CN112947066B - Manipulator improved finite time inversion control method - Google Patents
Manipulator improved finite time inversion control method Download PDFInfo
- Publication number
- CN112947066B CN112947066B CN202110105530.2A CN202110105530A CN112947066B CN 112947066 B CN112947066 B CN 112947066B CN 202110105530 A CN202110105530 A CN 202110105530A CN 112947066 B CN112947066 B CN 112947066B
- Authority
- CN
- China
- Prior art keywords
- layer
- manipulator
- neural network
- input
- output
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Images
Classifications
-
- G—PHYSICS
- G05—CONTROLLING; REGULATING
- G05B—CONTROL OR REGULATING SYSTEMS IN GENERAL; FUNCTIONAL ELEMENTS OF SUCH SYSTEMS; MONITORING OR TESTING ARRANGEMENTS FOR SUCH SYSTEMS OR ELEMENTS
- G05B13/00—Adaptive control systems, i.e. systems automatically adjusting themselves to have a performance which is optimum according to some preassigned criterion
- G05B13/02—Adaptive control systems, i.e. systems automatically adjusting themselves to have a performance which is optimum according to some preassigned criterion electric
- G05B13/04—Adaptive control systems, i.e. systems automatically adjusting themselves to have a performance which is optimum according to some preassigned criterion electric involving the use of models or simulators
- G05B13/042—Adaptive control systems, i.e. systems automatically adjusting themselves to have a performance which is optimum according to some preassigned criterion electric involving the use of models or simulators in which a parameter or coefficient is automatically adjusted to optimise the performance
Abstract
The invention discloses a manipulator improved finite time reversal control method, which comprises the following steps: s1, building a manipulator power model; s2, designing a manipulator finite time reversal controller; and S3, inverting the unknown state in the controller by using fuzzy neural network approximation. The control method of the invention utilizes the finite time inversion method to design the manipulator controller, so that the position of each joint of the manipulator reaches the expected target position, and adopts the structure and parameter full-regulation fuzzy neural network structure to design the self-adaptive fuzzy neural network control system, thereby overcoming the defect that the finite time inversion control strategy needs the accurate information of the system and further improving the robustness of the system.
Description
Technical Field
The invention relates to a manipulator control technology, in particular to an improved finite time reversal control method for a manipulator.
Background
The manipulator is a mechanical device which has the action function similar to that of a human arm, can grab and place objects in space or perform other operations, can replace heavy labor of people to realize mechanization and automation of production, can operate in a harmful environment to protect personal safety, and is widely applied to departments of mechanical manufacture, metallurgy, electronics, light industry, atomic energy and the like.
Due to the fact that factors such as uncertainty of joint parameters of the manipulator exist, model uncertainty exists in a manipulator dynamic model, and the manipulator control accuracy based on model control is affected. At present, aiming at the problem that uncertainty factors such as model uncertainty and external interference influence the control precision of a manipulator joint, the uncertainty factors are mostly approximated or compensated through an uncertainty estimator, and the adaptive law of parameters in the estimator is determined through stability analysis. A neural network and a fuzzy system with universal approximation capability are widely applied to uncertainty approximation compensation, but the controller for manipulator control needs to estimate system uncertainty upper bound information in advance, so that certain limitations exist.
Disclosure of Invention
The invention aims to overcome the defects of the prior art and provides a manipulator improved finite time reversal control method.
In order to achieve the purpose, the invention adopts the technical scheme that: a manipulator improved finite time reversal control method comprises the following steps:
step 1, building a mechanical arm power model
Consider a rigid manipulator with n joints, whose kinetic equation is:
wherein the ratio of q,respectively representing the position, speed and acceleration vector of each joint of the manipulator, M (q) epsilon Rn×nIn order to include an inertia matrix of the additional mass,g (q) e R as a coupling matrix of centrifugal force and Coriolis forcen×1In the form of a gravity matrix vector,for friction torque, τd∈Rn×1For unknown applied interference, T is equal to Rn×1N is the degree of freedom of the joint;
step 2, designing a manipulator finite time reversal controller
2.1 define the tracking error as
e1=q-qd (2)
Wherein q isdIs the desired position of the joint;
derived from (2)
Selecting a virtual control quantity
Wherein c is1Is a non-zero normal number;
taking the Lyapunov function as
2.1 e2can be expressed as
Then designing a nonsingular terminal sliding mode surface as
Wherein λ is1>0 is a constant, p1,p2Is odd number, 1<p2/p1<2,
Further, we can get
The finite time inversion controller can be designed as
τBSC=u1+u2 (10)
Wherein, tauM≥||τd||;
S3, aiming at the formula (10), the fuzzy neural network is used for approaching an unknown state in the inversion controller, and the input and output relation of the fuzzy neural network is defined as follows:
lk(k=1,...,Ny) The k-th output of the rule layer is represented,representing the connection weight matrix between the fuzzification layer and the regular layer, here taken as the unit vector, NyIs the total number of rule layers.
In a specific implementation mode, the fuzzy neural network consists of an input layer, a fuzzification layer, a fuzzy inference layer and an output layer, wherein the network input is the tracking deviation e1Output as control force tauFNNThe signal propagation and the function of each layer in the fuzzy neural network are represented as follows:
a first layer: input layer
Each node of the layer is directly connected with each component of the input quantity, and the input quantity is transmitted to the next layer;
a second layer: obscuration layer
A gaussian function is used as the membership function,representing the tracking offset vector e1The elements (A) and (B) in (B),andthe centre vector and the base width of the membership function of the ith input variable, the jth fuzzy set, respectively, i.e.
Easy to calculate, using NpiRepresenting individual numbers of membership functions, and we define the sets of adaptive parameter vectors b and c representing all the basis width and center vectors of Gaussian membership functions, respectively, i.e. WhereinRepresenting the total number of membership functions;
and a third layer: rule layer
The layer adopts a fuzzy inference mechanism, and the output of each node is the product of all input signals of the node, namely
In the formula Ik(k=1,...,Ny) The k-th output of the regular layer is represented,representing a blurred layer andthe connection weight matrix between regular layers, here taken as a unit vector, NyIs the total number of rules;
a fourth layer: output layer
The nodes of the layer represent output variables, each node yo (o 1.., N)o) The output of (2) being the sum of all input signals of the node, i.e.
Further, the input-output relationship of the fuzzy neural network is defined as follows:
as a specific embodiment, the fuzzy neural network includes two self-regulation strategies of data learning and data deletion, specifically:
data learning strategy
The fuzzy rule is determined step by step based on the condition thatWhere ψ is a spherical potential energy representing the novelty of the input data, given by the following equation:
when a new fuzzy rule needs to be added, its parameters are initialized,
where k is the overlap parameter of a pre-given fuzzy rule,
data deletion policy
According to the universal approximation theory, there is an optimal control forceSatisfy the requirement of
Where ε is the minimum reconstruction error vector, W*,b*And c*Optimal parameters of W, b and c respectively;
the output control force of the fuzzy neural network is assumed to be of the following form:
wherein the content of the first and second substances,andare each W*,b*And c*Is determined by the estimated value of (c),
defining an approximation error:
using Taylor series expansion, one can obtain
Wherein the content of the first and second substances,b*and c*Is the optimum value for b and c,andis b*And c*Estimated value of, OnvIs a high-order term of the signal,
then (25) into (24) can be obtained
further, the formula (9) can be rewritten as
Wherein, the first and the second end of the pipe are connected with each other,d=-E-τdand d is the total uncertainty,
the self-adaptive law of the weight, the base width and the central vector of the designed fuzzy neural network can be designed as follows:
wherein S isiIs an element of S, betaiIs an element in beta, sigmaω,σb,σcIs a normal number, and is,is thatIs determined by the estimated value of (c),is omegaiIs set to the optimum value of (a) or (b),
defining a lyapunov function as
Derivative and substitute (27) into
Bringing (28) - (30) into (32) to obtain
If τ is satisfiedMMore than or equal to d |, the designed manipulator improves finite time inversionThe control method is stable.
Due to the application of the technical scheme, compared with the prior art, the invention has the following advantages: according to the manipulator-improved finite-time inversion control method, the fuzzy neural network structure is fully adjusted through structure and parameters, the adaptive fuzzy neural network control system is designed to approach the finite-time inversion controller, and the limitation that an uncertain function needs to be predicted and the upper bound of interference is relaxed, so that the defect that a finite-time inversion control strategy needs system accurate information is overcome, and the system robustness is further improved.
Drawings
Fig. 1 is a block diagram of a manipulator-improved finite-time inversion control method according to the present invention.
Detailed Description
The technical solution of the present invention is further described with reference to the accompanying drawings and specific embodiments.
A manipulator improved finite time reversal control method comprises the following steps:
firstly, building a mechanical hand power model
Consider a rigid manipulator with n joints, whose kinetic equation is:
wherein the ratio of q,the position, speed and acceleration vector of each joint of the manipulator, M (q) epsilon Rn×nIn order to include an inertial matrix of the additional mass,g (q) epsilon R, which is a coupling matrix of centrifugal force and Coriolis forcen×1In the form of a gravity matrix vector,for friction torque, τd∈Rn×1For unknown applied interference, T is equal to Rn×1N is the degree of freedom of the joint;
(II) finite time reversal controller of design manipulator
The control objective of the robot position tracking control system is to design an appropriate control input tau so that the position q of each joint of the robot tracks the desired position qdThe design steps are as follows:
2.1 define the tracking error as
e1=q-qd (2)
Derived from (2)
Selecting virtual control quantities
Wherein c is1Is a non-zero normal number;
taking the Lyapunov function as
If e2When the value is equal to 0, thenThe stability requirement is met, so the design needs to be continued;
2.1 e2of (2)The number can be expressed as
Then designing a nonsingular terminal sliding mode surface as
Wherein λ is1>0 is a constant, p1,p2Is odd number, 1<p2/p1<2,
Further, we can get
The finite time inversion controller can be designed as
τBSC=u1+u2 (10)
Wherein, tauM≥||τd||;
Constructing a Lyapunov function as
To V2Performing a time derivative may result in:
by substituting formula (9) for formula (14)
From the above formula, the designed manipulator finite time reversal controller is stable. However, the above-mentioned controllers require detailed system information and an upper uncertainty bound τ in practical systemsMIt is difficult to determine, and these problems indicate that the above controller is difficult to implement in practical applications. Therefore, to overcome these problems, an improved finite time inversion controller has been proposed.
(III) design mechanical arm improved finite time inversion controller
The designed manipulator-improved finite-time inversion controller adopts a fuzzy neural network to approximate the inversion controller (10), thereby overcoming the defect that the inversion controller needs detailed system information. The fuzzy neural network comprises input layer, fuzzy inference layer and output layer, the network input is tracking deviation e1Output as control force tauFNN. The signal propagation and the function of the layers in the fuzzy neural network are represented as follows:
a first layer: input layer
Each node of the layer is directly connected with each component of the input quantity, and the input quantity is transmitted to the next layer;
a second layer: blurring layer
A gaussian function is used as the membership function,representing the tracking offset vector e1The elements (A) and (B) in (B),andthe affiliation of the jth input variable and the jth fuzzy set respectivelyThe center vector and the base width of the generic function, i.e.
Easy to calculate, using NpiRepresent individual numbers of membership functions and we define the adaptive parameter vectors b and c to represent the set of all the base width and center vectors of Gaussian membership functions, respectively, i.e. WhereinRepresenting the total number of membership functions;
and a third layer: rule layer
The layer adopts a fuzzy inference mechanism, and the output of each node is the product of all input signals of the node, namely
In the formula Ik(k=1,...,Ny) The k-th output of the regular layer is represented,representing the connection weight matrix between the fuzzification layer and the regular layer, here taken as the unit vector, NyIs the total number of rules;
a fourth layer: output layer
The nodes of the layer represent output variables, each node yo(o=1,...,No) The output of (2) being the sum of all input signals of the node, i.e.
Further, the input-output relationship of the fuzzy neural network is defined as follows:
y=[y1 y2…yNo]=τFNN=Wl
particularly, the meta-cognition fuzzy neural model considers two self-regulation strategies of data learning and data deletion, and is beneficial to efficiently executing a real-time control task.
First is the data learning strategy. The data learning process of the meta-cognitive fuzzy neural network involves the online evolution and parameter updating of the rules closest to the current input data.
The fuzzy rule is determined step by step according to the following conditions, namelyAnd psi<EsWhere ψ is a spherical potential energy, representing the novelty of the input data, given by the following equation:
when a new fuzzy rule needs to be added, its parameters are initialized,
wherein κ is preThe overlapping parameters of the fuzzy rule are given firstWhen the time is needed, the rule parameters are adjusted,
in the learning process, the contribution degree of a certain rule to the output may be reduced. In this case, insignificant rules should be removed from the rule base to avoid overcomputing. The contribution of the q-th rule is given by:
wherein, the first and the second end of the pipe are connected with each other,n represents the dimension of the input.
If the contribution degree of the q rule to the input is lower than the threshold value EpThen the rule is deleted.
Data deletion policy
Specifically, when the current tracking error is close to the error of the last iterative calculation process of the neural network, network parameters do not need to be updated, excessive learning is avoided, and the calculation burden is reduced.
According to the universal approximation theory, there is an optimal control forceSatisfy the requirements of
Where ε is the minimum reconstruction error vector, W*,b*And c*The optimal parameters of W, b and c are respectively;
the output control force of the fuzzy neural network is assumed to be of the following form:
wherein, the first and the second end of the pipe are connected with each other,andare each W*,b*And c*Is determined by the estimated value of (c),
defining an approximation error:
using Taylor series expansion, one can obtain
Wherein, the first and the second end of the pipe are connected with each other,b*and c*Is the optimum value for b and c,andis b*And c*Estimated value of, OnvIs a high-order term of the magnetic field,
then (25) into (24) can be obtained
further, the formula (9) can be rewritten as
Wherein the content of the first and second substances,d=-E-τdand d is a lumped uncertainty, the sum of the uncertainty,
the self-adaptive law of the weight, the base width and the central vector of the designed fuzzy neural network can be designed as follows:
wherein S isiIs an element of S, betaiIs an element of beta, σω,σb,σcIs a normal number, and is,is thatIs determined by the estimated value of (c),is omegaiIs set to the optimum value of (a) or (b),
defining a lyapunov function as
Derivative and substitute (27) into
Bringing (28) - (30) into (32) to obtain
If τ is satisfiedMAnd the designed manipulator improved finite time inversion control method is stable.
The above embodiments are merely illustrative of the technical ideas and features of the present invention, and the purpose thereof is to enable those skilled in the art to understand the contents of the present invention and implement the present invention, and not to limit the protection scope of the present invention. All equivalent changes and modifications made according to the spirit of the present invention should be covered within the protection scope of the present invention.
Claims (1)
1. A manipulator improved finite time reversal control method is characterized by comprising the following steps:
step 1, building a mechanical arm power model
Consider a rigid manipulator with n joints, whose kinetic equation is:
wherein the ratio of q,the position, speed and acceleration vector of each joint of the manipulator, M (q) epsilon Rn×nIn order to include an inertial matrix of the additional mass,g (q) epsilon R, which is a coupling matrix of centrifugal force and Coriolis forcen×1Is a vector of a gravity matrix and is,for friction torque, τd∈Rn×1For unknown applied interference, tau epsilon Rn×1N is the degree of freedom of the joint;
step 2, designing a manipulator finite time reversal controller
2.1 define the tracking offset as
e1=q-qd (2)
Wherein q isdIs the desired position of the joint;
derived from formula (2)
Selecting virtual control quantities
Wherein c is1Is a non-zero normal constant;
taking the Lyapunov function as
If e2When the value is equal to 0, thenThe stability requirement is met, and the design is continued;
2.1 e2is expressed as
Then designing a nonsingular terminal sliding mode surface as
Wherein λ is1>0 is a constant, p1,p2Is odd number, 1<p2/p1<2,
Further obtain
The finite time inversion controller is designed as
τBSC=u1+u2 (10)
Wherein, tauM≥||τd||;
S3, for equation (10), defining the input-output relationship of the fuzzy neural network by approximating the unknown state in the inversion controller with the fuzzy neural network, the specific process is as follows:
the fuzzy neural network consists of an input layer, a fuzzy inference layer and an output layer, wherein the network input is tracking deviation e1Output as control force tauFNNThe signal propagation and the function of each layer in the fuzzy neural network are represented as follows:
a first layer: input layer
Each node of the layer is directly connected with each component of the input quantity, and the input quantity is transmitted to the next layer;
a second layer: obscuration layer
Using Gaussian-like functions as membership functions Representative tracking offset e1The element of (1), i ═ 1.·, n;andthe central vector and the base width of the membership function of the ith input variable jth fuzzy set are respectively, wherein i is 1piI.e. by
Easy to calculate, using NpiRepresenting individual numbers of membership functions and defining adaptive parameter vectors b and c representing the set of all base width and center vectors of Gaussian membership functions, respectively, i.e. WhereinRepresenting the total number of membership functions;
and a third layer: rule layer
The layer adopts a fuzzy inference mechanism, and the output of each node is the product of all input signals of the node, namely
In the formula IkDenotes the kth output of the regular layer, where k 1y,Representing the connection weight matrix between the fuzzification layer and the regular layer, here taken as the unit vector, NyIs the total number of rules;
a fourth layer: output layer
The nodes of the layer represent output variables, each node yoThe output of (a) is the sum of all input signals of the node, o 1oI.e. by
Further, the input and output relationship of the fuzzy neural network is defined as follows:
in addition, the fuzzy neural network comprises two self-regulation strategies of data learning and data deletion, specifically:
data learning strategy
The fuzzy rule is determined step by step based on the condition thatAnd psi<EsWhere ψ is a spherical potential energy, representing the novelty of the input data, given by the following equation:
when a new fuzzy rule needs to be added, its parameters are initialized,
wherein κ*Is an overlap parameter of a pre-given fuzzy rule,
data deletion policy
According to the universal approximation theory, there is an optimal control forceSatisfy the requirements of
Where ε is the minimum reconstruction error vector, W*,b*,c*The optimal parameters of W, b and c are respectively;
the output control force of the fuzzy neural network is assumed to be of the following form:
wherein the content of the first and second substances,are respectively W*,b*,c*Is determined by the estimated value of (c),
defining an approximation error:
using Taylor series expansion to obtain
Wherein the content of the first and second substances,b*and c*Is the optimum value of b and c,andis b*And c*Estimated value of (a), OnvIs a high-order term of the signal,
then (25) is substituted into (24) to obtain
further, formula (9) is rewritten as
Wherein the content of the first and second substances,d=-E-τdand d is a lumped uncertainty, the sum of the uncertainty,
the adaptive law design of the weight, the base width and the central vector of the fuzzy neural network is as follows:
wherein S isiIs an element of S, betaiIs an element in beta, sigmaω,σb,σcIs a normal number, and is,is thatIs determined by the estimated value of (c),is omegaiIs determined to be the optimum value of (c),
defining the lyapunov function as
by applying the derivation of formula (31) and the substitution of formula (27)
Bringing formula (28) -formula (30) into formula (32) to obtain
If τ is satisfiedMAnd if the absolute value is more than or equal to | d |, the designed manipulator improved finite time inversion control method is stable.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202110105530.2A CN112947066B (en) | 2021-01-26 | 2021-01-26 | Manipulator improved finite time inversion control method |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202110105530.2A CN112947066B (en) | 2021-01-26 | 2021-01-26 | Manipulator improved finite time inversion control method |
Publications (2)
Publication Number | Publication Date |
---|---|
CN112947066A CN112947066A (en) | 2021-06-11 |
CN112947066B true CN112947066B (en) | 2022-07-22 |
Family
ID=76237211
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202110105530.2A Active CN112947066B (en) | 2021-01-26 | 2021-01-26 | Manipulator improved finite time inversion control method |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN112947066B (en) |
Families Citing this family (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN114844430B (en) * | 2022-06-06 | 2024-03-01 | 苏州泰科贝尔直驱电机有限公司 | Fuzzy neural network control method for magnetic suspension switch reluctance motor |
Family Cites Families (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN108284442B (en) * | 2017-01-24 | 2021-01-26 | 中国北方车辆研究所 | Mechanical arm flexible joint control method based on fuzzy neural network |
CN108227504B (en) * | 2018-01-25 | 2020-10-27 | 河海大学常州校区 | Micro-gyroscope fractional order self-adaptive fuzzy neural inversion terminal sliding mode control method |
CN109103884A (en) * | 2018-09-18 | 2018-12-28 | 河海大学常州校区 | Active Power Filter-APF back stepping control method based on metacognition fuzzy neural network |
CN109103885A (en) * | 2018-09-18 | 2018-12-28 | 河海大学常州校区 | Active Power Filter-APF metacognition fuzzy Neural Network Control Method |
CN108828961B (en) * | 2018-09-18 | 2021-07-30 | 河海大学常州校区 | Sliding mode control method of active power filter based on element cognition fuzzy neural network |
CN110877333B (en) * | 2019-04-12 | 2023-01-17 | 国网宁夏电力有限公司电力科学研究院 | Flexible joint mechanical arm control method |
-
2021
- 2021-01-26 CN CN202110105530.2A patent/CN112947066B/en active Active
Also Published As
Publication number | Publication date |
---|---|
CN112947066A (en) | 2021-06-11 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN108614419B (en) | Adaptive neural network control method of arc micro-electro-mechanical system | |
Abouaïssa et al. | On the control of robot manipulator: A model-free approach | |
CN111716360B (en) | Fuzzy logic-based flexible joint mechanical arm sampling control method and device | |
CN110877333A (en) | Flexible joint mechanical arm control method | |
Qi et al. | Stable indirect adaptive control based on discrete-time T–S fuzzy model | |
Hazem et al. | A comparative study of the joint neuro-fuzzy friction models for a triple link rotary inverted pendulum | |
Pan et al. | Online data‐driven composite adaptive backstepping control with exact differentiators | |
CN111958606A (en) | Distributed robust tracking control method applied to multi-degree-of-freedom mechanical arm | |
Perrusquía et al. | Robot position/force control in unknown environment using hybrid reinforcement learning | |
CN112947066B (en) | Manipulator improved finite time inversion control method | |
CN112051734A (en) | Wheeled mobile robot event triggering tracking control method based on deterministic learning | |
Yu et al. | Global fast terminal sliding mode control for robotic manipulators | |
Li et al. | Event-triggered-based cooperative game optimal tracking control for modular robot manipulator with constrained input | |
Ginting et al. | Attitude control of a quadrotor with fuzzy logic controller on SO (3) | |
CN112338914B (en) | Single-link manipulator fuzzy control algorithm based on random system under output limitation and input hysteresis | |
Montoya‐Cháirez et al. | Combined adaptive neural network and regressor‐based trajectory tracking control of flexible joint robots | |
Li et al. | A Novel Adaptive Sliding Mode Control of Robot Manipulator Based on RBF Neural Network and Exponential Convergence Observer | |
CN112757298B (en) | Intelligent inversion control method for manipulator | |
Nguyen | Non-Negative Adaptive Mechanism-Based Sliding Mode Control for Parallel Manipulators with Uncertainties. | |
Alavandar et al. | New hybrid adaptive neuro-fuzzy algorithms for manipulator control with uncertainties–Comparative study | |
Ye et al. | Event-based adaptive fuzzy asymptotic tracking control of quadrotor unmanned aerial vehicle with obstacle avoidance | |
He et al. | A Semiparametric Model-Based Friction Compensation Method for Multijoint Industrial Robot | |
Haqshenas M et al. | A finite‐time adaptive Taylor series tracking control of electrically‐driven wheeled mobile robots | |
Chen et al. | A fuzzy compensator for uncertainty of industrial robots | |
Qian et al. | RBF-neural network adaptive control of mobile manipulator |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant | ||
GR01 | Patent grant |