Event-Triggered State Estimation for Discrete-Time Multidelayed Neural Networks With Stochastic Parameters and Incomplete Measurements
read more
Citations
The sector bound approach to quantized feedback control | NOVA. The University of Newcastle's Digital Repository
Synchronization Control for A Class of Discrete Time-Delay Complex Dynamical Networks: A Dynamic Event-Triggered Approach
A survey on sliding mode control for networked control systems
An overview of recent developments in Lyapunov–Krasovskii functionals and stability criteria for recurrent neural networks with time-varying delays
Neural-Network-Based Output-Feedback Control Under Round-Robin Scheduling Protocols
References
Neural computation of decisions in optimization problems
Simple 'neural' optimization networks: An A/D converter, signal decision circuit, and a linear programming circuit
Distributed Event-Triggered Control for Multi-Agent Systems
A Novel Connectionist System for Unconstrained Handwriting Recognition
The sector bound approach to quantized feedback control
Related Papers (5)
Distributed recursive filtering for stochastic systems under uniform quantizations and deception attacks through sensor networks
Frequently Asked Questions (11)
Q2. What future works have the authors mentioned in the paper "Event-triggered state estimation for discrete-time multi-delayed neural networks with stochastic parameters and incomplete measurements" ?
Future research topics include the extension of the results to the continuoustime delayed neural networks ( see e. g. [ 22 ], [ 28 ] ) with incomplete information. ΗTd ( k ) Qηd ( k ) − ηT ( k ) Pη ( k ) }. ( 29 ) By noting ( 3 ) and ( 14 ), the term containing ÃT ( k ) P Ã ( k ) can be computed as follows: E { ηT ( k ) HT2 ÃT ( k ) P Ã ( k ) H2η ( k ) } =E { ηT ( k ) HT2 [ Ã ( k ) Ã ( k ) ] T [ P11 P12 P21 P22 ] [ Ã ( k ) Ã ( k ) ] H2η ( k ) } =E { ηT ( k ) HT2 ( E ◦ ( P11 + P12 + P21 + P22 ) ) H2η ( k ) } =E { ηT ( k ) HT2 P̃H2η ( k ) }. ( 30 ) For the term w̄T ( k ) BTPBw̄ ( k ), the authors have E { w̄T ( k ) BTPBw̄ ( k ) } ≤λmax ( BTPB ) E { wT ( k ) w ( k ) + vT ( k ) v ( k ) } =ϑ. ( 31 ) By using the elementary inequality −2ab ≤ a2 + b2, it can be obtained that E { −2β3ηT ( k ) ATPH1GUη ( k ) } ≤ E { β3ηT ( k ) ATPAη ( k ) +β3η T ( k ) UTGTHT1 PH1GUη ( k ) }, ( 32 ) E { −2β2ηT ( k ) ATPH1Gs̄ ( k ) } ≤ E { β2ηT ( k ) ATPAη ( k ) +β2s̄ T ( k ) GTHT1 PH1Gs̄ ( k ) }, ( 33 ) E { −2β2ξT ( k ) GTHT1 PH1Gs̄ ( k ) } ≤ E { β2ξT ( k ) GTHT1 PH1Gξ ( k ) +β2s̄ T ( k ) GTHT1 PH1Gs̄ ( k ) }, ( 34 ) E { −2β3GT According to the definition of functional V ( k ), it can be obtained that V ( k ) ≤λmax ( P ) ∥η ( k ) ∥2 + n∑ i=1 n∑ j=1 λmax ( Qij ) k−1∑ l=k−τij ∥η ( l ) ∥2. ( 39 ) Introducing a scalar α > 1, the authors compute E { αk+1V ( k + 1 ) − αkV ( k ) } =E { αk+1 ( V ( k + 1 ) − V ( k ) ) + αk ( α− 1 ) V ( k ) } ≤αkϕ ( α ) E { ∥η ( k ) ∥2 } + αk+1 ( λ4δ + ϑ ) + αk n∑ i=1 n∑ j=1 φij ( α ) k−1∑ l=k−τij E { ∥η ( l ) ∥2 }. ( 40 ) where ϕ ( α ) and φij ( α ) are defined in ( 25 ).
Q3. what is the state variable of neuron i?
R is the state variable of neuron i; w0ij and w 1 ij are the interconnection strength and the delayed interconnection strength between neurons i and j, respectively; τij is a constant representing the delay from neuron i to j; gj(·) denotes the neuron activation function; ωi(k) is a zero mean Gaussian white-noise process; and bi is a deterministic constant while ai(k) is a random variable satisfying 0 < ai(k) < 1.Remark 1: The neural networks given by (1) is in nature a Hopfield neural network.
Q4. What is the deterministic constant of ai(k)?
In this paper, both the quantization effects and sensor saturations are taken into account and the network measurement model is given as follows [33]:y(k) =δ(α(k), 1)Cx(k) + δ(α(k), 2)s(Cx(k))+ δ(α(k), 3)q(Cx(k)) + v(k) (6)where y(k) ∈
Q5. What is the purpose of this paper?
0 < µ < 1, ν > 0, κ̄ > 0 such thatE{∥η(k)∥2} ≤ µkν + κ(k), and lim k→+∞ κ(k) = κ̄. (15)The aim of this paper is to design an event-triggered estimator with the form (10) for the multi-delayed neural networks (1) with incomplete measurements described by (6).
Q6. ?
ḠT (k)W̄T1 PW̄1Ḡ(k) + ξT (k)GTHT1 PH1Gξ(k) + ηT (k)HT2 ÃT (k)P Ã(k)H2η(k) + β21η T (k)HT2 C TGTHT1 PH1GCH2η(k) + β22 s̄ T (k)GTHT1 PH1Gs̄(k) + w̄T (k)BTPBw̄(k) + β23η T (k)UTGTHT1 PH1GUη(k) + δ̃α21 (k)η T (k)HT2 C TGTHT1 PH1GCH2η(k) + δ̃α22 (k)s̄ T (k)GTHT1 PH1Gs̄(k) + δ̃α23 (k)η T (k)UTGTHT1 PH1GUη(k) + 2ηT (k)ATPW̄0G(k) + 2ηT (k)ATPW̄1Ḡ(k) + 2ηT (k)ATPH1Gξ(k) − 2β1ηT (k)ATPH1GCH2η(k) − 2β2ηT (k)ATPH1Gs̄(k) − 2β3ηT (k)ATPH1GUη(k) + 2GT (k)W̄T0 PW̄1Ḡ(k) + 2GT (k)W̄T0 PH1Gξ(k) − 2β1GT (k)W̄T0 PH1GCH2η(k) − 2β2GT (k)W̄T0 PH1Gs̄(k) − 2β3GT (k)W̄T0 PH1GUη(k) + 2ḠT (k)W̄T1 PH1Gξ(k) − 2β1ḠT (k)W̄T1 PH1GCH2η(k) − 2β2ḠT (k)W̄T1 PH1Gs̄(k) − 2β3ḠT (k)W̄T1 PH1GUη(k) − 2β1ξT (k)GTHT1 PH1GCH2η(k) − 2β2ξT (k)GTHT1 PH1Gs̄(k) − 2β3ξT (k)GTHT1 PH1GUη(k) + 2β1β2η T (k)HT2 C TGTHT1 PH1Gs̄(k) + 2β1β3η T (k)HT2 C TGTHT1 PH1GUη(k) + 2β2β3s̄ T (k)GTHT1 PH1GUη(k) + 2δ̃α2 (k)δ̃ α 3 (k)s̄ T (k)GTHT1 PH1GUη(k) + 2δ̃α1 (k)δ̃ α 3 (k)η
Q7. what is the deterministic constant of ai(k)?
(2)The random variables ωi(k) and ai(k) have the following statistical propertiesE{ai(k)} = āi, E{ai(k)aj(k)} = ãij , E{ωi(k)ωj(k)} = qij ,(3)where āi, ãij and qij are known constants.
Q8. Considering EV (0)0(1 T 0 )?
there exists a scalar α0 > 1 such that ζ(α0) = 0. Then, it follows from (43) thatE{αT0 V (T )} − E{V (0)}≤α0(1− α T 0 )1− α0 (λ4δ + ϑ)+ 1α0 − 1 n∑ i=1 n∑ j=1 τijφij(α0)× (ατij0 − 1) max−τij≤l≤0 E{∥η(l)∥2}(45)ConsideringE{V
Q9. what is the 0 ?
By using the Schur complement lemma, it is easily known that Φ < 0 is equivalent toΨ̃ = Ξ̄11 0 Ξ13 0 Ξ̄15 Ξ̄16 Ξ̄17 0 0 ∗ Ξ22 Ξ̄23 0 Ξ̄25 0 0 Ξ̄28 0 ∗ ∗ Ξ33 0 0 0 0 0 Ξ̄39 ∗ ∗ ∗ Ξ44 0 0 0 0 0 ∗ ∗ ∗ ∗ Ξ55 0 0 0 0 ∗ ∗ ∗ ∗ ∗ Ξ66 0 0 0 ∗ ∗ ∗ ∗ ∗ ∗ Ξ77 0 0 ∗ ∗ ∗ ∗ ∗ ∗ ∗ Ξ88 0 ∗ ∗ ∗ ∗ ∗ ∗ ∗ ∗ Ξ99 < 0whereΞ̄11 =H T 2 P̃H2 − P + λ1M̄ + n∑ i=1 n∑ j=1 Qij− λ3HT2 CTKCH2, Ξ̄15 =ATP − β1HT2 CTGTHT1 P, Ξ̄17 = √ 5β3U TGTHT1 P,Ξ̄16 = [√ β2 + β3ATP √ β1(1− β1)HT2 CTGTHT1 P ] ,Ξ̄23 = [ −β2GTHT1 PW̄0 −β2GTHT1 PW̄1 0 ]T ,Ξ̄25 = [ PW̄0 PW̄1 PH1G ]T , Ξ̄39 = √ 3β2G THT1 P,Ξ̄28 = [ 0 0 √ β2 + β3PH1G ]T .Let’s now deal with the uncertainty induced by quantization effect.
Q10. What is the value of the error in the estimation system?
By letting the estimation error be e(k) = x(k) − x̂(k), it follows from (5) and (10) thate(k + 1) =Ã(k)x(k) + (1− β1)GCx(k) + (Ā−GC)e(k) +W0g̃(k) +Gξ(k)+
Q11. what is the definition of a functional V?
According to the definition of functional V (k), it can be obtained thatV (k) ≤λmax(P )∥η(k)∥2+ n∑i=1 n∑ j=1 λmax(Qij) k−1∑l=k−τij∥η(l)∥2. (39)Introducing a scalar α > 1, the authors computeE{αk+1V (k + 1)− αkV (k)} =E{αk+1(V (k + 1)− V (k)) + αk(α− 1)V (k)} ≤αkϕ(α)E{∥η(k)∥2}+ αk+1(λ4δ + ϑ)+ αk n∑i=1 n∑ j=1 φij(α) k−1∑l=k−τijE{∥η(l)∥2}.(40)where ϕ(α) and φij(α) are defined in (25).