A kind of uncatalyzed coking dissipation filtering method of nonlinear network networked control systems
Technical field
The present invention relates to nonlinear network networked control systems and dissipation filtering, more particularly to a kind of to have time delay and packet loss
Nonlinear network networked control systems uncatalyzed coking dissipation filtering method.
Background technology
Network control system (networked control are referred to as by the closed-loop control system that communication network is formed
Systems, is abbreviated NCSs), NCSs has the advantages that convenient for installation and maintenance, flexibility is high and is easy to reconstruct.However, communication network
Introducing result in system and there is problems with:1) network delay:Data in communication network transmission because network blockage or
The reasons such as external interference so that there is network delay in network control system;2) packet loss:In data transmission procedure because
The reason such as network blockage and resource contention can cause the problem of data-bag lost.Extraneous uncertain factor may result in simultaneously
Systematic function reduces even unstability.Therefore, make NCSs that there is fault-tolerant ability and preferable interference free performance is kept with very heavy
The theory significance wanted and more practical value.
For time delay present in NCSs and packet loss problem, many scholars and expert have done numerous studies.Horse leap etc.
In paper " robust dissipation filterings of unknown time-delay Discrete-time Nonlinear Systems ", time delay is have studied to the stability of a system and dissipation
The impact of performance.Lin Qiongbin etc. have studied packet loss pair in " having the dissipation fuzzy filter of many data packetloss nonlinear systems "
The impact of the stability of a system and dissipative performance, Zhang Peng etc. are at paper " the robust dissipation filtering devices of Linear uncertain time-delay systems "
In, the impact of Study system parameter uncertainty and time delay to the stability and dissipative performance of system.Above-mentioned dissipation filtering
The research of aspect has considered only time delay or packet loss, and is not subject to external interference in view of wave filter inherent parameters
Some changes can be also produced, but time delay and packet loss are simultaneous and wave filter inherent parameters in an actual situation
Can be interfered change, so seeming particularly significant using uncatalyzed coking filtering, it can make system immediate stability, Disturbance Rejection
More preferably, filtering estimation effect is more preferable for level.Li Xiuying etc. " has network system H of a step stochastic Time-Delay and many packet losses in paper∞
Wave filter is designed ", packet loss and time lag are studied to the stability of a system and H∞The impact of performance, and system model is augmented,
Higher-dimension wave filter is devised, so substantially increases the computation burden that the wave filter of design higher-dimension can bring, so design wave filter
When consider system and do not model state, can make system that there is stronger robustness, and to design full rank wave filter be to keep away
The higher-dimension wave filter estimation of augmented state is exempted from, computation burden and design cost can have been significantly reduced.
The content of the invention
For the problem that above-mentioned technology is present, the invention provides a kind of uncatalyzed coking dissipation filtering of network control system
Method.Exist in the case of time delay, packet loss and wave filter have Parameter Perturbation in view of network control system, devising non-
Fragile dissipation filtering device so that network control system remains to keep Stochastic stable, and strict dissipativity in these cases.
The technical solution adopted in the present invention is:A kind of uncatalyzed coking dissipation filtering side of nonlinear network networked control systems
Method, comprises the following steps:
1) set up nonlinear networked filtering error system model:
Wherein:x(k)∈RnIt is the state vector of system, xf(k)
∈RnIt is state estimation, y (k) ∈ RrIt is that the measurement that wave filter is received is exported, w (k) ∈ RpIt is external interference signals, f (k, x
(k)) meet Lipschitz condition Nonlinear Vector items | | f (k, x (k)) | |≤| | Wx (k) | |, w (k) ∈ l2[0, ∞) to disturb
Dynamic input, e (k)=z (k)-zfK () is filtering error, z (k) ∈ RmIt is to be estimated signal, zf(k)∈RmIt is the defeated of wave filter
Go out;
Afd=Af+ΔAf, Bfd=Bf+ΔBf, Cfd=Cf+ΔCf
Wherein:Af,Bf,CfIt is filter parameter matrix, A ∈ Rn×n, B ∈ Rn×p, C ∈ Rr×n, D ∈ Rr×p, L1∈Rm×n, L2
∈Rm×p, 0 and I is for null matrix and unit matrix, Afd=Af+ΔAf, Bfd=Bf+ΔBf, Cfd=Cf+ΔCf;ΔAf=H1F1(k)
E1, Δ Bf=H2F2(k)E2, Δ Cf=H3F3(k)E3For filter parameter perturbation matrices, Af∈Rn×n, Bf∈Rn×r, Cf∈Rm×nFor
Filter parameter matrix, H1∈Rn×d, H2∈Rn×h, H3∈Rm×a, E1∈Rd×n, E2∈Rh×r, E3∈Ra×n, F1(k)∈Rd×d, F2
(k)∈Rh×h, F3(k)∈Ra×a;
Definition:
Introduce:ξ (k)=(1- θ (k)) δ (k+1),
Then there is following statistical property (E represents mathematic expectaion):
Wherein:δkAnd θkIt is uncorrelated random variables,
By model:
Wherein:y(k)∈RrIt is measurement output;
Time delay probability of happening is
Drop probabilities are
The probability of acceptance is data on timey(k)∈RrIt is measurement output;
2) Lyapunov functions are constructed;
Wherein:P is positive definite symmetric matrices;
3) calculate uncatalyzed coking dissipation filtering device parameter matrix Af, Bf, CfWith system performance index γ, system Stochastic stable and
Uncatalyzed coking dissipative control device exist adequate condition be:
For following linear MATRIX INEQUALITIES:
Wherein:
Ψ3=diag {-ε1I,-ε1I,-ε2I,-ε2I,-ε3I,-ε3I,-ε4I,-ε4I}
Π3=diag {-Π ,-Π ,-Π ,-Π }
Π4=diag {-P2,-P2,-P2,-P2}
Wherein, W, V are nonsingular constant matrices, are met, WVT=I-XZ-1;
X∈Rn×n, Z ∈ Rn×n,P2∈R2r×2r, εi> 0 (i=1,2,3,4) is
For known variables, other variables are known, can be drawn according to systematic parameter or directly be given, using Matlab LMI works
Tool case is solved, if there is symmetric positive definite matrix X, Z, P2And matrixWith scalar ε i > 0 (i=1,2,3,
4), then networking filtering error system is Stochastic stable and has strict dissipativity, and uncatalyzed coking filter parameter matrix is γ=Σ (| | e (k) | |)/Σ (| | w (k) | |), and can be after
It is continuous to carry out step 4);If above-mentioned known variables are without solution, networking filtering error system is not Stochastic stable and is unsatisfactory for tight
Lattice dissipativeness, it is impossible to obtain the parameter matrix of uncatalyzed coking wave filter, cannot also carry out step 4);
4) calculate uncatalyzed coking H∞Filter parameter matrix Af, Bf, Cf, each matrix parameter is taken as:Q=-I, R=γ2I, S=
0, H∞The lower Optimal Disturbance Rejection of filtering compares γoptThe condition of optimization is:
Make e=γ2If following optimization problem is set up:
X=XT> 0, Z=ZT> 0, X-Z > 0,εi> 0 (i=1,2,3,4)
The minimal disturbances inhibiting rate of systemThe parameter matrix of uncatalyzed coking dissipation filtering device can also be optimised for simultaneously
Compared with prior art, the present invention has following Advantageous Effects:
1) present invention is directed to the nonlinear network networked control systems with time delay and packet loss, while considering filter parameter
Perturbation and the impact of external disturbance, establish networking filtering error system model, the corresponding stability of a system and dissipation filtering
Solution;
2) present invention considers the probability of happening of random loss and time delay, random loss and time delay and meets Bernoulli point
Cloth, more practical significance;
3) present invention take into account the perturbation of filter parameter, optimize system performance index so that control based on network system
System is with more preferable interference free performance;
4) present invention is filtered suitable for Dissipative, including H∞Filtering reduces the guarantor of the uncatalyzed coking filter design method
Keeping property.
5) system is considered during present invention design wave filter and do not model state, make system that there is stronger robustness, and
The wave filter of design is full rank, it is to avoid the higher-dimension wave filter of augmented state is estimated, can significantly reduce computation burden and set
Meter cost.
Description of the drawings
Accompanying drawing 1 is the flow chart of nonlinear network networked control systems uncatalyzed coking dissipation filtering method.
When accompanying drawing 2 has general uncatalyzed coking (Q, S, R) dissipation filtering device, variable z (k) to be estimated is estimated with whichResponse
Figure.
Accompanying drawing 3 has uncatalyzed coking H∞During wave filter, variable z (k) to be estimated is estimated with whichResponse diagram.
Specific embodiment
Below in conjunction with the accompanying drawings the specific embodiment of the present invention is described further.
Referring to the drawings 1, a kind of uncatalyzed coking H of nonlinear network networked control systems∞Fault tolerant control method, including following step
Suddenly:
Step 1:Set up networking filtering error system model
Consider following nonlinear network networked control systems
Wherein:x(k)∈RnIt is the state vector of system, z (k) ∈ RmIt is to be estimated signal,It is that measurement is exported, w
(k)∈RpIt is external interference signals, f (k, x (k)) meets Lipschitz condition Nonlinear Vector items | | f (k, x (k)) | |≤| |
Wx(k)||;w(k)∈l2[0, ∞), matrix A, B, C, D, L1,L2Respectively there is the known constant matrix of corresponding dimension.
Assume that sensor is that clock drives,Wave filter is transmitted through the network to after packing, due to communicate bandwidth and
The reasons such as network congestion, packet can occur inevitable time lag in the transmission and even lose, this phenomenon be it is random, one
As using the random sequence of Bernoulli distributions describing these situations.
When having random delay:
When having random loss:
Wherein:y(k)∈RPIt is that the measurement that wave filter is received is exported, θkIt is that the Bernoulli that value is 0 and 1 is distributed
Random sequence, meets
By the stochastic variable for introducing 2 Bernoulli distribution, by random delay and packet loss phenomenon with a model retouching
State:
Wherein:δkAnd θkIt is uncorrelated random variables,By model
(2) understand, time delay probability of happening is
Drop probabilities are
The probability of acceptance is data on time
Designing wave filter is:
WhereinIt is state estimation, zf(k)∈RqIt is the estimation of z (k), Afd,Bfd,CfdWith filter parameter
It is uncertain and meet following form:
Wherein:xf(k)∈RnIt is state estimation, zf(k)∈RmIt is the output of wave filter, Afd, Bfd, CfdIt is not true with parameter
It is qualitative and meet following form:
Wherein:Af,Bf,CfIt is filter parameter matrix, HiAnd Ei(i=1,2,3) it is known normal matrix;FiK () meets:
Fi(k)TFi(k)≤I, i=1,2,3,4 (6)
Definition filtering error is e (k)=z (k)-zf(k).In order to further obtain filtering error system, new change is introduced
Amount, order
ξ (k)=(1- θ (k)) δ (k+1) (7)
Then there is following statistical property (E represents mathematic expectaion):
Formula (7) is substituted into formula (2) to obtain
Definition:
θ (k) ξ (k)=ξ (k) (1- ξ (k)) δ (k+1)=0 is noticed, then by formula (1) and formula (3), filtering as follows can be obtained and missed
Difference system:
Wherein:
It is as follows that supply function E (w, z, T) can be measured in definition:
Wherein:Q∈Rm×m, R ∈ Rp×pFor known symmetrical matrix and Q < 0, S ∈ Rm×pFor known constant matrix.
Step 2:Construction Lyapunov functions
Wherein:P is symmetric positive definite matrix.
When w (k)=0,
Utilize:
Due to fT(k)f(k)≤xT(k)WTWx (k), accordingly, there exist τ > 0, τ xT(k)WTWx(k)-τfT(k)f(k)≥0
Wherein:
Γ2=[GT 0 0 0]T
M1,2=M1-M2,
N1,2=N1-N2
Step 3:Using Lyapunov Theory of Stability and LMI analysis method, control based on network system is obtained
The solution of adequate condition and filter parameter that system Stochastic stable and uncatalyzed coking dissipation filtering device are present, step are as follows:
Step 3.1:Based on the Lyapunov functions that step 2 is constructed, using Lyapunov Theory of Stability and linear matrix
Inequality analysis method, first determines whether the Stochastic stable and strict dissipativity of networking filtering error system, obtains networking filter
The adequate condition that wave error system Stochastic stable and dissipation filtering device are present.
Lemma is mended by Schur to obtain:Φ < 0 are equivalent to:
I.e.Wherein, μ is the minimal eigenvalue of-Φ.Thus can obtain,
So filtering error system (10) is Stochastic stable.
When w (k) ≠ 0, constructing Lyapunov functions in the same manner can obtain:
Due to:fT(k)f(k)≤xT(k)WTWx (k), therefore, deposit 0 τ x of τ >T(k)WTWx(k)-τfT(k)f(k)≥0
Wherein:
G2=[GT 0 0 0 0]T
As Λ < 0, sufficiently small a > 0 are certainly existed so that (Λ+adiag (0, I)) < 0, so:
Formula (15) k is sued for peace from 0 to T available:
Because filtering error system (10) is Stochastic stable, it is initial zero under conditions ofAndInstitute
With:To arbitrary T > 0 and all non-zeros w (k),Meet strict dissipativity.Mended using Schur and drawn
Reason, Λ < 0 are equivalent to
The adequate condition that networking filtering error system Stochastic stable shown in system (10) and dissipation filtering device are present is:When
During external disturbance w (k) ≠ 0, it is known that Q, S, R and Q < 0 and Afd,Bfd,Cfd, there is positive definite symmetric matrices P and formula (16) set up,
When the adequate condition of step 3.1 is set up, then execution step 3.2;If the adequate condition of step 3.1 is false, system is not
Stochastic stable and uncatalyzed coking dissipation filtering device just do not exist, it is impossible to execution step 3.2.
Step 3.2:The solution of filter parameter.
By P=diag { P1,P2Inequality (16) is substituted into, then to inequality (16) both sides premultiplication diag { Σ2,I,I,I,
Σ1,I,Σ1,I,Σ1,I,Σ1,I,I}T, diag { Σ are taken advantage of on the right side2,I,I,I,Σ1,I,Σ1,I,Σ1,I,Σ1, I, I } pass through again
Equivalence transformation can obtain formula (17):
Further can be calculated:
Wherein:
Ξc=[L1 L1Z-1 -CfVT],Ξd=L2
Uncatalyzed coking filter parameter matrix and Parameter Perturbation matrix can be tried to achieve by following formula:
Wherein:W, V are nonsingular constant matrices, are met:WVT=I-XZ-1。
Diag { I, Z, I, I, I, I, Z, I, Z, I, Z, I, Z, I, I, I, I, I } is taken advantage of to formula (18) premultiplication, the right side in conjunction with formula
(19) obtain Ψ1, then and filter joint parameter (4) and (5), by the indeterminate of formula with determine item and separate, utilization
Schur mends lemma and obtains formula (20).
Wherein:
Ψ3=diag {-ε1I,-ε1I,-ε2I,-ε2I,-ε3I,-ε3I,-ε4I,-ε4I}
Π3=diag {-Π ,-Π ,-Π ,-Π }
Π4=diag {-P2,-P2,-P2,-P2}
X∈Rn×n, Z ∈ Rn×n, P2∈P2r×2r,εi> 0 (i=1,2,3,4) is known variables, its
Its variable is known, can be drawn according to systematic parameter or directly be given, and is solved using Matlab LMI tool boxes,
If there is symmetric positive definite matrix X, Z, P2MatrixWith scalar εi(i=1,2,3,4), then networking is filtered > 0
Error system is Stochastic stable and has strict dissipativity, and uncatalyzed coking filter parameter matrix is And step 4 can be proceeded);If above-mentioned known variables are without solution, networking filtering
Error system is not Stochastic stable and is unsatisfactory for strict dissipativity, it is impossible to obtains the parameter matrix of uncatalyzed coking wave filter, can not yet
With carry out step 4);
Step 4:Q, S are worked as in consideration, the dissipation filtering problem of system, wherein H when R chooses different value∞Filtering can be considered as one
As dissipation filtering a kind of special case.If Dissipative filtering, then asked using γ=Σ (| | e (k) | |)/Σ (| | w (k) | |)
Go out corresponding system performance index γ;If the H of standard∞Filtering, i.e., each matrix parameter are taken as:Q=-I, R=γ2I,
S=0, provides H∞The lower Optimal Disturbance Rejection of filtering compares γoptThe condition of optimization is:
Make e=γ2If following optimization problem is set up:
X=XT> 0, Z=ZT> 0,εi> 0 (i=1,2,3,4)
The Optimal Disturbance Rejection ratio of systemThe parameter matrix of uncatalyzed coking dissipation filtering device can also be optimised for simultaneously
Embodiment:
Using a kind of uncatalyzed coking dissipation filtering method of network control system proposed by the present invention, in no external disturbance
In the case of i.e. w (k)=0 when, networking filtering error system is Stochastic stable.When there is external disturbance, system is also
Stochastic stable and with certain antijamming capability.Concrete methods of realizing is as follows:
Step 1:Consider following network control system:
Y (k)=[2-3 5] x (k)+2w (k)
Z (k)=[- 0.1 0.3-0.2] x (k)
Consideration is given below parameter:
H3=[1 0 0], E2=0.02
Here takeWhen there is Parameter Perturbation F1(k)=F2(k)=F3(k)=F4During (k)=I, it is assumed that
System disturbance input w (k)=1/k2, it is considered to work as Q, S, when R chooses different value, the dissipation filtering problem of system, wherein H∞Filtering
It is considered as a kind of special case of Dissipative filtering.
Step 2:Each parameter of Dissipative filtering is taken as Q=-0.9, S=0.5, R=12.According to formula (20), utilize
Matlab LMI tool boxes uncatalyzed coking filtering parameter, and obtained with γ=∑ (| | e (k) | |)/∑ (| | w (k) | |) respectively and be
Unite corresponding performance indications γ,
Cf=[- 0.0854 0.2958-0.2085]
γ=0.7331
However, making Δ Af, Δ Bf, Δ CfPerformance indications γ for drawing traditional wave filter when being zero are 0.7375, contrast
It was found that, performance indications γ of uncatalyzed coking wave filter are substantially little than traditional performance of filter index γ, this explanation institute of the present invention
Uncatalyzed coking wave filter has more preferable Disturbance Rejection performance compared with conventional filter.
Due to the time delay that there is uncertainty probability in network control system and packet loss, can obtain under different performance index
Dissipation filtering device such as table 1.
Performance indications under 1 different delay of table and packet loss
Note 2:Time delay probability is equal toThe probability of packet loss is equal to
As it can be seen from table 1 the probability of packet loss and time delay can be changed by changing the expectation of θ (k) and δ (k), with losing
The probability of bag and time delay increases, and performance indications can also become big, and the estimation effect of uncatalyzed coking dissipation filtering device can be deteriorated, and illustrate packet loss
The performance of system is had an impact with time delay.
Step 3:H∞Each parameter of filtering is taken as:Q=-I, R=γ2I, S=0.In the same manner, according to formula (21), utilize
Matlab LMI tool boxes obtain uncatalyzed coking filtering parameter, and obtain corresponding Optimal Disturbance Rejection than γ opt,
Cf=[0.1911-0.3255 0.0739]
γopt=3.2713
However, making Δ Af, Δ Bf, Δ CfPerformance indications γ for drawing traditional wave filter when being zero are 4.0021, contrast
It was found that, the Optimal Disturbance Rejection of uncatalyzed coking wave filter compares γoptThan traditional performance of filter index γoptLittle, this explanation is originally
Invention uncatalyzed coking wave filter used has more preferable antijamming capability compared with conventional filter.
Due to the time delay that there is uncertainty probability in network control system and packet loss, different Optimal Disturbance Rejections can be obtained
Uncatalyzed coking H than under∞Wave filter such as table 2.
Performance indications under 2 different delay of table and packet loss
From table 2 it can be seen that the probability of packet loss and time delay can be changed by changing the expectation of θ (k) and δ (k), with losing
The probability of bag and time delay increases, and Optimal Disturbance Rejection can become big than also, and the antijamming capability of system can be deteriorated, uncatalyzed coking H∞Filtering
The estimation effect of device can be deteriorated, and illustrate that packet loss and time delay are had an impact to the performance of system.
Step 4:Using in step 2 and 3When Matlab LMI tool boxes solve result, use
Matlab simulates corresponding variable z (k) to be estimated of network control system and estimates z with whichfThe response of (k), such as accompanying drawing 2 and attached
Shown in Fig. 3.
By accompanying drawing 2 and accompanying drawing 3 as can be seen that variable z (k) to be estimated of system estimates z with whichfK () is bounded stability,
Illustrate the design of uncatalyzed coking dissipation filtering device and uncatalyzed coking H when there is certainty time delay and packet loss in a network environment∞Wave filter
Design is effective.
It is more than presently preferred embodiments of the present invention, not makees any pro forma restriction, every foundation to the present invention
The technical spirit of the present invention belongs to inventive technique to any simple modification made for any of the above embodiments, equivalent variations and modification
In the range of scheme.