to knee joint ectoskeleton control field, a kind of ectoskeleton based on landform identification technology is referred in particular to
Band priori torque non-model control method.
Background technique
The increasing of the elderly makes the auxiliary devices such as crutch, walk helper and wheelchair obtain demand to greatly increase, and active knee closes
Section ectoskeleton can help handicapped people to walk, and carry out active power-assisted.The control of knee joint ectoskeleton is the pass of active power-assisted
Where key, it is therefore desirable to which a kind of workload is small, controls the sufficiently high control strategy of precision.It is traditional to be controlled based on model
Method can be led to the problem of because of modeling difficulty or inaccuracy but control effect is bad.Data drive control refers to controller
Design is unrelated with the mathematical model of controlled device, but between online and offline I/O data and data based on controlled process
Implicit information carrys out control object.
Number of patent application: CN201810036674, a kind of lower limb rehabilitation exoskeleton system and its walking control method design
A kind of lower limb rehabilitation exoskeleton system and its walking control method.The upper body inclination data of this method acquisition ectoskeleton wearer,
The real time data of plantar pressure, joint of lower extremity angle identifies current gait phase.According to current gait to ectoskeleton
Movement controlled, leading leg for ectoskeleton is liftoff into the swing process that will be landed of leading leg, and the main supporting leg of controller is protected
Hold generally upstanding state.And criterion control is shifted according to center of gravity when leading leg liftoff and carries out liftoff wobbling action, so that upper body is inclined
Angle and plantar pressure are maintained in predetermined interval, effectively eliminate lateral tilting moment, and human body major part weight passes through ectoskeleton
Supporting leg bar is transmitted to ground, mitigates the burden of wearer's supporting leg.The patent has the disadvantages that (1) ectoskeleton is controlling
During leading leg, power-assisted movement is made without the accurate control moment of specific control method；(2) different terrain pair is not considered
The influence of walking.
Application number: CN201611124145, one kind being used for lower limb exoskeleton man-machine system closed-loop control sensor-based system.Inclining
Angle transducer, pressure sensor, acceleration transducer, torque sensor and code sensor are communicatively connected to signal acquisition platform
On.Joint driven torque is controlled according to collected new information ectoskeleton controller, then is communicated with drive system, thus shape
At closed-loop control.The patent has the disadvantages that the influence for not considering specific landform for torque of walking.
Summary of the invention
In view of the deficiencies of the prior art, the technical problems to be solved by the present invention are: providing a kind of knee joint ectoskeleton
Band priori torque non-model control method.The URG-04LX-UG01 scanning laser range finder of Bei Yang company, Japan is used first
Landform is identified.Knee joint angle appropriate is selected according to landform, uses the MFA control with priori torque
Method (MFAC) controls knee joint ectoskeleton, avoids in lower limb exoskeleton control the foundation of complex model and model not
Accurate problem.
The present invention solves technical solution used by the technical problem:
A kind of band priori torque non-model control method of ectoskeleton, steps are as follows:
The first step, laser ranging system setting
Scanning laser range finder is fixed on the human body any position between user waist and knee joint, perpendicular to water
Landform in front of flat scanning, adjusts the target graduation position of laser range finder, makes its alignment underface i.e. standpoint of human body,
Acquisition range is nearby arranged in target graduation position, and the length of acquisition range is 70~90 °, and 200 are acquired in data acquisition range
The data of~400 points；
Second step, data acquisition and pretreatment
Scanning laser range finder is surveyed the topography range data with human motion, and the sample data measured is carried out
Pretreatment rejects the significantly greater point of distance between consecutive points to improve the quality of data.257 points of each topographical scan away from
From data constitute an one-dimensional vector as one group of data, and according to level land, stair activity, five kinds of landform of upper downslope class
Other to mark to vector, i.e., when value is 1, corresponding scanning landform is level land, when value is 2, corresponding scanning
Shape be go upstairs, and so on value be 5 when, scanned landform be downslope.Every kind of landform acquires 1500 groups of data and combines shape
At sample database；
Third step carries out landform identification
The five kinds of terrain datas shown according to fig. 2, it can be seen that the distance between different terrain feature difference is obvious, selects
Effective classification to landform may be implemented in method according to distance classification.Therefore LVQ classifier is trained using sample database.
This patent selects LVQ classification method to be first randomly provided 5 one-dimensional vector conducts for one-dimensional input sample
Prototype vector, each prototype vector indicate a kind of terrain category.After inputting new sample, calculates input sample and each is former
The Euclidean distance of type vector, choose and its apart from nearest prototype vector generic be the sample class.Greatest iteration is set
Number is 1000 times.
Step1: default prototype vector number q is 5, and the terrain category label t of each prototype vector is arranged1,t2,...tq, t table
Show category label.It chooses learning rate α ∈ (0,1), initializes one group of prototype vector p1,p2,...pq。
Step2: j-th of sample (x is randomly selected from acquisition dataj,yj), calculate xjWith each prototype vector pi's
Distance obtains dji:
dji=| | xj-pi||2, i=1,2 ... q (1)
And it finds and xjApart from nearest prototype vector pk, wherein k=argmini∈{1,2,...q}dji.
Step3: judge choose sample category label it is whether consistent with prototype vector category label come update prototype to
Measure pkIf the category label of the two is identical, p is enabledkTo xjDirection draw close, otherwise update pkFar from xjDirection.
In formula (2), yjIndicate xjTerrain category label, tkIndicate pkCategory label.
Step4: judge whether to be more than the number of iterations, be to export prototype vector, otherwise return to Step2.
Can be realized after the one group of prototype vector that learns to classification of landform, to arbitrary input sample x, will be divided to
In its classification representated by the nearest prototype vector.
Landform is identified using trained classifier, and result is used for the ectoskeleton of following 5th step
Control.
4th step establishes human body lower limbs model and obtains priori torque
The human body lower limbs model that can simulate human body walking is established in SolidWorks software, each to human mould lower limb type
After articular portion setting constraint, then in each joint motions pair setting rotate driving of human body lower limbs model, respectively closed with the lower limb of acquisition
Angle is saved as drive volume；
When controlling ectoskeleton movement, by setting the motion profile of ectoskeleton joint angles, control is applied to ectoskeleton
On torque keep the human body consistent with the movement of ectoskeleton, realize ectoskeleton to the power-assisted of human motion, using ViconMX tri-
Tie up gait analysis system collection analysis joint of lower extremity angle and motion feature；
It is sticked in human hip, knee joint, thigh, shank, ankle-joint, toe and heel lower extremity left and right sides infrared
Reflective spot is as marker, and when there is two or more cameras to take same marker, three-dimensional gait analysis system is
The position that can determine marker, by the Chang, Kua width of the height, weight, leg of experimenter, knee joint width and ankle-joint width these
Essential information is input in three-dimensional gait analysis system, and the motion profile of human body and the angle letter in each joint is obtained by calculation
Breath；
Manikin is driven with the above-mentioned each joint angles being calculated, i.e., it is the joint angles of different terrain are defeated
Enter into model, obtains corresponding priori torque u under different terrainprior；
5th step establishes knee joint ectoskeleton dynamical linearization model
Man machine exoskeleton model.Using lagrangian dynamics analytic approach, lower limb dynamic system universal model can be obtained
Following form:
In formula (3), θ is joint angles vector,For joint angular acceleration,For joint angular acceleration, M (θ) is positive definite
Inertia matrix,It is gravity continuous item matrix for Ge Shili and centrifugal force continuous item matrix, G (θ), τ is ectoskeleton joint
Torque, τhThe torque of ectoskeleton is acted on wherein for human body, and parameters are to consider wearing lower limb exoskeleton to manikin shadow
Parameter after sound.
Formula (3), which can arrange, is
Control situation when θ is knee joint angle in invention herein, thus system is converted to single-input single-output system.It is fixed
Adopted ectoskeleton knee joint torque τ is input u, and knee joint angle θ is output y.Wherein, knee joint angle is defined as the femur longitudinal axis and prolongs
The angle of long line and shin bone longitudinal axis parallel lines, buckling are positive, and stretching, extension is negative.
For a discrete system, C (k), G (k), M (k) in k moment formula (4) are specifically to be worth, and human body is made
Torque τ for ectoskeletonh(k) it is also bounded, formula (4) discrete can turn to following form:
When the sampling time of discrete system is sufficiently small, if the sampling time is T,It is represented by
It willSubstitution formula (5) can obtain:
It is hereinafter described for convenience, the output variation of two adjacent moments is defined here, input variation is respectively as follows:
Δ y (k)=y (k)-y (k-1) (9)
Δ u (k)=u (k)-u (k-1) (10)
When | | [y (k), u (k)]T| | when ≠ 0, it there will necessarily be time-varying parameter vector Φ (k)=[φ of pseudo- gradient1(k),φ2
(k)], Φ (k) is a time-varying vector, φ1(k),φ2It (k) is element therein, so that system (8) can be converted into full format and move
State inearized model
Δ y (k+1)=φ1(k)Δy(k)+φ2(k)Δu(k) (11)
Gradient time-varying vector Φ (k)=[φ in formula (11)1(k),φ2(k)] value can be solved by following method.Due to
Φ (k) is a time-varying vector, so its exact value is difficult to obtain, estimates pseudo- gradient using modified projection algorithm for estimating
The value of vector Φ (k).
In formula (12), μ > 0 is weight factor,It is the estimated value to Φ (k), differential is carried out to Φ (k) and is enabled
As a result zero, it can obtain:
Step factor η ∈ in formula (13) and (14) (0,2] it estimatesWith
6th step, knee joint ectoskeleton controller design
To prevent control algolithm from generating the tracking error and excessive control input of stable state, using input criterion control as follows
Function designs controller.
J(uMFA(k))=(y*(k+1)-y(k+1))2+λ(uMFA(k)-uMFA(k-1))2 (15)
Wherein, λ is the weight factor for limiting control input quantity, uMFA(k) it is exported for MFA control, y*
(k+1) formula (11) are substituted into formula (15) to u for desired output signalMFA(k) it derivation and enables as a result zero, can be controlled
Output are as follows:
In formula (16), and step factor ρ ∈ (0,1], since gradient time-varying vector Φ (k) is difficult to calculate, use
(13) it is replaced with the pseudo-time gradient of (14) formula.U can be solved by being substituted into (16) formulaMFA(k) i.e. model-free adaption control
Make the value of output.
The knee joint priori torque u of corresponding 4th step is chosen according to third step landform recognition resultprior.In order to avoid elder generation
Excessive influence of the torque on control moment is tested, exports u for MFA controlMFAWith knee joint priori torque upriorDistribution
Weight is respectively 1-F and F, and wherein [0,1] F ∈, influence of the more big then priori torque of F to control moment are bigger.
Finally obtaining controller form is formula (17):
U (k)=(1-F) uMFA(k)+Fuprior(k) (17)
Above-mentioned band priori torque non-model control method, wherein used equipment is obtained by known approach.
Compared with prior art, marked improvement of the invention is as follows:
(1) to the foundation foundation of human body ectoskeleton model Lagrangian method in the present invention.Then further foundation
Former discrete system is converted into full format dynamical linearization model using the time-varying parameter vector of pseudo- gradient by model discretization,
In the case where not producing bigger effect to calculated result, calculation process is enormously simplified.
(2) present invention algorithmically compared to traditional model-free methods more considers priori torque.The input of controller is
MFA control exports uMFAWith knee joint priori torque upriorMultiply weight 1-F respectively and is added again with F.It can be according to reality
Situation adjusts weight F to improving the performance of controller, reduces the gap between actual value and expectation.Simulation result is aobvious
Show in support stage phase output angle and expected angle error within 3 °, is difficult to putting based on MFA control method
Dynamic stage phase makes knee joint angle quickly follow the variation of expected angle, and merges the model-free adaption of priori moment information
Control method can realize good tracking effect in stage shaking peroid.It can be seen that the side for considering priori torque by Fig. 6, Fig. 7
Method error is more much smaller than what is do not accounted for.
Detailed description of the invention
Present invention will be further explained below with reference to the attached drawings and examples.
Fig. 1 is data acquisition schematic diagram in mesorelief of the present invention identification.
Fig. 2 (a) is the data of the level land landform of sensor acquisition in the present invention.
Fig. 2 (b) be the present invention in sensor acquisition level land to landform of going upstairs data.
Fig. 2 (c) be the present invention in sensor acquisition level land to landform of going downstairs data.
Fig. 2 (d) be the present invention in sensor acquisition level land to upslope landform data.
Fig. 2 (e) be the present invention in sensor acquisition level land to downslope landform data.
Fig. 3 is LVQ algorithm flow chart in the present invention.
Fig. 4 is ViconMX three-dimensional gait analysis system acquisition system in the present invention
Fig. 5 is the control system architecture figure of knee joint ectoskeleton in the present invention.
Fig. 6 is the emulation gait figure that ectoskeleton is dressed in the present invention.
Fig. 7 be the present invention in the torque containing priori MFAC and tradition without priori torque MFAC method knee joint angle with
With effect contrast figure.
Fig. 8 be the present invention in the torque containing priori MFAC and tradition without priori torque MFAC method knee joint angle with
With error comparison diagram.
Specific embodiment
Embodiment illustrated in fig. 1 shows data acquisition in landform identification.The URG-04LX- of Japanese Bei Yang company
UG01 scanning laser range finder and computer are attached setting, after be fixed on the waist of user, user is in various landform
Walking determination data.
Fig. 2 (a)-(e) illustrated embodiment shows five kinds of terrain datas of laser range finder acquisition.To the data after acquisition
It is pre-processed, distance differs biggish point to improve the quality of data between rejecting consecutive points.It is horizontal in the figure of five kinds of terrain datas
The ground level distance of coordinate representation landform, ordinate indicate the vertical distance with ground, and the point in figure indicates laser range finder
257 points of acquisition.
Embodiment illustrated in fig. 3 is the process handled data, the i.e. algorithm flow chart of LVQ.First pass through the algorithm into
Row training.After the completion of training after inputting new sample, the Euclidean distance of input sample and each prototype vector is calculated, is chosen
With its apart from nearest prototype vector generic be the sample class.
Embodiment illustrated in fig. 4 shows ViconMX three-dimensional gait analysis system acquisition system specific structure.Use three-dimensional gait
Each joint angles are used to carry out manikin driving and obtained by analysis system collection analysis joint of lower extremity angle and motion feature
Priori torque.
The control system of knee joint ectoskeleton is shown in embodiment illustrated in fig. 5.The corresponding phase is selected according to different terrain
Hope that angle track is controlled to coordinate man-machine system under different road conditions.Model-free adaption is designed according to input criterion function
Controller, desired joint angles y*(k) it is made the difference with the output y (k-1) at k-1 moment and obtains Δ y (k), Δ y (k) and uMFA
(k-1) it is inputted together as no mould adaptive controller, show that it exports uMFA.Wherein, y (k-1) and uMFA(k-1) pass through delay
Link Z-1To obtain.UMFAWith the priori torque u selected according to landformpriorIt is added according to a certain percentage and is used as knee joint dermoskeleton
The input torque of bone model.Use the Δ u obtained by differentiatorMFA(k-1) gradient estimate vector is calculated with Δ y (k)To adjust MFAC.Priori torque is the joint moment emulated under different road conditions, is added in controller output
Data variation trend when adding a certain proportion of priori torque to walk using human normal between torque improves the accurate of control
Property.
Fig. 6 institute embodiment, which is shown, drives knee joint using the control moment that MATLAB is calculated, by expectation joint angle
Degree drives other each joints, the human body walking gait figure emulated.Compared to the control method that priori torque is not added, the method control
Effect processed is more preferable, demonstrates the feasibility of the application control method.
The simulated effect and tracking error comparison diagram that Fig. 7, Fig. 8 are.Simulation result shows the controlling party based on MFAC
Method is difficult to make knee joint angle quickly follow the variation of expected angle in stage shaking peroid, and merges priori moment information
MFAC control method can realize that better tracking effect, tracking error are smaller in stage shaking peroid.
Detailed description are as follows for the theoretical foundation of the method for the present invention:
Knee joint ectoskeleton kinematic system is a high granular time nonlinear system.MFA control
(MFAC) Discrete time Nonlinear Systems are directed to, controlled device is controlled based on the inputoutput data of system, passes through dynamic
Linearization technique converts the value at complication system discrete point to the time-varying linearized data of incremental form.Using in each dynamic
The quasi equivalent dynamical linearization modelling controller of a kind of void that operating point is established, therefore MFA control algorithm is several
Be free of the parameter model of controlled process.Using projection algorithm for estimating to time-varying parameter puppet partial derivative (the pseudo- ladder in data model
Degree) estimated, theoretically, the I/O based on system establishes dynamical linearization model, MFA control algorithm
There is no Unmarried pregnancy problems, have certain interference rejection ability.
MFA control algorithm (MFAC) be based on dynamical linearization method can be divided into tight format dynamical linearization, partially
Format dynamical linearization and three kinds of full format dynamical linearization are described below by taking full format dynamical linearization as an example and carry out to MFAC
It introduces.
For SISO discrete-time system, form can be indicated are as follows:
Y (k+1)=f (y (k) ..., y (k-ny),u(k),...,u(k-nu)) (18)
Wherein, y (k) ∈ R, u (k) ∈ R respectively indicate system in the output and input at k moment；ny, nuIt is two unknown
Positive integer；F (...) indicates unknown nonlinear function.
Assume when (18) formula meets:
Assuming that 1: in addition to finite time point, f (...) is about n-thyThe partial derivative of+2 variables is continuous.
Assuming that 2: system (18) meets generalized Lipschitz condition, for any k1≠k2,k1,k2>=0 and u (k1)≠u
(k2) when, have
|y(k1+1)-y(k2+1)|≤b|u(k1)-u(k2)| (19)
Wherein b > 0 is a constant, y (ki+ 1)=f (y (ki),...,y(ki-ny),u(ki),...,u(ki-nu)),i
=1,2.For convenience of the narration to system, the output variation of two adjacent moments is defined here, input variation is respectively
Δ y (k)=y (k)-y (k-1) (20)
Δ u (k)=u (k)-u (k-1) (21)
Assume 1 for meeting, it is assumed that 2 system, as | Δ u (k) | when ≠ 0, system (18) can dynamical linearization be expressed as
Following data model
Δ y (k+1)=φc(k)Δu(k) (22)
Wherein φcIt (k) is the time-varying parameter introduced to any time k bounded, referred to as pseudo- partial derivative.
Proof procedure is as follows:
It can be obtained by formula (18), (20)
It enables
By assuming 1 and Cauchy Order Derivatives in Differential Mid-Value Theorem, (23) formula can be converted are as follows:
WhereinIt is y (k+1) about n-thyThe partial derivative of+2 variables [y (k) ..., y (k-ny),u(k)...,
u(k-nu)][y(k),...,y(k-ny),u(k-1),u(k-1),...,u(k-nu)] between value at certain point.
When the k moment, the equation for containing variable η (k) is considered:
Due to Δ u (k) ≠ 0, therefore equation (26) existence and unique solution η*(k), it enables
Therefore (23) formula can be written as
Δ y (k+1)=φc(k)Δu(k) (28)
Proof finishes.
Priori torque is the joint moment emulated under different road conditions in the present invention.First landform is identified, then
Corresponding priori torque is set according to different landform, priori torque is an important parameter in control method.In controller
Data variation trend when the middle a certain proportion of priori torque of addition can use human normal walking between torque improves control
Accuracy, and then guarantee control precision.
Embodiment 1
The first step, laser ranging system setting
The URG-04LX-UG01 scanning laser range finder of Bei Yang company, Japan is selected, the ranging of phase shift rangefinder method is based on.
The light source of sensor is the infrared laser that wavelength is 785nm, has 1 grade of laser safety, ranging range is that 20mm is arrived
5600mm, its measurement error is ± 30mm within the scope of 60mm to 1000mm, measures and misses in the range of 1000mm to 4095mm
Difference is within the 3% of measurement distance.The weight of sensor is only 160g, is convenient for wearable design.Scanning angle range is 240 °,
0.36 ° of angular resolution (every 360 ° of scanning can generate 1024 data points), scan frequency 10Hz/s, i.e., each scan period
It is completed in 0.1s, meets the needs of to sensor rapidity.
Scanning laser range finder is fixed on user waist leftward position, perpendicular to landform in front of horizontal Surface scan,
It is the standpoint of human body immediately below the position alignment that 180 ° of laser range finder.The acquisition range of scanning laser range finder is set
It sets, data acquisition range is the data of 257 points in 110 °~200 °.Meanwhile the range by reducing sensor scanning can
Avoid the range error of sensor to the influence of recognition result, the measurement range of scanning laser range finder used herein is in 1m
When between 4m, range error is being measured within distance.
Second step, data acquisition and pretreatment
Scanning laser range finder is mounted on position between human lumbar, range data of surveying the topography with human motion, due to
It is only 0.1 second that it is practical, which to scan a cycle, for sensor, and human body is during this period of time walked about apart from smaller, the identification to terrain data
The data for influencing less, therefore testing measurement are that laser range finder in human body loins fixed bit sets measurement.Each landform is swept
The range data for retouching 257 points constitutes an one-dimensional vector as one group of data, and according to level land, stair activity, upper oblique
The classification of the five kinds of landform in slope marks vector, i.e., when value is 1, corresponding scanning landform is level land, when value is 2,
Corresponding scanning landform be go upstairs, and so on value be 5 when, scanned landform be downslope.For measuring every group of number
According to being pre-processed, the biggish point of distance between consecutive points is rejected in 257 points of same group to improve the quality of data.Between consecutive points
Away from it is excessive it is larger may be measuring error, therefore reject consecutive points spacing be greater than 300mm point.Every kind of landform acquires 1500 groups of numbers
According to identifying data sample using treated data as landform.
Third step carries out ground row identification
The five kinds of terrain datas shown according to Fig. 2 (a)-(e) in a upper section, it can be seen that the distance between different terrain
Feature difference is obvious, selects the method according to distance classification that effective classification to landform may be implemented.It is instructed using sample data
Practice for training LVQ classifier.
The present invention selects LVQ classification method, and for one-dimensional input sample, LVQ is by there is the study of supervision to obtain one group
Prototype vector, each prototype vector indicate a kind of classification.After inputting new sample, calculate input sample and each prototype to
The Euclidean distance of amount, choose and its apart from nearest prototype vector generic be the sample class.
Step1: default prototype vector number is q, and the terrain category label t of each prototype vector is arranged1,t2,...tq, wherein
t∈(1,2,...,5).It chooses learning rate α ∈ (0,1), initializes one group of prototype vector p1,p2,...pq。
Step2: j-th of sample (x is randomly selected from acquisition dataj,yj), calculate xjWith each prototype vector pi's
Distance obtains dji:
dji=| | xj-pi||2, i=1,2 ... q (1)
And it finds and xjApart from nearest prototype vector pk, wherein k=argmini∈{1,2,...q}dji.
Step3: judge choose sample category label it is whether consistent with prototype vector category label come update prototype to
Measure pkIf the category label of the two is identical, p is enabledkTo xjDirection draw close, otherwise update pkFar from xjDirection.
In formula (2), yjIndicate xjTerrain category label, tkIndicate pkCategory label.
Step4: judge whether to be more than the number of iterations, be to export prototype vector, otherwise return to Step2.
Can be realized after the one group of prototype vector that learns to road conditions classify, to arbitrary input sample x, will be divided to
In its classification representated by the nearest prototype vector.
Landform is identified using trained classifier, and result is used for next ectoskeleton and is controlled.
4th step establishes human body lower limbs model and obtains priori torque
Corresponding lower limb model is established in SolidWorks software, as shown in Figure 1, being saved as parasolid's
Format is simultaneously directed into ADAMS software.In ADAMS software, freedom of motion is constrained by kinematic pair module, commonly
Constraints module has fixed joint (FixedJoint), revolute (Revolute Joint), sliding pair (Translational
Joint) etc..Fixed joint the constraint relationship is that two components are fixed together, so that not having opposite movement between two components；
Revolute constrains two components can only relative rotation around rotary shaft at certain point；Sliding two components of secondary constraint can only be along certain
Slide axis sliding.Herein according to the functional relationship in the practical walking process of human body to model add joint pair, wherein head and
It is fixed to keep the opposing stationary of the two that fixed joint is added between upper trunk；Revolute is added between two arms and trunk to simulate people
Swinging arm in body walking；The hip joint of left and right leg, knee joint and ankle-joint addition revolute bend and stretch fortune simulate lower limb
It is dynamic.The normal walking of human body is mainly the movement in sagittal plane, therefore adds planar contact pair at trunk on human body, so that human mould
Type is walked in sagittal plane, establishes ground model and contact of the foot with ground is arranged to simulate foot by Contact module
With the interaction force on ground.
After to the setting constraint of each articular portion of manikin, in order to turn joint according to the expected angle track of setting
It is dynamic, it needs to apply corresponding Motion driving in each constraint pair.Driving is actually other freedom that kinematic pair is unconstrained
Degree further constraint, allows it to change according to certain rule.Common driving has rotation to drive and slide driving, rotates driving restraint
First component, by the rule rotation of setting, slides with respect to second component of first component of driving restraint relative to second component
According to the rule sliding of setting.Rotate driving is arranged in each joint motions pair of manikin, is made with each joint angles of the lower limb of acquisition
For drive volume.
When controlling ectoskeleton movement, by setting the motion profile of ectoskeleton joint angles, control is applied to ectoskeleton
On torque keep the human body consistent with the movement of ectoskeleton, realize ectoskeleton to the power-assisted of human motion.This patent uses
ViconMX three-dimensional gait analysis system acquisition analyzes joint of lower extremity angle and motion feature.
ViconMX three-dimensional gait analysis system includes 6 high speed MX infrared shooting heads, MX component, PC host and the periphery MX
External member is as shown in Figure 4.Wherein MXNet is used to provide power supply for video camera, provides the information of PC host and video camera as intermediary
Exchange.MX Link is to enhance the data-transformation facility between host and MX Link to improve the real-time of system.MX
Control provides interface, including myoelectricity instrument, force plate etc. between system and third party device.WORKSTATION is system
Core processing software, for system initial calibration and to video camera acquisition image procossing.
It is sticked in human hip, knee joint, thigh, shank, ankle-joint, toe and heel lower extremity left and right sides infrared
Reflective spot is as marker, and when there is two or more cameras to take same marker, system software processing part is
The position that can determine marker, by the Chang, Kua width of the height, weight, leg of experimenter, knee joint width and ankle-joint width these
Essential information is input in ViconMX three-dimensional gait analysis system, motion profile and each pass by the available human body of calculating
Save desired angle information.
Model is driven with desired angle, the joint angles of different terrain are input in model, are corresponded to
Priori torque uprior。
5th step establishes knee joint ectoskeleton dynamical linearization model
Man machine exoskeleton model is as shown in figure attachment 1, and position is used between wherein scanning laser range finder is placed on human lumbar
The following form of lower limb dynamic system universal model can be obtained in lagrangian dynamics analytic approach:
In formula (3), θ is joint angles vector, and M (θ) is positive definite inertia matrix,For Ge Shili and centrifugal force phase
Item matrix is closed, G (θ) is gravity continuous item matrix, and τ is ectoskeleton joint moment, τhFor human body act on ectoskeleton torque its
In, parameters are to consider wearing lower limb exoskeleton on the parameter after manikin influence.
Formula (3), which can arrange, is
Control situation when θ is knee joint angle in invention herein, thus system is converted to single-input single-output system.It is fixed
Adopted ectoskeleton knee joint torque τ is input u, and knee joint angle θ is output y.Wherein, knee joint angle is defined as the femur longitudinal axis and prolongs
The angle of long line and shin bone longitudinal axis parallel lines, buckling are positive, and stretching, extension is negative.
For a discrete system, C (k), G (k), M (k) in k moment formula (4) are specifically to be worth, and human body is made
Torque τ for ectoskeletonh(k) it is also bounded, formula (4) discrete can turn to following form:
When the sampling time of discrete system is sufficiently small, if the sampling time is T,It is represented by
It willSubstitution formula (5) can obtain:
It is hereinafter described for convenience, the output variation of two adjacent moments is defined here, input variation is respectively as follows:
Δ y (k)=y (k)-y (k-1) (9)
Δ u (k)=u (k)-u (k-1) (10)
When | | [y (k), u (k)]T| | when ≠ 0, it there will necessarily be time-varying parameter vector Φ (k)=[φ of pseudo- gradient1(k),φ2
(k)], so that system (8) can be converted into full format dynamical linearization model.
Δ y (k+1)=φ1(k)Δy(k)+φ2(k)Δu(k) (11)
Value Φ (k)=[φ of gradient time-varying vector in formula (11)1(k),φ2(k)] it can be solved by following method.Φ(k)
It is a time-varying vector, so its exact value is difficult to obtain, pseudo- gradient vector is estimated using modified projection algorithm for estimating
The value of Φ (k).
In formula (12), μ > 0 is weight factor,It is the estimated value to Φ (k), differential is carried out to Φ (k) and is enabled
As a result zero, it can obtain:
Step factor η ∈ in formula (13) and (14) (0,2] it estimatesWith
5th step, knee joint ectoskeleton controller design
To prevent control algolithm from generating the tracking error and excessive control input of stable state, using input criterion control as follows
Function designs controller.
J(uMFA(k))=(y*(k+1)-y(k+1))2+λ(uMFA(k)-uMFA(k-1))2 (15)
Wherein, λ is the weight factor for limiting control input quantity, uMFA(k) it is exported for MFA control, y*
(k+1) formula (11) are substituted into formula (15) to u for desired output signalMFA(k) it derivation and enables as a result zero, can be controlled
Output are as follows:
In formula (16), and step factor ρ ∈ (0,1], since gradient time-varying vector Φ (k) is difficult to calculate, use
(13) it is replaced with the pseudo-time gradient of (14) formula.U can be solved by being substituted into (16) formulaMFA(k) i.e. model-free adaption control
Make the value of output.
uMFA(k) the control output calculated for model-free adaptive controller, in order to preferably to knee joint ectoskeleton control
System imports each joint angles data of acquisition in ADAMS simulation software, by driving function CUBSPL function to manikin
It is driven, knee joint priori torque can be obtained in emulation.Adding a certain proportion of priori torque in the controller can use people
Data variation trend when body normal walking between torque improves the accuracy of control.
The knee joint priori torque u of corresponding 4th step is chosen according to third step landform recognition resultprior.In order to avoid
Excessive influence of the priori torque on control moment exports u for MFA controlMFAWith knee joint priori torque upriorPoint
It is respectively 1-F and F with weight, wherein [0,1] F ∈, influence of the more big then priori torque of F to control moment are bigger.
It is as follows to finally obtain controller form:
U (k)=(1-F) uMFA(k)+Fuprior(k) (17)
6th step, software emulation verifying
Associative simulation is carried out using SolidWorks software, ADAMS simulation software and MATLAB.SolidWorks software base
On plinth, rigid model is established according to human body actual size size, while having built the joint of MATLAB/Simulink and ADAMS
Experiment simulation platform carries out simplation verification using above-mentioned control algolithm.
For simulating level land road conditions, knee joint is driven using the control moment that MATLAB is calculated, by it is expected joint
Angle drives other each joints.Fig. 6 is the obtained human body walking gait figure of emulation, wherein each parameter of controller be set as ρ=
1, η=1, λ=0.8, μ=2, F=0.1 are the weight of priori Torque distribution 10% in the controller.It can be seen in gait figure
Normal walking may be implemented in human body under ectoskeleton power-assisted out, falls phenomena such as toppling over without falling, illustrates the control method proposed herein
There is good effect to the Power assisted control of knee joint ectoskeleton.
Selected simulation time is three gait cycles of human body normal walking, the simulated effect that Fig. 7, Fig. 8 are
It restrains and forces since 42 ° in emulation start time knee joint angle due to model initialization with tracking error comparison diagram
Nearly expected angle.Simulation result is shown in support phase stage output angle and expected angle error within 3 °, based on MFAC's
Control method is difficult to make knee joint angle quickly follow the variation of expected angle in stage shaking peroid, and merges priori torque letter
The MFAC control method of breath can realize good tracking effect in stage shaking peroid.
The present invention does not address place and is suitable for the prior art.
Claims (4)
1. a kind of the step of band priori torque non-model control method of ectoskeleton, this method, is:
The first step, laser ranging system setting:
Scanning laser range finder is fixed on the human body any position between user waist and knee joint, perpendicular to horizontal plane
Scanning front landform, adjusts the target graduation position of laser range finder, makes its alignment underface i.e. standpoint of human body, in target
Acquisition range is nearby arranged in graduation position, and the length of acquisition range is 70~90 °, and 200~400 are acquired in data acquisition range
The data of a point；
Second step, data acquisition and pretreatment:
Scanning laser range finder is surveyed the topography range data with human motion, and the sample data measured is located in advance
Reason, rejects the significantly greater point of distance between consecutive points, the range data of 257 points of each topographical scan constitute one it is one-dimensional
Vector marks vector as one group of data, and according to the classification of five kinds of level land, stair activity, upper downslope landform, i.e., ought take
When value is 1, corresponding scanning landform is level land, and when value is 2, corresponding scanning landform is to go upstairs, and so on
When value is 5, scanned landform is downslope；Every kind of landform acquisition is no less than 1000 groups of data and combines to form sample database；
Third step trains classifier using the sample database of second step, carries out landform identification；
4th step establishes human body lower limbs model and obtains priori torque:
The human body lower limbs model that can simulate human body walking is established in SolidWorks software, to each joint of human mould lower limb type
After section sets constraint, then in each joint motions pair setting rotate driving of human body lower limbs model, with each joint angle of the lower limb of acquisition
Degree is used as drive volume；
When controlling ectoskeleton movement, by setting the motion profile of ectoskeleton joint angles, control is applied on ectoskeleton
Torque keeps the human body consistent with the movement of ectoskeleton, realizes that ectoskeleton to the power-assisted of human motion, is walked using ViconMX three-dimensional
State analysis system collection analysis joint of lower extremity angle and motion feature；
Infrared reflecting is sticked in human hip, knee joint, thigh, shank, ankle-joint, toe and heel lower extremity left and right sides
Point is used as marker, and when there is two or more cameras to take same marker, three-dimensional gait analysis system can be true
The position for determining marker, by height, weight, the leg length of experimenter, Kua width, knee joint width and ankle-joint width, these are basic
The motion profile of human body and the angle information in each joint is obtained by calculation into three-dimensional gait analysis system in information input；
Manikin is driven with the above-mentioned each joint angles being calculated, i.e., the joint angles of different terrain are input to
In model, corresponding priori torque u under different terrain is obtainedprior；
5th step establishes knee joint ectoskeleton dynamical linearization model:
Man machine exoskeleton model is established using lagrangian dynamics analytic approach, defining ectoskeleton knee joint torque τ is input u,
Knee joint angle θ is that output y, knee joint angle θ are defined as the angle of femur longitudinal axis extension line Yu shin bone longitudinal axis parallel lines, buckling
It is positive, stretching, extension is negative；Then the man machine exoskeleton model discretization of foundation, using pseudo- gradient time-varying parameter vector it is former from
Dissipating system converting is dynamical linearization model, and the value of gradient time-varying vector is estimated with projection algorithm for estimating；
6th step, knee joint ectoskeleton controller design:
Input criterion function, which is controlled, using formula (15) designs knee joint ectoskeleton controller,
J(uMFA(k))=(y*(k+1)-y(k+1))2+λ(uMFA(k)-uMFA(k-1))2 (15)
Wherein, λ is the weight factor for limiting control input quantity, uMFA(k) it is exported for MFA control, y*(k+1)
It is expected output signal, dynamical linearization model is substituted into formula (15) to uMFA(k) it derivation and enables as a result zero, obtains nothing
Model self-adapted control output are as follows:
In formula (16), ρ is step factor, and ρ ∈ (0,1], φ1(k),φ2It (k) is gradient time-varying vector, y (k) is the k moment
Output；Δ y (k)=y (k)-y (k-1)；
To arbitrary input sample x, using third step, trained classifier identifies landform, according to third step
Shape recognition result chooses the knee joint priori torque u of corresponding 4th stepprior, and u is exported for MFA controlMFAWith
Knee joint priori torque upriorDistributing weight is respectively 1-F and F, and wherein [0,1] F ∈, obtains knee joint ectoskeleton controller u
(k) form is formula (17):
U (k)=(1-F) uMFA(k)+Fuprior(k) (17)。
2. control method according to claim 1, which is characterized in that the classifier is LVQ classifier, is first randomly provided
For the vectors of 5 257 dimensions as 5 prototype vectors, each prototype vector indicates a kind of terrain category, calculate input sample with it is each
The Euclidean distance of a prototype vector, choose and its apart from nearest prototype vector generic be the sample class.
3. control method according to claim 1, which is characterized in that dynamical linearization model is divided into tight format dynamic linear
Change, inclined format dynamical linearization or full format dynamical linearization.
4. control method according to claim 3, which is characterized in that full format dynamical linearization model is formula (11)
Δ y (k+1)=φ1(k)Δy(k)+φ2(k)Δu(k) (11)
Value Φ (k)=[φ of gradient time-varying vector in formula (11)1(k),φ2(k)] formula (13) is respectively adopted and (14) calculate
φ1(k),φ2(k) estimated valueWith
Wherein, μ > 0 is weight factor, η be step factor η ∈ (0,2],It is the estimated value to Φ (k), Φ (k)=[φ1
(k),φ2(k)]。
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910618865.7A CN110327187B (en) | 2019-07-10 | 2019-07-10 | Model-free control method with prior moment for exoskeleton |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910618865.7A CN110327187B (en) | 2019-07-10 | 2019-07-10 | Model-free control method with prior moment for exoskeleton |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN110327187A true CN110327187A (en) | 2019-10-15 |
| CN110327187B CN110327187B (en) | 2021-07-13 |
Family
ID=68145132
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910618865.7A Expired - Fee Related CN110327187B (en) | 2019-07-10 | 2019-07-10 | Model-free control method with prior moment for exoskeleton |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110327187B (en) |
Cited By (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110716557A (en) * | 2019-11-18 | 2020-01-21 | 河北工业大学 | Machine parameter identification and contact force monitoring method based on priori dynamics knowledge |
| CN110711114A (en) * | 2019-10-17 | 2020-01-21 | 河北工业大学 | Exoskeleton active disturbance rejection control method based on dynamic model |
| CN111403019A (en) * | 2020-03-20 | 2020-07-10 | 国家康复辅具研究中心秦皇岛研究院 | Ankle joint prosthesis model establishing method, model-free control method and verification method |
| CN111568432A (en) * | 2020-04-06 | 2020-08-25 | 沈阳工业大学 | Fall detection method of omnibearing walking training robot based on fuzzy inference |
| CN111930135A (en) * | 2020-08-12 | 2020-11-13 | 深圳航天科技创新研究院 | Active power-assisted control method and device based on terrain judgment and exoskeleton robot |
| CN112008726A (en) * | 2020-08-28 | 2020-12-01 | 浙江协力机器人技术有限公司 | Composite finite time control method based on exoskeleton robot actuator |
| CN113977582A (en) * | 2021-11-12 | 2022-01-28 | 杭州电子科技大学 | Lower limb exoskeleton stair climbing control method based on DMP |
Citations (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| EP0488691A3 (en) * | 1990-11-30 | 1993-02-24 | Honda Giken Kogyo Kabushiki Kaisha | System for controlling locomotion of legged walking robot |
| JP2006075456A (en) * | 2004-09-13 | 2006-03-23 | Tohoku Univ | Wearing type support system based on human body model |
| CN102551995A (en) * | 2012-01-20 | 2012-07-11 | 沈阳工业大学 | Real-time detecting method for knee joint angles and device |
| WO2012100250A1 (en) * | 2011-01-21 | 2012-07-26 | Iwalk, Inc. | Terrain adaptive powered joint orthosis |
| CN103263339A (en) * | 2013-05-17 | 2013-08-28 | 大连交通大学 | Exoskeleton walk-assisting robot for old people and bionic control method for anti-falling gaits |
| CN104698848A (en) * | 2015-02-11 | 2015-06-10 | 电子科技大学 | Control method for rehabilitation training of lower extremity exoskeleton rehabilitation robot |
| CN106125087A (en) * | 2016-06-15 | 2016-11-16 | 清研华宇智能机器人（天津）有限责任公司 | Dancing Robot indoor based on laser radar pedestrian tracting method |
| CN106778445A (en) * | 2015-11-20 | 2017-05-31 | 沈阳新松机器人自动化股份有限公司 | Service robot vision based on Face datection leads method |
| CN109848990A (en) * | 2019-01-28 | 2019-06-07 | 南京理工大学 | Knee joint ectoskeleton gain-variable model-free angle control method based on PSO |
-
2019
- 2019-07-10 CN CN201910618865.7A patent/CN110327187B/en not_active Expired - Fee Related
Patent Citations (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| EP0488691A3 (en) * | 1990-11-30 | 1993-02-24 | Honda Giken Kogyo Kabushiki Kaisha | System for controlling locomotion of legged walking robot |
| JP2006075456A (en) * | 2004-09-13 | 2006-03-23 | Tohoku Univ | Wearing type support system based on human body model |
| WO2012100250A1 (en) * | 2011-01-21 | 2012-07-26 | Iwalk, Inc. | Terrain adaptive powered joint orthosis |
| CN102551995A (en) * | 2012-01-20 | 2012-07-11 | 沈阳工业大学 | Real-time detecting method for knee joint angles and device |
| CN103263339A (en) * | 2013-05-17 | 2013-08-28 | 大连交通大学 | Exoskeleton walk-assisting robot for old people and bionic control method for anti-falling gaits |
| CN104698848A (en) * | 2015-02-11 | 2015-06-10 | 电子科技大学 | Control method for rehabilitation training of lower extremity exoskeleton rehabilitation robot |
| CN106778445A (en) * | 2015-11-20 | 2017-05-31 | 沈阳新松机器人自动化股份有限公司 | Service robot vision based on Face datection leads method |
| CN106125087A (en) * | 2016-06-15 | 2016-11-16 | 清研华宇智能机器人（天津）有限责任公司 | Dancing Robot indoor based on laser radar pedestrian tracting method |
| CN109848990A (en) * | 2019-01-28 | 2019-06-07 | 南京理工大学 | Knee joint ectoskeleton gain-variable model-free angle control method based on PSO |
Non-Patent Citations (8)
| Title |
|---|
| NILI ELIANA KRAUSZ，TOMMASO LENZI，LEVI J. HARGROVE: "《Depth Sensing for Improved Control of Lower Limb Prostheses》", 《TRANSACTIONS ON BIOMEDICAL ENGINEERING》 * |
| RYAN STEGER, SUNG HOON KIM, AND H. KAZEROONI: "《Control Scheme and Networked Control Architecture for the Berkeley Lower Extremity Exoskeleton (BLEEX)》", 《INTERNATIONAL CONFERENCE ON ROBOTICS AND AUTOMATION》 * |
| YANG LIU, YONGSHENG GAO, FEIYUN XIAO, JIE ZHAO: "Research on the Cable-pulley Underactuated Lower Limb Exoskeleton》", 《INTERNATIONAL CONFERENCE ON MECHATRONICS AND AUTOMATION> * |
| 宋立博，李劲松: "轮式移动机器人嵌入式自适应控制器设计与仿真", 《控制理论与应用》 * |
| 李醒，王晓峰: "《不确定机器人系统无模型自适应滑模控制方法》", 《南京理工大学学报》 * |
| 柴虎: "《下肢外骨骼机器人跟随系统的研究》", 《中国优秀硕士学位论文全文数据库（电子期刊）》 * |
| 王晓峰，李醒，王建辉: "《基于无模型自适应的外骨骼式上肢康复机器人主动交互训练控制方法》", 《自动化学报》 * |
| 王永，费燕琼，宋立博: "《模块化移动机器人的无模型自适应控制研究》", 《机电工程》 * |
Cited By (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110711114A (en) * | 2019-10-17 | 2020-01-21 | 河北工业大学 | Exoskeleton active disturbance rejection control method based on dynamic model |
| CN110716557A (en) * | 2019-11-18 | 2020-01-21 | 河北工业大学 | Machine parameter identification and contact force monitoring method based on priori dynamics knowledge |
| CN111403019A (en) * | 2020-03-20 | 2020-07-10 | 国家康复辅具研究中心秦皇岛研究院 | Ankle joint prosthesis model establishing method, model-free control method and verification method |
| CN111403019B (en) * | 2020-03-20 | 2023-08-29 | 国家康复辅具研究中心秦皇岛研究院 | Method for establishing ankle joint artificial limb model, model-free control method and verification method |
| CN111568432A (en) * | 2020-04-06 | 2020-08-25 | 沈阳工业大学 | Fall detection method of omnibearing walking training robot based on fuzzy inference |
| CN111568432B (en) * | 2020-04-06 | 2023-04-07 | 沈阳工业大学 | Fall detection method of omnibearing walking training robot based on fuzzy inference |
| CN111930135A (en) * | 2020-08-12 | 2020-11-13 | 深圳航天科技创新研究院 | Active power-assisted control method and device based on terrain judgment and exoskeleton robot |
| CN111930135B (en) * | 2020-08-12 | 2023-09-05 | 深圳航天科技创新研究院 | Active power-assisted control method and device based on terrain judgment and exoskeleton robot |
| CN112008726A (en) * | 2020-08-28 | 2020-12-01 | 浙江协力机器人技术有限公司 | Composite finite time control method based on exoskeleton robot actuator |
| CN112008726B (en) * | 2020-08-28 | 2021-08-27 | 浙江协力机器人技术有限公司 | Composite finite time control method based on exoskeleton robot actuator |
| CN113977582A (en) * | 2021-11-12 | 2022-01-28 | 杭州电子科技大学 | Lower limb exoskeleton stair climbing control method based on DMP |
Also Published As
| Publication number | Publication date |
|---|---|
| CN110327187B (en) | 2021-07-13 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110327187A (en) | A kind of band priori torque non-model control method of ectoskeleton | |
| Lin et al. | A leg configuration measurement system for full-body pose estimates in a hexapod robot | |
| CN106983589B (en) | A kind of active above-knee prosthesis TSM control method based on interference observer | |
| Liu et al. | Triaxial joint moment estimation using a wearable three-dimensional gait analysis system | |
| KR102336728B1 (en) | A system for analyzing gait | |
| CN108388887A (en) | Biped robot's Analytical Methods of Kinematics based on toddlerhood child's Gait extraction | |
| Bonnet et al. | Fast determination of the planar body segment inertial parameters using affordable sensors | |
| Febrer-Nafría et al. | Comparison of different optimal control formulations for generating dynamically consistent crutch walking simulations using a torque-driven model | |
| CN110193830B (en) | Ankle joint gait prediction method based on RBF neural network | |
| Li et al. | Reconstructing walking dynamics from two shank-mounted inertial measurement units | |
| Wu et al. | Gait phase prediction for lower limb exoskeleton robots | |
| CN109817048B (en) | Recognition system for role posture and walking direction in virtual simulation environment | |
| CN112472531A (en) | Gait smoothing algorithm of lower limb exoskeleton robot for medical rehabilitation and assisted walking | |
| Minh et al. | Model Predictive Control for Modeling and Simulation of Human Gait Motions | |
| CN110456902A (en) | It is mobile to control the skeleton pattern in computer system to track user | |
| Mistry et al. | Sit-to-stand task on a humanoid robot from human demonstration | |
| CN113283116B (en) | Multi-information fusion human motion analysis method and device | |
| CN110442947A (en) | Merge the lower limb robot dynamics emulation platform and emulation mode of equilibrium strategy | |
| Fang et al. | Modelling of the toe trajectory during normal gait using circle-fit approximation | |
| CN108013880A (en) | A kind of instantaneous aroused in interest measuring method of human elbow anterior flexion and rear stretching around instantaneous aroused in interest movement | |
| CN208591046U (en) | The detection device of unstable motion data | |
| CN114795604A (en) | Lower limb prosthesis coordination control method and system based on non-zero sum game | |
| IMBESI | Estimation of ground reaction forces with applications for ecological monitoring of joint loading: a combined musculoskeletal and optimization based proof of concept | |
| Stojićević et al. | A bipedal mechanical walker with balancing mechanism | |
| Matthews et al. | In-situ step size estimation using a kinetic model of human gait |
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: 20210713 |
|
| CF01 | Termination of patent right due to non-payment of annual fee |